November 4, 2025 • 53 mins
The source provides an extensive overview of the global economic and social impacts of Artificial Intelligence, primarily drawing from the 2025 UNCTAD Technology and Innovation Report. It highlights the dual nature of AI, projecting a massive $4.8 trillion market by 2033 while warning that the technology could disrupt up to 40% of jobs worldwide due to automation. The analysis explains how AI uniquely targets cognitive, high-skill work in advanced economies and routine tasks in emerging markets, threatening to exacerbate inequality by favoring capital over labor and concentrating benefits in nations like the U.S. and China. Crucially, the episode examines mechanisms of job displacement and creation, using case studies from India, Kenya, and Germany to illustrate real-world disruption, concluding with a policy roadmap for inclusive governance and widespread reskilling.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Welcome to the deep dive. This is where we take
those really dense global reports, the ones packed with data,
and try to well unpack them into clear, actionable knowledge
for you.
Speaker 2 (00:11):
And today we're tackling a big one, a document that
honestly has huge implications for pretty much everyone, every economy,
every career path, and definitely the future of global equity.
Speaker 1 (00:22):
We're talking about the UNC Tech Technology and Innovation Report
twenty twenty five. Came out early this year April third.
I believe that's right.
Speaker 2 (00:29):
And you know this report, it's more than just a forecast.
It really feels like an urgent map for the next decade.
It lays out this picture of artificial intelligence AI.
Speaker 1 (00:39):
It's everywhere now exactly.
Speaker 2 (00:40):
And the picture is well, it's complicated, it's dazzling. On
one side, you know this incredible engine for productivity, potentially
explosive growth.
Speaker 3 (00:48):
Well there's a butt, a big butt.
Speaker 2 (00:50):
It also presents AI as potentially the biggest driver for
widening global division we've seen since maybe the industrial age.
Speaker 3 (00:57):
It's really starts.
Speaker 1 (00:58):
Okay, let's ground this immediately. The scale, because the numbers
involved here are just almost hard to grasp.
Speaker 3 (01:04):
They are astronomical.
Speaker 2 (01:06):
We're looking at the global AI market, the UNC TAD
projection it's expected to swell to get this four point
eight trillion dollars by twenty thirty three.
Speaker 1 (01:16):
Four point eight trillion. Just say that again, four.
Speaker 3 (01:19):
Point eight trillion dollars.
Speaker 1 (01:21):
Wow, okay, context, help us understand what four point eight
trillion dollars actually looks like?
Speaker 2 (01:25):
Right? Think about Germany's entire economy right now. It's GDP,
Europe's biggest yuconic, that's the one. The AI market alone
is projected to be roughly that size in less than
ten years.
Speaker 1 (01:36):
So this isn't just another tech sector. It's becoming a
global economic force in its own.
Speaker 2 (01:40):
Right, absolutely a major economic superpower, purely based on its
market value and how it's spreading into well everything, every industry.
Speaker 1 (01:47):
Now, what's really driving that value? According to the report,
is it just hype or specific applications?
Speaker 3 (01:53):
No, it's specific.
Speaker 2 (01:54):
The UNC TAD analysis points to massive productivity leaps, primarily
in three core areas, which are healthcare diagnostics think AI
reading scans better than humans. Then there's autonomous logistics, self
driving trucks, automated warehouses, and finally, personalized education delivered at scale.
Speaker 1 (02:12):
So it's not just the flashy consumer stuff we see,
not at all.
Speaker 2 (02:16):
It's about automating these fundamental economic functions. The promise is
genuinely transformative.
Speaker 1 (02:21):
Okay, the promise is huge. But you mentioned the tension,
the dual warning at the heart of this report.
Speaker 3 (02:26):
Let's get into that exactly.
Speaker 2 (02:27):
So while AI offers this unprecedented economic boom, these huge
productivity gains, the unc TAD report is very clear, very stark.
It warns AI could.
Speaker 1 (02:40):
Disrupt disrupt how many up.
Speaker 2 (02:42):
To forty percent of jobs worldwide through automation, through job displacement.
Speaker 1 (02:48):
That's the headline number, isn't it.
Speaker 2 (02:49):
It jumps out at you, it's designed to, and it's
a shocking figure. But the real story, the complexity that
the report really digs into, is the geopolitical angle. Because
a sorearm wave of time technology, unlike maybe previous ones,
it structurally favors and this is their phrase, capital over labor.
Speaker 1 (03:05):
Capital over labor. Okay, unpack that for us. What does
that mean in practice?
Speaker 2 (03:09):
Well, think about how developing economies have grown over the
last say, four decades. What was their main competitive advantage?
Speaker 1 (03:16):
Low cost labor, mostly manufacturing, call centers, that kind.
Speaker 2 (03:19):
Of thing exactly, that was the primary way millions were
lifted out of poverty. Countries built their development ladder.
Speaker 3 (03:25):
On that edge.
Speaker 2 (03:26):
But ai AI systematically erodes that.
Speaker 1 (03:29):
Edge because it replaces the chief human labor with.
Speaker 2 (03:32):
With computation with capital investment in servers and software, zero
wage computational capital effectively, so.
Speaker 1 (03:39):
The whole development model is being flipped on.
Speaker 3 (03:40):
Its head fundamentally.
Speaker 2 (03:42):
Instead of needing a large, relatively cheap workforce, countries now
need immense investment in things like compute power, specialized chips,
robust digital infrastructure just to stay in the game.
Speaker 1 (03:55):
Which most developing countries just.
Speaker 2 (03:56):
Don't have precisely. And that disparity, that sift and what's
needed is why the unc TAD Secretary General Rebecca grinspand
sounded the alarm so strongly.
Speaker 1 (04:05):
What did she say?
Speaker 2 (04:06):
Her quote was direct countries should act now. The message
is clear. The window to shape this is closing. If
we just wait and see then.
Speaker 1 (04:14):
That four point eight trillion dollar boom.
Speaker 2 (04:16):
It'll overwhelmingly flow to a handful of wealthy nations and
powerful corporations. The report warns this could lead to a
catastrophic global division. Our policy choices made right now will
decide if AI actually serves global progress or just deepens inequality.
Speaker 1 (04:33):
Okay, that perfectly sets up our mission for this deep dive.
Then we need to really get into what that forty
percent effective figure means. We need to map out that
inequality chasm. The report describes North versus South, capital versus.
Speaker 2 (04:45):
Labor, and crucially examine the solutions the pragmatic policy Playbook
unc TAD is proposing to actually manage this huge transformation.
Speaker 1 (04:55):
Right, let's start with a big number, the one causing
the most anxiety. That and you hear that number, and honestly,
the first reaction is probably panic images of mass unemployment cues.
It sounds almost like hyperbole, maybe designed to grab headlines.
Speaker 2 (05:09):
It's a natural reaction. And yeah, it's a startling figure.
But is it credible? I think the credibility comes from
the nuance and the methodology used by both UNCTAD and
the IMF who they cite.
Speaker 1 (05:18):
Okay, so how do they arrive at forty percent? What's
the method?
Speaker 2 (05:21):
It's not just counting job titles. It's based on these
occupational exposure models. They break down jobs into their individual tasks.
Speaker 1 (05:28):
Ah so, not is a lawyer automatable, but which tasks
that lawyers do.
Speaker 2 (05:34):
Are automatable precisely, it's much more granular and based on
that task analysis. They define affected in two ways. A
job is affected if AI can either complement the human
doing the.
Speaker 1 (05:46):
Work, meaning make them more productive, help them.
Speaker 2 (05:48):
Out exactly, boost their efficiency their output. Or the second way,
if AI can substitute for the human tasks.
Speaker 1 (05:56):
Leading to displacement or maybe the task just vanishing for
humans together.
Speaker 2 (06:00):
Right, So affected really means relevant to current or near
future AI capabilities. It's about the technical feasibility of applying
AI to the stuff people actually do in their jobs.
Speaker 1 (06:11):
Got it. So it's not a prediction of forty percent.
Speaker 3 (06:13):
Unemployment, definitely not.
Speaker 2 (06:14):
It's a measure of exposure, and as you'd expect, that
exposure varies massively depending on the type of economy. The
report clearly maps out three global tiers.
Speaker 1 (06:23):
Okay, let's break those down. Where's the exposure highest?
Speaker 3 (06:26):
Its highest?
Speaker 2 (06:26):
In the advanced economies they estimate a staggering sixty percent
exposure rate there.
Speaker 1 (06:31):
Sixty percent? Wow? Why so high? In the developed world?
You think they'd be more resilient.
Speaker 2 (06:36):
It's counterintuitive, isn't it, But it's because advanced economies are
primarily knowledge intensive. Their main functions revolved around processing information,
analyzing data, generating text finding patterns.
Speaker 1 (06:47):
Exactly the kind of cognitive work that large language models
and generative AI are getting really good at.
Speaker 3 (06:53):
You nailed it.
Speaker 2 (06:54):
This is where that in version of the script we
talked about really hits home. It's not just automating manual
laboring anymore.
Speaker 1 (07:00):
Give us some concrete examples. We're talking white collar jobs, right,
professional roles people thought were secure.
Speaker 2 (07:05):
Absolutely. Think about, say, junior lawyers. A lot of their
early work involves legal discovery, summarizing old cases, drafting basic
motions or sections of contract.
Speaker 4 (07:16):
The grunt work basically kinda yeah, But AI tools specifically
trained on legal data can now produce a solid first
draft of a complex brief, or sift through thousands of
discovery documents in minutes, not weeks.
Speaker 3 (07:30):
The human lawyer's role then shifts.
Speaker 1 (07:32):
From drafting it to checking it.
Speaker 2 (07:34):
Exactly, from creating to editing and verifying, which is still vital,
but it changes the whole structure.
Speaker 1 (07:40):
And it hits the training pipeline hard. Presumably, how do
you learn to draft If the AI does the.
Speaker 2 (07:45):
First pass, that's a massive challenge Those entry level roles
where you learn the ropes, they might just evaporate.
Speaker 3 (07:51):
Look at accounting too.
Speaker 1 (07:53):
Financial service is another huge sector in advanced economies.
Speaker 2 (07:56):
Right think about reconciling ledgers, basic tax prep, odd analysis patterns.
These are routine cognitive tasks. AI can ingest the data,
match things up, flag anomalies almost instantly.
Speaker 1 (08:08):
Even highly specialized fields aren't immune medicine.
Speaker 2 (08:12):
Even there, AI is already showing remarkable results, sometimes outperforming
experienced human radiologists and spotting subtle anomalies on X raysor
mammograms consistently.
Speaker 1 (08:21):
So the expert the radiologist becomes more of a supervisor,
a verifier of the AI's findings.
Speaker 2 (08:28):
Potentially, Yes, still a crucial role, requiring deep expertise, but
perhaps needing fewer hours overall, or maybe fewer specialists in
total for the same volume of work.
Speaker 1 (08:38):
And this aligns with other research I think you mentioned McKenzie.
Speaker 3 (08:41):
Yeah.
Speaker 2 (08:41):
A twenty twenty five Mackenzy update estimated that forty five
percent of tasks just within the finance and insurance sectors
alone are highly automatable with current tech forty five.
Speaker 1 (08:51):
Percent, So the substitution potential is hitting right at the
core of the global NORDS Economic Engine. Okay, let's contrast
that sixty percent with the next tier.
Speaker 2 (09:00):
That would be emerging markets. Here, the exposure rate drops,
but it's still significant forty.
Speaker 1 (09:06):
Percent, still very high. What kind of jobs are most
exposed there?
Speaker 2 (09:09):
It tends to be less complex work, often stuff that's
been standardized for outsourcing we're talking about. There's really large
scale customer service operations, call centers running on scripts.
Speaker 3 (09:19):
Yeah.
Speaker 2 (09:19):
Yeah, high volume data entry, basic tech support, level one stuff,
simple translation tasks. The backbone of the BPO business process
outsourcing industry for the last two decades.
Speaker 1 (09:30):
And this is where that thread of offshoring reversal really bites,
isn't it.
Speaker 2 (09:34):
It's a huge threat and the speed is alarming. Why
would a US or European company continue to offshore a
call center to say, Manila, or software testing to Bengaluru
purely for lower labor.
Speaker 1 (09:48):
Costs if they can deploy an AI chatbot or an
automated testing suite back home that costs virtually nothing per interaction.
Speaker 2 (09:55):
Exactly zero payroll, zero benefits, infinitely scalable. Age advantage isn't
just reduced, it's often completely neutralized by zero wage automation.
This could pull the rug out from under the economic
model of many middle income nations.
Speaker 1 (10:09):
Okay, then the final tier, low income countries. The exposure
rate there is the lowest.
Speaker 2 (10:15):
It is estimated at twenty six percent.
Speaker 1 (10:17):
Why lower? What protects them? Relatively speaking?
Speaker 2 (10:19):
It reflects their current economic structure. These economies are still
heavily reliant on manual physical labor, which AI and robotics
have historically found harder to replicate things like complex farming,
construction work, traditional crafts, resource extraction.
Speaker 1 (10:33):
But the reports suggest this is temporary protection very much.
Speaker 2 (10:36):
So unc TET is clear this relative safety won't last.
We're already seeing AI driven drones used in agriculture for
crop monitoring, and robotic harvesters are starting to appear in
large scale farming as robotics get better eyes and hands, basically.
Speaker 1 (10:52):
Better dexterity, better visual perception.
Speaker 2 (10:54):
Right, that twenty six percent figure is likely to climb
potentially quite rapidly in the coming year.
Speaker 1 (11:00):
So across all these tiers sixty percent twenty six percent.
The key question isn't just if a job is exposed,
but how that split.
Speaker 2 (11:09):
You mentioned earlier that fifty to fifty split between complementarity
and substitution That is the most critical distinction for anyone
thinking about policy.
Speaker 1 (11:16):
So roughly half the jobs exposed might actually get better.
Speaker 2 (11:19):
Potentially, yes, they become AI augmented. Think of a digital marketer.
They could use AI to sift through millions of customer commons,
understand sentiment instantly, and then deploy incredibly personalized ad campaigns
almost in real time. Their effectiveness skyrocket.
Speaker 1 (11:34):
Making them much more valuable exactly.
Speaker 3 (11:36):
But then there's the other half. Those jobs risk.
Speaker 2 (11:39):
Being diminished or outright eliminated because the AI can perform
the core task better, faster, or cheaper.
Speaker 1 (11:45):
Which brings us back to that inversion of the script.
It's not about replacing hands anymore. It's about replacing cognitive functions.
Speaker 2 (11:51):
It completely flips the historical relationship between skills and job security.
The Industrial Revolution mostly hit physical labor farm armours, artisans,
later automation factory assembly.
Speaker 1 (12:04):
Lines, but AI goes for the brain work.
Speaker 2 (12:06):
It targets cognitive labor. Those high skill, often high wage
roles involving complex pattern matching calculations, sophisticated language generation. They
are suddenly vulnerable. This demands a total rethink of education, training, everything,
and the.
Speaker 1 (12:21):
Speed of this change That's what makes this wave so different,
so much harder to manage than past disruptions.
Speaker 2 (12:27):
The velocity is unprecedented. We absolutely cannot overstate this pass shifts.
Even disruptive ones usually unfolded over decades. Society policy they
had some time, albeit often not enough to.
Speaker 1 (12:39):
Adapt, Like the Industrial Revolution.
Speaker 2 (12:40):
Example, right when steam power displaced say forty percent of
the UK's hand loom levers in the early eighteen hundreds.
That was brutal for those individuals, but the transition to
factory work took maybe fifty years. Even electricity in the
nineteen twenties automating around thirty percent of manufacturing cask happened
along inside the gradual growth of new industries over decades.
Speaker 1 (13:02):
But AI moves at the speed of the.
Speaker 2 (13:04):
Internet precisely, which creates this kind of political and social vertigo.
The report uses chat GPT as a benchmark, reaching one
hundred million users in just two months. That's faster than TikTok,
faster than Instagram, faster than any major technology adoption curve
we've ever seen.
Speaker 1 (13:20):
And what does that speed mean practically for workers and governments.
Speaker 3 (13:23):
It means the.
Speaker 2 (13:23):
Time lag between a job task being automated away and
a new compensating role potentially emerging is dramatically compressed. There's
less time to adjust, retrain, relocate. Policy can't afford its
usual bureaucratic pace. It needs to react almost in real time,
or the social friction, the human cost of this displacement
could become truly unmanageable.
Speaker 1 (13:43):
Okay, the speed is terrifying because, as you said, the
displacement isn't some future threat. It's happening now. The machine
is running. Let's look at the concrete examples UNCTAD gathered.
Where is AI actively eliminating roles today, starting with creative
fields which many thought were uniquely human.
Speaker 2 (14:02):
Yeah, that idea of creative work being somehow safe is
really being challenged. Tools like mid Journey, Deli stable Diffusion
they can generate high quality images, illustrations, even add concepts
from simple text prompts.
Speaker 1 (14:15):
So the need for junior designers' concept artists it's shrinking.
Speaker 2 (14:19):
Why hire a team for weeks to brainstorm ideas When
an AI can spit out fifty different, usable, legally clear
concepts in literally an hour. It structurally removes the need
for a lot of that initial entry level creative work.
Speaker 1 (14:32):
And in areas with huge volume. Like customer seemas, the
numbers are stark.
Speaker 2 (14:36):
The Klara example cited in the report is a perfect
quantifiable case. Their AI chatbot handled two point three million
customer conversations in just one month in twenty twenty.
Speaker 1 (14:45):
Four, two point three million, and the human equivalent.
Speaker 2 (14:48):
They estimated that workload was equivalent to about seven hundred
full time human agents. That wasn't about making agents more efficient.
That was a direct substitution to cut costs, plain and simple.
Speaker 1 (14:59):
And it's not just lower skilled roles you mentioned. High
value professional services are seeing this too.
Speaker 2 (15:03):
Law for instance, absolutely legal tech is booming. Systems like
Harvey AI, which we touched on, can draft contracts, analyze precedents,
write initial briefs reportedly ninety percent faster than junior lawyers
or pair.
Speaker 1 (15:17):
Of legals ninety percent faster. How does a human compete
with that on cost or time?
Speaker 2 (15:23):
They can't for those specific tasks. It fundamentally changes the
economics of legal work and again the entry point into
the profession. We even saw companies like Duo Lingo in
the language learning space cut ten percent of their human
contractors in twenty twenty four, explicitly saying AI could now
generate and refine lessons more effectively.
Speaker 1 (15:40):
So white collar knowledge work is clearly in the firing line.
But what about the physical world manufacturing logistics, That.
Speaker 3 (15:47):
Trend is also accelerating.
Speaker 2 (15:49):
We all remember the Foxconn story from twenty sixteen replacing
sixty thousand iPhone assemblers with robots. But the difference now
is that the robots, the cobots are getting.
Speaker 1 (16:00):
Smarter cobots, collaborative robots.
Speaker 2 (16:02):
Yeah, they have better vision systems, more dexterity, they can
handle tasks that aren't perfectly repetitive, making them viable replacements
even in smaller factories or for more customized production lines.
Speaker 1 (16:13):
And even in the tech industry itself, the heartland of
AI development.
Speaker 2 (16:17):
The evidence is right there in the layoff numbers. The
US tech sector shed around two hundred and sixty thousand
jobs in twenty twenty three and early twenty twenty four,
and significantly in about thirty percent of those layoff announcements,
major companies like Microsoft and Google specifically mentioned leveraging AI
as a factor enabling them to reduce headcount.
Speaker 1 (16:35):
So the corporate world is actively planning for this. It's
becoming strategy.
Speaker 2 (16:39):
It seems so the world Economic Forum survey found forty
one percent of companies globally planned some level of AI
driven workforce reduction within the next three years. This isn't
just speculation. It's baked into corporate planning based on expected
cost savings and efficiency gains.
Speaker 1 (16:54):
Okay, that's the displacement side, and it sounds pretty relentless.
But let's pivot. We have to ask about the other
side of the coin, the creation counterbalance. Where are the
new jobs? History usually shows and technology creates more jobs
than it destroys eventually.
Speaker 2 (17:10):
That's the historical pattern and it's the basis for a
lot of optimism. The classic analogy is the ATM.
Speaker 1 (17:16):
Right. ATMs replaced many bank tellers.
Speaker 2 (17:19):
But they also lowered the cost of running branches, so
banks open more branches, hired more loan officers, financial advisors,
it staff to manage the networks. Net Employment and banking
actually went up even though the specific role of the
teller declined.
Speaker 3 (17:33):
The jobs changed.
Speaker 1 (17:34):
So the hope is AI follows that pattern. What specific
new roles are emerging directly because of AI?
Speaker 2 (17:40):
Broadly, you can group them into a few categories. First,
there are the roles needed to manage and guide the
AI itself. Prompt engineers people who are experts at crafting
the right questions and instructions to get reliable, useful, ethical
outputs from these powerful, but sometimes unpredictable AI models.
Speaker 1 (17:57):
So prompt engineering, Why is that so crucial? If the
AI is supposedly smart.
Speaker 2 (18:01):
That's a great point. It's because these large models are generalists.
They know a lot, but they can also hallucinate, make
things up, or give answers that are technically right but
completely miss the business context. A good prompt engineer is
like a translator, bridging the gap between a vague human
need and the specific input the AI needs to deliver value.
They constrain the AI, guide it, focus it. It turns
(18:24):
raw AI power into a usable business tool.
Speaker 1 (18:27):
Okay, prompt engineers, what else?
Speaker 2 (18:28):
We definitely need AI at the CIS people specifically tasked
with auditing models for bias, fairness, and safety. And then
there's a huge need for data curators and AI trainers.
Speaker 1 (18:38):
Data curators like librarians.
Speaker 3 (18:40):
For data sort of yeah.
Speaker 2 (18:42):
Ensuring the vast data sets used to train these models
are clean, representative, unbiased, and legally compliant garbage in, garbage outright.
High quality data is paramount It's why LinkedIn listed AI
trainer as its fastest growing job title for twenty twenty five.
Someone has to prepare that data.
Speaker 1 (18:59):
Okay, so managing the AI are their entirely new industries
emerging too.
Speaker 2 (19:03):
Absolutely, think about the whole ecosystem that will be needed
to support autonomous vehicles. Maintenance technicians, remote monitoring centers, regulatory experts,
mapping specialists. Projections suggest this could create maybe a million
jobs in the US alone by twenty thirty.
Speaker 1 (19:19):
A whole new sector.
Speaker 2 (19:21):
Or consider personalized medicine. Using AI to analyze genomic data
or patient histories can lead to incredibly tailored treatments, but
you need experts, genetic counselors, specialized nurses, data analysts to
interpret those AI findings and integrate them into actual patient
care plans. These are human machine collaboration roles that barely
(19:41):
existed five years ago.
Speaker 1 (19:43):
And then there's the other half of that fifty to
fifty split we discussed augmentation, the people whose jobs aren't
replaced what are supercharged by AI.
Speaker 2 (19:50):
This is where a lot of the optimism lies. The
unc TAD report highlights examples like doctors using AI agnostic tools.
The AI handles the pattern recognition on scans faster, freeing
up maybe twenty percent more of the doctor's time.
Speaker 1 (20:03):
Time They can then spend on.
Speaker 2 (20:04):
On the human element, talking to patients, explaining complex conditions,
providing empathy, dealing with unique cases, the stuff AI can't do.
It potentially transforms the job into something higher value, more
focused on uniquely human skills, and maybe even more satisfying.
Speaker 1 (20:20):
Okay, so we have displacement but also creation and augmentation.
Sounds like it could balance out, maybe even be a
net positive. But there's a huge catch, isn't there.
Speaker 2 (20:28):
There's a massive catch. It's the reskilling challenge, the transition friction.
It's the gap between the jobs being lost and the
jobs being created.
Speaker 1 (20:36):
And bridging that gap requires what money?
Speaker 2 (20:40):
A lot of money and infrastructure. Unc TED puts a
number on it. They estimate empowering the global workforce for
this transition requires investing something like one trillion dollars globally
by twenty thirty. That's just for reskilling programs, education system upgrades,
training subsidies a trillion.
Speaker 1 (20:56):
Dollars A trillion. Are we seeing successful efforts to bridge
that gap up anywhere? Even on a smaller scale.
Speaker 2 (21:02):
We're seeing pockets of innovation. Singapore's Skills Future program is
often cited. It is every citizen credits like a voucher
they can use for approved training courses throughout their life.
They've subsidized AI related courses for hundreds of thousands of workers.
Speaker 1 (21:17):
Already proactive government support.
Speaker 2 (21:19):
Yes, Or look at Rowanda. They're focusing on very specific,
high value niche skills. They're training people in advanced drone
repair and maintenance, recognizing that's a growing field where they
can build expertise.
Speaker 1 (21:32):
But these seem like exceptions. The overall challenge remains immense right,
especially for older workers or those without advanced education.
Speaker 2 (21:40):
That's the harsh reality the report doesn't shy away from.
It's incredibly difficult for say a fifty year old former
data entry clerk or factory worker, maybe without a college degree,
to suddenly retrain as an AI alignment specialist requiring advanced
math and coding skills. The cognitive leap is just too
large for many people without significant sustain and support.
Speaker 1 (22:00):
So if we don't invest that one trillion dollars, if
we don't build those.
Speaker 2 (22:04):
Bridges, then the theoretical net job game becomes meaningless for
the millions who are displaced the benefits of AI, the
higher wages, the new opportunities will overwhelmingly flow to those
who already have the high skills and educational background needed
to grab them. It becomes a recipe for massive social
and economic polarization.
Speaker 1 (22:24):
Okay, So that failure to bridge the skills gap, combined
with the basic mechanics of AI, leads us straight into
the core warning of the unc TAD report. AI is
poised to dramatically worsen inequality both within nations and between
them unless we actively intervene.
Speaker 2 (22:41):
Exactly, and it starts with that core mechanism we mentioned.
AI favors capital over labor.
Speaker 1 (22:46):
Remind us how that works again, well, Building.
Speaker 2 (22:48):
And running advanced AI requires huge upfront investment in physical infrastructure.
We're talking massive data centers, enormous amounts of electricity, specialized
cooling systems, and critically those very expensive high ends the
graphics processing unit.
Speaker 1 (23:01):
It's computer chips, right.
Speaker 3 (23:03):
These are all capital assets.
Speaker 2 (23:05):
So when AI generates value, the primary financial returns the
profits flow directly back to the owners of that capital,
the companies that own the servers, the shareholders of the
chip manufacturers, not primarily to the human workforce involved, and.
Speaker 1 (23:19):
The most visible sign of this is the soaring value
of the company's making those chips.
Speaker 2 (23:23):
It's the clearest illustration. Look at Nvidia, their market value
hitting three trillion dollars in twenty twenty four. That wasn't
driven by selling more video games. It was almost entirely
fueled by the insatiable demand for their specialized chips needed
to train and run large AI models. That wealth accumulation
is incredibly concentrated.
Speaker 1 (23:43):
Okay, let's look at the impact within countries.
Speaker 3 (23:45):
First.
Speaker 1 (23:46):
The report talks about a K shaped recovery effect from AI.
What does that mean? Visualize the K for us?
Speaker 2 (23:52):
A K shape is a perfect metaphor for the growing divergence.
Imagine the letter K, the upper arms slanting upwards. That
represents the people in jobs where AI complements their skills significantly,
The augmentees exactly, the senior engineers fine tuning the models,
the data scientists designing the systems, the managers leveraging AI
insights for strategic decisions. These people become hyper productive, high
(24:15):
leverage employees. Their skills are in high demand, and their
scarcity drives up their wages dramatically.
Speaker 1 (24:21):
How much what kind of increases are we seeing.
Speaker 2 (24:24):
The report notes these high complementarity roles are seeing real
wage growth of twenty to thirty percent, sometimes even more.
Speaker 1 (24:30):
Twenty Your thirty percent is huge. How does AI actually
justify that kind of pay bump? For say, senior software engineer.
Speaker 2 (24:38):
Because their job transforms, they're no longer spending hours writing
routine code. The AI can handle a lot of that
boilerplate stuff. Instead, they focus one hundred percent of their
time on the really complex parts the system architecture, integrating
the AI into legacy systems, solving unique edge cases, overseeing
the AI's performance.
Speaker 3 (24:56):
They're essentially leveraging.
Speaker 2 (24:58):
The AI to do the work that might have previously
taken a team of five or ten junior coders. The
company captures that value, and the market rewards the leveraging
engineer with the significant chunk of it.
Speaker 1 (25:08):
Okay, that's the upward armor of the K. What about
the downward arm.
Speaker 2 (25:11):
That represents the stagnation or even decline for workers in
jobs where AI primarily substitutes for their skills, or where
the complementarity is very low. Think administrative assistance, basic data processors,
content writers producing routine articles, maybe some customer service.
Speaker 1 (25:27):
Roles their skills become less valuable or redundant.
Speaker 2 (25:30):
Exactly, demand falls, wages stagnate or even decrease in real terms.
The imf RAN simulations mentioned in the report that painted
a stark picture. AI could boost the incomes of the
top ten percent by around fifteen percent, while the bottom
fifty percent might see their average incomes actually dip by
five percent due to this pressure.
Speaker 1 (25:49):
A widening gap, and this divergence is reinforced by education
levels massively.
Speaker 2 (25:53):
Education acts like a gatekeeper here. The data, particularly from
the US, shows that jobs with high AI exposure do
do command a wage premium, maybe twenty five percent higher
pay than non exposed jobs. But and this is critical,
that premium flows almost exclusively to workers who have a
college degree. If you don't have that degree, AI exposure
is far more likely to mean substitution and wage pressure,
(26:15):
not augmentation and a pay rise. It hardens the socioeconomic divide, and.
Speaker 1 (26:19):
We also need to consider the impact on different demographic groups.
Right who holds these routine cognitive jobs that are most
vulnerable to substitution.
Speaker 2 (26:27):
Disproportionately These roles administrative support, data entry, some service jobs
are held by women and minority groups. Especially in economies
with less social mobility. So the automation of these jobs
risks heading already disadvantaged community's hardest, exacerbating existing inequalities and
making it tougher for them to access the resources needed
(26:48):
for reskilling.
Speaker 1 (26:48):
Okay, that's the inequality within nations. Now let's turn to
what U and C. TAD calls the real bombshell, the
widening inequality between nations, the AI divide. This is where
the geopolitical risk gets released.
Speaker 2 (27:00):
Serious, it is deeply alarming. The concentration of AI power
and resources is just staggering. The report found that just
one hundred companies, and get this, ninety percent of them
are based in either the US or China control about
forty percent of all global spending on AI research and development.
Speaker 1 (27:18):
One hundred firms controlling almost half the global R and D.
Speaker 2 (27:22):
And think back to that comparison we made earlier. The
combined market value of just three tech giants Apple, Microsoft,
Nvidia is greater than the entire annual economic output the
GDP of the entire African continent.
Speaker 1 (27:36):
Wow, just let that sink. In three companies versus a continent.
Speaker 2 (27:39):
That single data point tells you almost everything you need
to know about the concentration risk here.
Speaker 1 (27:43):
And this financial power is directly tied to control over
the essential infrastructure, the compute power.
Speaker 2 (27:49):
Compute power is the critical bottleneck. It's the new oil,
as some say, but unlike oil, it's not easily substitutable,
and the supply of the most advanced chips is incredibly centralized.
The US currently can controls access to around sixty percent
of the world's advanced semiconductor.
Speaker 1 (28:03):
Supply sixty percent, and Africa's.
Speaker 2 (28:05):
Share a devastating point one percent, zero point one. If
you don't have access to these high end chips, you
simply cannot train your own large sophisticated foundational AI models
tailored to your own languages, your culture, your specific economic challenges.
You're immediately dependent.
Speaker 1 (28:22):
Dependent on the models built in the north, which leads
to this idea of data colonialism.
Speaker 2 (28:27):
It's the modern digital equivalent of resource extraction, but maybe
even more insidious.
Speaker 3 (28:32):
Think about it.
Speaker 2 (28:33):
The raw material is the data generated every day by
people in developing countries. Their social media posts, their search queries,
their online interaction of data, Your data, my data, everyone's data.
That data is harvested, often with opaque consent, and used
to train the huge, powerful AI models primarily developed in
the US and China, and then those sophisticated AI services
(28:55):
are sold back globally, including to the very countries whose
data help build them. The profits flow north. The South
provides the raw material for free or nearly free, and
then pays to buy back the finished product. It creates
this perpetual cycle of dependency, and this.
Speaker 1 (29:09):
Whole dynamic fundamentally breaks the traditional development ladder that countries
used to climb. The report used Bangladesh's garment sector as
a key example.
Speaker 2 (29:18):
It's a powerful and worrying case study. Bangladesh's garment industry
is the backbone of its economy, worth around fifty billion
dollars a year, employing about four million people, mostly women.
It's their primary export earner, built entirely on that low
cost labor advantage.
Speaker 1 (29:35):
But AI threatens that how.
Speaker 2 (29:37):
Unc tad estimates that advancements in robotics, specifically AI powered
vision systems, combined with increasingly dexterous robotic arms basically AI
sewing bots, could automate tasks threatening up to sixty percent
of those garment jobs by twenty thirty.
Speaker 1 (29:52):
Sixty percent of four million jobs. That's potentially catastrophic is.
Speaker 2 (29:56):
If a country loses its main competitive edge, cheap scalable
labor to automation, before it has managed to transition its
economy and workforce towards higher skill, knowledge based industries. It
risks getting stuck trapped in poverty, unable to climb that
development ladder because the lower rungs have been automated away.
Speaker 1 (30:11):
And adding insult to injury. The report highlights that the
Global South is also being excluded from shaping the rules
of the game the governance aspect.
Speaker 2 (30:21):
Yes, the governance exclusion is critical. UNCTAD found that one
hundred and eighteen nations from the Global South were effectively
absent from key international discussions trying to set ethical guidelines
and standards for AI development and employment, things like the
G sevens Hiroshima AI process.
Speaker 1 (30:37):
So they don't get a say in the rules that
will govern a technology profoundly shaping their future.
Speaker 2 (30:42):
Exactly, if you don't have a seat at the table
when standards for data privacy, algorithmic fairness, safety protocols, or
even liability are being decided, you risk having rules imposed
on you that were designed by and for the Global North,
rules that might embed biases against your population, misunderstand your
cultural context, or simply not serve your development needs. You
(31:03):
become a rule taker, not a rule maker.
Speaker 1 (31:05):
Which brings us back to that huge pwcstinate. The report
mentions AI could add fifteen point seven trillion dollars to
global GDP by twenty thirty, But the AI divide means
that wealth won't be shared.
Speaker 2 (31:16):
That's the stark warning. Without deliberate intervention, the vast majority
of that fifteen point seven trillion dollar windfall will be
generated and captured in the north, primarily the US and China.
Instead of lifting all boats, AI could dramatically deepen global
inequality and instability. The stakes couldn't be higher.
Speaker 1 (31:35):
Okay, we've laid out the scale of the potential disruption,
that forty percent exposure and the profound risk of deepening inequality,
the AI divide and the K shape the UNCTAD report
isn't just about warnings. It's crucially about solutions. Let's move
to the case studies. They highlight the ground zero examples.
Speaker 2 (31:54):
Yes, these four cases are really important because they show
how different policy responses lead to different outcomes and make
the theory concrete and suggest a crucial pattern proactive policy
can genuinely soften the landing.
Speaker 1 (32:05):
Let's start with case one, India and its massive IT sector.
Speaker 2 (32:08):
Right, India's IT and business process management sector is a
global giant two hundred and fifty billion dollar industry employing
over five point four million people. It's been a huge
engine for middle class.
Speaker 1 (32:19):
Growth, but it's facing immediate disruption, particularly at the entry level.
Speaker 3 (32:23):
Exactly.
Speaker 2 (32:24):
AI coding assistance like GitHub, Copilot and others are becoming
incredibly proficient. They can now write maybe forty percent of
the routine boilerplate code automatically. Major Indian tech firms like
Infosis and Whipbrow have publicly stated they're aiming for high
levels of automation maybe fifty percent in areas like software
testing and maintenance.
Speaker 1 (32:43):
And the impact on jobs it's been.
Speaker 2 (32:46):
Felt immediately at the entry point. The report estimates that
around two hundred thousand fresher rolls those first jobs for
recent graduates effectively disappeared in the twenty twenty four to
twenty twenty five period because those tasks were automated.
Speaker 1 (33:00):
Two hundred thousand entry level jobs gone. That's a huge
blow to the pipeline. How did India respond? What's the
policy counterbalance.
Speaker 2 (33:07):
India's response has been quite rapid, largely focused on a
massive push for upskilling and reskilling, often through public private partnerships.
They've established large scale training hubs in major tech centers
like hydrobad and Bengaluru.
Speaker 1 (33:19):
Training people in what specifically, if coding.
Speaker 2 (33:21):
Is being automated, they're focusing on skills adjacent to AI
but less likely to be fully automated soon. A big
focus is on MLOPS machine learning operations MLOPS.
Speaker 1 (33:31):
We hear that term a lot. Explain why that's a
valuable pivot.
Speaker 2 (33:34):
MLOPS is essentially the practical side of AI deployment. It's
everything involved in taking an AI model developed in a
lab and actually getting it to work reliably and continuously
in the real world. This includes deploying the model, monitoring
its performance, maintaining it, retraining it when its performance degrades.
Speaker 1 (33:51):
Because AI models aren't static, they need ongoing work.
Speaker 2 (33:54):
Exactly, They suffer from model drift. Their accuracy can decline
as the real world data chain, so you need skilled
humans to constantly manage this life cycle. It's complex, it
requires expertise in both software engineering and data science, and
it's crucial for keeping these AI systems running effectively in businesses.
India recognized that while basic coding was vulnerable, managing the
(34:16):
deployed AI systems was a growing high value Niche They're
training maybe one hundred thousand people annually in.
Speaker 1 (34:22):
These hubs Okay shifting gears Case two Kenya and the
call center crunch. This illustrates that offshoring reversal we discussed right.
Speaker 2 (34:30):
Nairobi has built a reputation as Silicon Savannah, partly thanks
to its thriving BPO sector, employing around two hundred thousand people.
But AI voice spots are getting incredibly good, especially localized ones.
Speaker 1 (34:43):
Localized meaning they can handle local languages and.
Speaker 2 (34:45):
Accents precisely, AI bots that can flawlessly handle customer queries
in Swahili, understanding nuances far better than a generic global
bot could. This directly impacts the value proposition of Kenyon
call centers. Safari Coom, the big regional telecom operator, cut
about fifteen percent of its human call center agents in
twenty twenty four, explicitly citing the capabilities of these AI bots.
Speaker 1 (35:09):
Sounds like a familiar story automation hitting outsourcing, but ken
You found a different pivot.
Speaker 2 (35:13):
They did instead of just trying to protect the existing
call center jobs. They looked at moving up the AI
value chain through a government initiative called Ajira Digital.
Speaker 1 (35:22):
What does a geerdigital do?
Speaker 2 (35:23):
It focuses on training young Kenyans not to be call
center agents, but to become the human workforce behind the AI,
specifically in high quality data annotation, labeling and validation.
Speaker 1 (35:33):
So preparing the data that AI models.
Speaker 2 (35:35):
Need to learn exactly tasks like identifying objects in images
for computer vision systems, transcribing audio data accurately, or verifying
the quality of AI generated text. Kenya is essentially exporting
this labeled data as a service to global AI companies
like OpenAI and others. They're leveraging their linguistic skills and
(35:55):
local knowledge, turning it into a valuable input for the
global AI economy rather than just being replaced by it.
Speaker 1 (36:02):
Interesting pivot, Okay, Let's look at developed economy example K
three Germany and its middle stand. They seem to be
focusing more on augmentation.
Speaker 2 (36:10):
Germany's approach is quite different, heavily influenced by its strong
industrial base and its famous middle stand, the network of
small and medium sized manufacturing companies. Their strategy is very
clearly focused on AI augmentation, not displacement.
Speaker 1 (36:23):
How does that work? In practice?
Speaker 2 (36:25):
On a factory floor, they make heavy use of Cobot's
collaborative robots. These aren't designed to replace human workers entirely,
but to work alongside them. A cobot might hold a
heavy part in place while a human performs a complex weld,
or assist with lifting and moving materials. The result significant
productivity boosts. The report cites output increases of around thirty percent,
(36:47):
but crucially without large scale layoffs.
Speaker 1 (36:50):
And what policy levers enable this augmentation focus.
Speaker 3 (36:54):
Two main things.
Speaker 2 (36:55):
First, Germany's very strong tradition of vocational training the ousebuilding system,
which is at adapting skills. Second, targeted government policy. There
are significant subsidies available covering up to fifty percent of
the cost for companies to rescale their existing workforce to
work with these new robotic and AI tools, programming them, maintaining.
Speaker 1 (37:14):
Them, so they invest in the human alongside the machine.
Speaker 2 (37:16):
Precisely, the policy prioritizes enhancing human capital, ensuring workers adapt
and their value increases alongside the technology. This focus on collaboration,
backed by strong social partnership and investment, is a key
reason why German unemployment has remained remarkably low, around three percent,
despite technological change.
Speaker 1 (37:33):
Okay, one more case four the Philippines. This one highlights
a different vulnerability, the impact on digital gigwork and remittances.
Speaker 2 (37:42):
Yes, the Philippines is a major recipient of remittances from
overseas Filipino workers OFWs, and increasingly a significant chunk of
that income comes from online digital gigwork things like freelance, transcription,
virtual assistance, content moderation. Million of the families rely on
this income.
Speaker 1 (37:59):
In AI is hitting these gigs hard, very hard.
Speaker 2 (38:02):
Take online transcription platforms like rev dot com as AI
speech to text technology got dramatically better. These platforms automated
large parts of their workflow. When seventy percent of the
work disappears, the income stream for thousands of Filipino transcriptionists
collapses almost overnight.
Speaker 1 (38:18):
A sudden shock to household incomes. What's the policy response there?
Speaker 2 (38:21):
The Philippine government launched a program called two pad AI
in twenty twenty five. It's a large scale, targeted retraining
initiative specifically aimed at helping those displaced by automation in
the gig economy. The goal is to transition about fifty
thousand workers, particularly former transcriptionists, into higher value digital roles
(38:41):
like specialized virtual assistants, social media managers or data annotator.
Skills still in demand. They're actively trying to prevent the
value chain from collapsing entirely for these workers.
Speaker 1 (38:52):
So these four cases India, Kenya, Germany, Philippines, they really
underscore the UNCCAD conclusion, don't the outcome isn't predetermined by
the technology itself, not at all.
Speaker 2 (39:03):
It's driven by the structural policy choices that governments and
societies make in response to the technology. Proactive investment in skills,
infrastructure and social safety nets makes a huge difference between
a managed transition and a chaotic disruption.
Speaker 1 (39:16):
Which leads us perfectly into the specific policy recommendations from
unc TAD their proposed policy playbook for making AI more inclusive.
Speaker 2 (39:23):
Right, this is the roadmap they offer. It's pragmatic and
covers several key areas.
Speaker 1 (39:27):
Let's start with the foundation digital public infrastructure or DPI.
They cite India's Adhar system as a model. Why is
something like a national digital idea so crucial for AI.
Speaker 2 (39:39):
Equity Because DPI access the basic rails for participation in
the digital economy. A universal, secure digital identity system like Adhar,
which covers over one point three billion people in India,
enables access to services, finance, and importantly, it facilitates the creation.
Speaker 3 (39:56):
Of data commons.
Speaker 1 (39:57):
Data commons explain that concept. How is it different from
just making governm data open.
Speaker 2 (40:01):
Open data is often raw, messy, maybe inconsistently formatted. A
data commons is more structured. It's a shared pool of
high quality, ethically sourced, anonymized and standardized data. Think health records,
agricultural yields, traffic patterns, governed by clear rules.
Speaker 1 (40:15):
And why is that so important for developing countries in
the age of.
Speaker 2 (40:18):
AI Because it allows them to overcome the massive data
disadvantage they face. If you have a national data commons,
local researchers, startups and government agencies can access the high
quality data needed to train AI models that are relevant
to their specific context, models that understand local languages, local
crop diseases, local public health challenges.
Speaker 1 (40:39):
So it reduces dependency on the giant models trained on
northern data.
Speaker 3 (40:43):
Exactly.
Speaker 2 (40:43):
It helps level the playing field, allowing the global South
to build and benefit from its own AI ecosystems rather
than just being consumers of Western tech or suppliers.
Speaker 3 (40:53):
Of raw data.
Speaker 1 (40:54):
Okay, DPI is the foundation. The second pillar unc TAD
emphasizes is a skills revolution, and they suggest fronting it
by potentially taxing AI profits.
Speaker 2 (41:02):
Yes, this directly tackles that capital over labor imbalance. If
AI is generating massive profits for a relatively small number
of companies and capital owners. The argument is that society
via the state, need to recapture some of that windfall
to invest in the human.
Speaker 3 (41:17):
Capital of the broader workforce.
Speaker 2 (41:19):
The funding mechanism has to be there for lifelong learning
to become a reality, and they.
Speaker 1 (41:23):
Point to models like Denmark's Disruption councils what makes them effective.
Speaker 2 (41:27):
The Danish model is interesting because it's proactive and collaborative.
These councils bring together unions, industry representatives, and government officials.
Their job is to constantly scan the horizon identify which
sectors and skills are likely to be disrupted by technology
years in advance, so.
Speaker 1 (41:44):
They anticipate change rather than just reacting to.
Speaker 2 (41:46):
It precisely, and based on that foresight, they fund and
sometimes even mandate large scale retraining programs. Denmark aims to
retrain something like forty percent of its workforce every single decade.
Speaker 3 (41:59):
It's systemic.
Speaker 2 (42:00):
It accepts that skills have a shorter shelf life now
and requires continuous collective reinvestment as a national priority.
Speaker 1 (42:07):
Moving to the global level, the third pillar is global governance.
This includes mandating a seat for the Global South in
KEYAI bodies, but also that very ambitious proposal for an
AI solidarity fund.
Speaker 2 (42:18):
This is probably the most politically challenging but potentially transformative
idea in the report. The proposal is for a direct
global transfer mechanism.
Speaker 1 (42:27):
How would it work?
Speaker 2 (42:27):
The idea is to levy a small percentage. They suggest
one percent on the global revenues of the largest multinational
tech companies, the ones benefiting most from the digital economy
in AI.
Speaker 1 (42:38):
One percent of big tech revenue. That would be a
huge amount.
Speaker 2 (42:41):
Of money billions easily, and the proposal is to dedicate
that fund specifically to supporting developing countries, helping them build
their digital public infrastructure, fund their reskilling programs, improve their
access to compute power, essentially using a fraction of the
gains from digitalization to ensure a more equitable global transition.
Speaker 1 (43:01):
But the politics of implementing a global tax like that
seems incredibly difficult. How could it realistically be collected and managed?
Speaker 2 (43:08):
That's the trillion dollar question, isn't it. The report acknowledges
the huge hurdles implementing a global digital services tax requires
unprecedented multilateral cooperation, something that's historically been very hard to
achieve due to national interests and tax competition. It would
likely need a reformed international body, maybe the OECD or
(43:28):
a dedicated UN agency to oversee it. Perhaps nations would
collect the tax on revenues generated within their borders, then
pool it internationally for redistribution based on need and equity criteria.
Speaker 1 (43:40):
So it's less about the specific mechanism right now and
more about establishing the principle.
Speaker 3 (43:44):
I think.
Speaker 2 (43:45):
So establishing the principle that the enormous wealth generated by
global technologies like AI comes with a global responsibility to
manage the downsides and share the benefits more widely. It's
a call for fiscal solidarity in the digital age, okay.
Speaker 1 (43:58):
The fourth area is labor protections. As work changes, especially
with the rise of gigwork, how do we protect workers.
Speaker 2 (44:05):
With traditional stable employment potentially declining. The report stresses the
need for innovative social safety nets A key concept here
is portable benefits.
Speaker 1 (44:14):
Portable benefits explain what that means, especially for someone doing gigwork,
maybe across borders.
Speaker 2 (44:19):
Portable benefits mean that core social protections things like health
insurance contributions, retirement savings, unemployment insurance, sick leave, are tied
to the individual worker, not to a specific employer or job. So,
if you're a data labeler in Kenya working freelance contracts
for three different companies based in Europe in the US.
Speaker 1 (44:39):
Which traditional labor law struggles to.
Speaker 2 (44:41):
Cover exactly, portable benefits, perhaps managed by a government agency
or a third party platform, would allow you to accrue
these benefits consistently across all your gigs. When one contract ends,
you don't suddenly lose your health coverage or retirement contributions.
It provides a crucial layer of security and stability in
an increasingly fragmented way world of work. It decouples essential
(45:02):
security from the outdated notion of a single lifelong employer.
Speaker 1 (45:06):
The report also mentions exploring more structural changes shorter work
weeks UBI yes.
Speaker 2 (45:11):
These are seen as potential long term responses to productivity
gains from AI pilots like Spain's experiment with a thirty
two hour work week are testing if we can share
the productivity dividend by working less for the same day,
potentially creating more jobs overall. And universal Basic income UBI
pilots like those in Kenya or Finland are exploring whether
(45:32):
providing a regular, unconditional income floor can reduce the immense
financial stress of technological unemployment, giving people the breathing room
and security to retrain or pursue other paths.
Speaker 1 (45:44):
These are big ideas, still mostly experimental, very much so.
Speaker 2 (45:47):
But the report argues we need to be seriously exploring
and testing these concepts now, given the scale of the
potential disruption ahead.
Speaker 1 (45:53):
And finally, the playbook includes ethical AI mandates ensuring the
technology itself is developed and deployed responsibly.
Speaker 3 (46:00):
This is fundamental.
Speaker 2 (46:02):
If the tools themselves are biased or unsafe, none of
the other policies matter as much. Key recommendations include mandating
a human in the loop for high stakes decisions, meaning
for critical applications think medical diagnoses, loan applications, hiring decisions,
judicial sentencing recommendations. The AI can provide analysis or recommendations,
but a human being must always retain the final decision
(46:24):
making authority and accountability the AI assists it doesn't.
Speaker 1 (46:29):
Command, and auditing algorithms for bias absolutely crucial, especially since
many foundational models are trained predominantly on data from North
America and Europe.
Speaker 2 (46:38):
When these models are deployed globally, they can inadvertently perpetuate
or even amplify existing biases related to race, gender, language,
or cultural norms found in that training data. Rigorous independent
auditing is needed to detect and mitigate these biases before
they cause harm in diverse contexts.
Speaker 1 (46:56):
It's encouraging. I suppose that the report also notes some
movement from the corporate site big tech making pledges.
Speaker 2 (47:02):
Yes, there is some alignment, though you could argue it's
often driven by enlightened self interest. Microsoft pledging four billion
dollars for AI skilling initiatives globally, Google setting up AI
research hubs in Africa. These companies recognize that they need
a skilled global workforce to use their products, and that
widespread social instability caused by automation is ultimately bad for business.
(47:25):
Talent scarcity and market disruption are risks for them too,
So while it's not purely altruistic, these corporate commitments are
a necessary part of the solution mix hashtag tag tag
tag outro.
Speaker 1 (47:36):
So we've covered a lot of ground, the sheer scale
of the AI revolution, the potential four point eight trillion
dollar market, the forty percent job exposure warning, the deep
risks of inequality amplification, but also crucially the policy pathways
unc TAB proposes for a more equitable future.
Speaker 2 (47:53):
It really brings us to the philosophical heart of the debate,
doesn't it, Because even among experts there's still a deep
split on what the ultimate outcome will be.
Speaker 1 (48:00):
The optimists on one side.
Speaker 2 (48:01):
Right, the techno optimists, people like Eric Brynjolfson at Stanford.
Their core argument rests on history. They say, look, every
major technology shift in the past, steam electricity, computers destroyed
some jobs, but ultimately created far more new jobs and
new industries. Complementarity, they argue, always wins out over substitution in.
Speaker 1 (48:21):
The long run, and they have data to back that up.
Speaker 2 (48:24):
They point to studies like a recent one from MIT
in twenty twenty five showing that companies actively using AI
to augment their workers actually increase their overall hiring rate
by about eleven percent. For optimists, that four point eight
trillion dollars prize isn't a threat, It's a resource that
human ingenuity will inevitably harness for broad progress.
Speaker 1 (48:42):
But then you have the other side, the skeptics.
Speaker 2 (48:45):
The skeptics perhaps best represented by Carl Benedict Frey at Oxford.
Their counter argument is simple, but powerful history might not
be a reliable guide this time because one key variable
is different velocity, the sheer speed of AI development and deployment.
Speaker 1 (48:58):
They worry the changes too fast exactly.
Speaker 2 (49:01):
Frey warns of a real possibility of persistent technological unemployment,
not because new jobs won't eventually emerge, but because the
pace of job destruction might dramatically outstrip society's ability to adapt, retrain,
and absorb displaced workers into those new roles. If the
skills gap is too wide and the transition support is
too slow, we could face a prolonged period of social
(49:23):
disruption and pain, even if the long term picture is positive.
Speaker 1 (49:26):
So where does the unc TAD report land between those
two poles, optimism or skepticism.
Speaker 2 (49:32):
I'd say it carves out a pragmatic, action oriented middle ground.
It acknowledges the immense potential of AI, that potential fifteen
point seven trillion dollars boost a global GDP by twenty
thirty is real, but it insists that realizing that potential
equitably is not automatic. It requires deliberate political choices.
Speaker 1 (49:48):
It's not technodeterminism, but it's about policy precisely.
Speaker 2 (49:52):
The report essentially says, we have the potential resources generated
by AI to fund almost any necessary transition, universal basic services,
global reskilling, on an unprecedented scale, green energy infrastructure.
Speaker 3 (50:06):
The money could be there.
Speaker 2 (50:07):
The fundamental question is whether we globally and nationally have
the political will to implement the mechanisms like that AI
solidarity fund, like robust digital public infrastructure, like mandatory lifelong
learning systems needed to democratize access and distribute the benefits fairly.
Speaker 1 (50:25):
We started this deep dive talking about the human scale
that forty percent figure. Let's end by bringing it back
to the individuals caught up in this wave, because policy
ultimately has to work for that.
Speaker 3 (50:35):
Absolutely.
Speaker 2 (50:35):
The report is filled with data, but behind the percentages
are real people's lives. Think of Maria, who they might
mention hypothetically, a thirty eight year old transcriptionist in Manila.
Speaker 1 (50:43):
Whose gigwork vanished overnight due to automation.
Speaker 2 (50:46):
Right economic stability dawn just like that. But if policy
provides a bridge, maybe access to government funded training credits,
she could potentially enroll in an online course, maybe get
a certificate in AI data moderation, and perhaps find a
new role, maybe even earning twenty percent more. Her story
becomes one of managed transition enabled by support.
Speaker 1 (51:07):
Or think of rejestion in Bengaluru.
Speaker 2 (51:08):
Maybe he was a software quality assurance tester. His job,
stable and middle class gets automated away by sophisticated testing bots.
Speaker 3 (51:16):
Huge pressure.
Speaker 2 (51:17):
But again, if there are accessible government backed programs, maybe
he gets into a cybersecurity boot camp run by a
company like Tata. He pivots finds stability in a related
but higher demand field.
Speaker 1 (51:28):
These individual stories, even if hypothetical, they illustrate the point.
Speaker 2 (51:32):
They show that disruption doesn't have to mean destitution. It
can be managed, navigated, but it requires empathy, foresight, and
critically concrete action and investment in people.
Speaker 1 (51:41):
So the four point eight trillion dollar AI prize is there.
It could fund a global transformation, a creative renaissance, tackle
climate change.
Speaker 2 (51:49):
It could, but only if we make the conscious choice
to democratize access, manage the risks, and share the rewards.
Speaker 1 (51:55):
Because the alternative if we ignore that forty percent warning,
If we let the AI I divide deepen.
Speaker 2 (52:01):
The alternative is a deeply fractured world, A world of
AI augmented elites concentrated in a few regions controlling the
new means of production, and a vast global population facing
precarious work, stagnant wages, and diminishing opportunities. A world of
haves and have nots on a scale we've never seen before.
Speaker 1 (52:21):
But the future isn't set in stone.
Speaker 3 (52:23):
Not at all.
Speaker 2 (52:24):
The report's ultimate message is one of agency. The future
isn't pre coded by the technology. It will be shaped
by the policy choices we make or fail to make.
Right now, we are programming our economic destiny.
Speaker 1 (52:36):
Which brings it back to you. The listener, Rebecca Grinspan
at you n c Ted said, countries must act now
for global progress. If you had to push just one
policy lever first, based on everything we've discussed to nudge
us towards that more equitable future, what would it be.
Speaker 2 (52:50):
Is it investing massively in lifelong learning within your own nation,
perhaps funded by taxing AI profits, or is it pushing
for that ambitious, maybe idealistic global AI Solidarity fund to
ensure countries in the Global South.
Speaker 3 (53:04):
Aren't left behind, that they have.
Speaker 2 (53:06):
The basic tools like compute access and DPI to participate.
Speaker 1 (53:10):
Something to think about. The choices are complex, the stakes
are enormous, and the time to act is well now, we.
Speaker 3 (53:15):
Really must act now. The choices ours food for thought.
Speaker 1 (53:19):
Indeed, thank you for joining us for this deep dive.
We'll see you next time.
Welcome to the deep dive. This is where we take
those really dense global reports, the ones packed with data,
and try to well unpack them into clear, actionable knowledge
for you.
Speaker 2 (00:11):
And today we're tackling a big one, a document that
honestly has huge implications for pretty much everyone, every economy,
every career path, and definitely the future of global equity.
Speaker 1 (00:22):
We're talking about the UNC Tech Technology and Innovation Report
twenty twenty five. Came out early this year April third.
I believe that's right.
Speaker 2 (00:29):
And you know this report, it's more than just a forecast.
It really feels like an urgent map for the next decade.
It lays out this picture of artificial intelligence AI.
Speaker 1 (00:39):
It's everywhere now exactly.
Speaker 2 (00:40):
And the picture is well, it's complicated, it's dazzling. On
one side, you know this incredible engine for productivity, potentially
explosive growth.
Speaker 3 (00:48):
Well there's a butt, a big butt.
Speaker 2 (00:50):
It also presents AI as potentially the biggest driver for
widening global division we've seen since maybe the industrial age.
Speaker 3 (00:57):
It's really starts.
Speaker 1 (00:58):
Okay, let's ground this immediately. The scale, because the numbers
involved here are just almost hard to grasp.
Speaker 3 (01:04):
They are astronomical.
Speaker 2 (01:06):
We're looking at the global AI market, the UNC TAD
projection it's expected to swell to get this four point
eight trillion dollars by twenty thirty three.
Speaker 1 (01:16):
Four point eight trillion. Just say that again, four.
Speaker 3 (01:19):
Point eight trillion dollars.
Speaker 1 (01:21):
Wow, okay, context, help us understand what four point eight
trillion dollars actually looks like?
Speaker 2 (01:25):
Right? Think about Germany's entire economy right now. It's GDP,
Europe's biggest yuconic, that's the one. The AI market alone
is projected to be roughly that size in less than
ten years.
Speaker 1 (01:36):
So this isn't just another tech sector. It's becoming a
global economic force in its own.
Speaker 2 (01:40):
Right, absolutely a major economic superpower, purely based on its
market value and how it's spreading into well everything, every industry.
Speaker 1 (01:47):
Now, what's really driving that value? According to the report,
is it just hype or specific applications?
Speaker 3 (01:53):
No, it's specific.
Speaker 2 (01:54):
The UNC TAD analysis points to massive productivity leaps, primarily
in three core areas, which are healthcare diagnostics think AI
reading scans better than humans. Then there's autonomous logistics, self
driving trucks, automated warehouses, and finally, personalized education delivered at scale.
Speaker 1 (02:12):
So it's not just the flashy consumer stuff we see,
not at all.
Speaker 2 (02:16):
It's about automating these fundamental economic functions. The promise is
genuinely transformative.
Speaker 1 (02:21):
Okay, the promise is huge. But you mentioned the tension,
the dual warning at the heart of this report.
Speaker 3 (02:26):
Let's get into that exactly.
Speaker 2 (02:27):
So while AI offers this unprecedented economic boom, these huge
productivity gains, the unc TAD report is very clear, very stark.
It warns AI could.
Speaker 1 (02:40):
Disrupt disrupt how many up.
Speaker 2 (02:42):
To forty percent of jobs worldwide through automation, through job displacement.
Speaker 1 (02:48):
That's the headline number, isn't it.
Speaker 2 (02:49):
It jumps out at you, it's designed to, and it's
a shocking figure. But the real story, the complexity that
the report really digs into, is the geopolitical angle. Because
a sorearm wave of time technology, unlike maybe previous ones,
it structurally favors and this is their phrase, capital over labor.
Speaker 1 (03:05):
Capital over labor. Okay, unpack that for us. What does
that mean in practice?
Speaker 2 (03:09):
Well, think about how developing economies have grown over the
last say, four decades. What was their main competitive advantage?
Speaker 1 (03:16):
Low cost labor, mostly manufacturing, call centers, that kind.
Speaker 2 (03:19):
Of thing exactly, that was the primary way millions were
lifted out of poverty. Countries built their development ladder.
Speaker 3 (03:25):
On that edge.
Speaker 2 (03:26):
But ai AI systematically erodes that.
Speaker 1 (03:29):
Edge because it replaces the chief human labor with.
Speaker 2 (03:32):
With computation with capital investment in servers and software, zero
wage computational capital effectively, so.
Speaker 1 (03:39):
The whole development model is being flipped on.
Speaker 3 (03:40):
Its head fundamentally.
Speaker 2 (03:42):
Instead of needing a large, relatively cheap workforce, countries now
need immense investment in things like compute power, specialized chips,
robust digital infrastructure just to stay in the game.
Speaker 1 (03:55):
Which most developing countries just.
Speaker 2 (03:56):
Don't have precisely. And that disparity, that sift and what's
needed is why the unc TAD Secretary General Rebecca grinspand
sounded the alarm so strongly.
Speaker 1 (04:05):
What did she say?
Speaker 2 (04:06):
Her quote was direct countries should act now. The message
is clear. The window to shape this is closing. If
we just wait and see then.
Speaker 1 (04:14):
That four point eight trillion dollar boom.
Speaker 2 (04:16):
It'll overwhelmingly flow to a handful of wealthy nations and
powerful corporations. The report warns this could lead to a
catastrophic global division. Our policy choices made right now will
decide if AI actually serves global progress or just deepens inequality.
Speaker 1 (04:33):
Okay, that perfectly sets up our mission for this deep dive.
Then we need to really get into what that forty
percent effective figure means. We need to map out that
inequality chasm. The report describes North versus South, capital versus.
Speaker 2 (04:45):
Labor, and crucially examine the solutions the pragmatic policy Playbook
unc TAD is proposing to actually manage this huge transformation.
Speaker 1 (04:55):
Right, let's start with a big number, the one causing
the most anxiety. That and you hear that number, and honestly,
the first reaction is probably panic images of mass unemployment cues.
It sounds almost like hyperbole, maybe designed to grab headlines.
Speaker 2 (05:09):
It's a natural reaction. And yeah, it's a startling figure.
But is it credible? I think the credibility comes from
the nuance and the methodology used by both UNCTAD and
the IMF who they cite.
Speaker 1 (05:18):
Okay, so how do they arrive at forty percent? What's
the method?
Speaker 2 (05:21):
It's not just counting job titles. It's based on these
occupational exposure models. They break down jobs into their individual tasks.
Speaker 1 (05:28):
Ah so, not is a lawyer automatable, but which tasks
that lawyers do.
Speaker 2 (05:34):
Are automatable precisely, it's much more granular and based on
that task analysis. They define affected in two ways. A
job is affected if AI can either complement the human
doing the.
Speaker 1 (05:46):
Work, meaning make them more productive, help them.
Speaker 2 (05:48):
Out exactly, boost their efficiency their output. Or the second way,
if AI can substitute for the human tasks.
Speaker 1 (05:56):
Leading to displacement or maybe the task just vanishing for
humans together.
Speaker 2 (06:00):
Right, So affected really means relevant to current or near
future AI capabilities. It's about the technical feasibility of applying
AI to the stuff people actually do in their jobs.
Speaker 1 (06:11):
Got it. So it's not a prediction of forty percent.
Speaker 3 (06:13):
Unemployment, definitely not.
Speaker 2 (06:14):
It's a measure of exposure, and as you'd expect, that
exposure varies massively depending on the type of economy. The
report clearly maps out three global tiers.
Speaker 1 (06:23):
Okay, let's break those down. Where's the exposure highest?
Speaker 3 (06:26):
Its highest?
Speaker 2 (06:26):
In the advanced economies they estimate a staggering sixty percent
exposure rate there.
Speaker 1 (06:31):
Sixty percent? Wow? Why so high? In the developed world?
You think they'd be more resilient.
Speaker 2 (06:36):
It's counterintuitive, isn't it, But it's because advanced economies are
primarily knowledge intensive. Their main functions revolved around processing information,
analyzing data, generating text finding patterns.
Speaker 1 (06:47):
Exactly the kind of cognitive work that large language models
and generative AI are getting really good at.
Speaker 3 (06:53):
You nailed it.
Speaker 2 (06:54):
This is where that in version of the script we
talked about really hits home. It's not just automating manual
laboring anymore.
Speaker 1 (07:00):
Give us some concrete examples. We're talking white collar jobs, right,
professional roles people thought were secure.
Speaker 2 (07:05):
Absolutely. Think about, say, junior lawyers. A lot of their
early work involves legal discovery, summarizing old cases, drafting basic
motions or sections of contract.
Speaker 4 (07:16):
The grunt work basically kinda yeah, But AI tools specifically
trained on legal data can now produce a solid first
draft of a complex brief, or sift through thousands of
discovery documents in minutes, not weeks.
Speaker 3 (07:30):
The human lawyer's role then shifts.
Speaker 1 (07:32):
From drafting it to checking it.
Speaker 2 (07:34):
Exactly, from creating to editing and verifying, which is still vital,
but it changes the whole structure.
Speaker 1 (07:40):
And it hits the training pipeline hard. Presumably, how do
you learn to draft If the AI does the.
Speaker 2 (07:45):
First pass, that's a massive challenge Those entry level roles
where you learn the ropes, they might just evaporate.
Speaker 3 (07:51):
Look at accounting too.
Speaker 1 (07:53):
Financial service is another huge sector in advanced economies.
Speaker 2 (07:56):
Right think about reconciling ledgers, basic tax prep, odd analysis patterns.
These are routine cognitive tasks. AI can ingest the data,
match things up, flag anomalies almost instantly.
Speaker 1 (08:08):
Even highly specialized fields aren't immune medicine.
Speaker 2 (08:12):
Even there, AI is already showing remarkable results, sometimes outperforming
experienced human radiologists and spotting subtle anomalies on X raysor
mammograms consistently.
Speaker 1 (08:21):
So the expert the radiologist becomes more of a supervisor,
a verifier of the AI's findings.
Speaker 2 (08:28):
Potentially, Yes, still a crucial role, requiring deep expertise, but
perhaps needing fewer hours overall, or maybe fewer specialists in
total for the same volume of work.
Speaker 1 (08:38):
And this aligns with other research I think you mentioned McKenzie.
Speaker 3 (08:41):
Yeah.
Speaker 2 (08:41):
A twenty twenty five Mackenzy update estimated that forty five
percent of tasks just within the finance and insurance sectors
alone are highly automatable with current tech forty five.
Speaker 1 (08:51):
Percent, So the substitution potential is hitting right at the
core of the global NORDS Economic Engine. Okay, let's contrast
that sixty percent with the next tier.
Speaker 2 (09:00):
That would be emerging markets. Here, the exposure rate drops,
but it's still significant forty.
Speaker 1 (09:06):
Percent, still very high. What kind of jobs are most
exposed there?
Speaker 2 (09:09):
It tends to be less complex work, often stuff that's
been standardized for outsourcing we're talking about. There's really large
scale customer service operations, call centers running on scripts.
Speaker 3 (09:19):
Yeah.
Speaker 2 (09:19):
Yeah, high volume data entry, basic tech support, level one stuff,
simple translation tasks. The backbone of the BPO business process
outsourcing industry for the last two decades.
Speaker 1 (09:30):
And this is where that thread of offshoring reversal really bites,
isn't it.
Speaker 2 (09:34):
It's a huge threat and the speed is alarming. Why
would a US or European company continue to offshore a
call center to say, Manila, or software testing to Bengaluru
purely for lower labor.
Speaker 1 (09:48):
Costs if they can deploy an AI chatbot or an
automated testing suite back home that costs virtually nothing per interaction.
Speaker 2 (09:55):
Exactly zero payroll, zero benefits, infinitely scalable. Age advantage isn't
just reduced, it's often completely neutralized by zero wage automation.
This could pull the rug out from under the economic
model of many middle income nations.
Speaker 1 (10:09):
Okay, then the final tier, low income countries. The exposure
rate there is the lowest.
Speaker 2 (10:15):
It is estimated at twenty six percent.
Speaker 1 (10:17):
Why lower? What protects them? Relatively speaking?
Speaker 2 (10:19):
It reflects their current economic structure. These economies are still
heavily reliant on manual physical labor, which AI and robotics
have historically found harder to replicate things like complex farming,
construction work, traditional crafts, resource extraction.
Speaker 1 (10:33):
But the reports suggest this is temporary protection very much.
Speaker 2 (10:36):
So unc TET is clear this relative safety won't last.
We're already seeing AI driven drones used in agriculture for
crop monitoring, and robotic harvesters are starting to appear in
large scale farming as robotics get better eyes and hands, basically.
Speaker 1 (10:52):
Better dexterity, better visual perception.
Speaker 2 (10:54):
Right, that twenty six percent figure is likely to climb
potentially quite rapidly in the coming year.
Speaker 1 (11:00):
So across all these tiers sixty percent twenty six percent.
The key question isn't just if a job is exposed,
but how that split.
Speaker 2 (11:09):
You mentioned earlier that fifty to fifty split between complementarity
and substitution That is the most critical distinction for anyone
thinking about policy.
Speaker 1 (11:16):
So roughly half the jobs exposed might actually get better.
Speaker 2 (11:19):
Potentially, yes, they become AI augmented. Think of a digital marketer.
They could use AI to sift through millions of customer commons,
understand sentiment instantly, and then deploy incredibly personalized ad campaigns
almost in real time. Their effectiveness skyrocket.
Speaker 1 (11:34):
Making them much more valuable exactly.
Speaker 3 (11:36):
But then there's the other half. Those jobs risk.
Speaker 2 (11:39):
Being diminished or outright eliminated because the AI can perform
the core task better, faster, or cheaper.
Speaker 1 (11:45):
Which brings us back to that inversion of the script.
It's not about replacing hands anymore. It's about replacing cognitive functions.
Speaker 2 (11:51):
It completely flips the historical relationship between skills and job security.
The Industrial Revolution mostly hit physical labor farm armours, artisans,
later automation factory assembly.
Speaker 1 (12:04):
Lines, but AI goes for the brain work.
Speaker 2 (12:06):
It targets cognitive labor. Those high skill, often high wage
roles involving complex pattern matching calculations, sophisticated language generation. They
are suddenly vulnerable. This demands a total rethink of education, training, everything,
and the.
Speaker 1 (12:21):
Speed of this change That's what makes this wave so different,
so much harder to manage than past disruptions.
Speaker 2 (12:27):
The velocity is unprecedented. We absolutely cannot overstate this pass shifts.
Even disruptive ones usually unfolded over decades. Society policy they
had some time, albeit often not enough to.
Speaker 1 (12:39):
Adapt, Like the Industrial Revolution.
Speaker 2 (12:40):
Example, right when steam power displaced say forty percent of
the UK's hand loom levers in the early eighteen hundreds.
That was brutal for those individuals, but the transition to
factory work took maybe fifty years. Even electricity in the
nineteen twenties automating around thirty percent of manufacturing cask happened
along inside the gradual growth of new industries over decades.
Speaker 1 (13:02):
But AI moves at the speed of the.
Speaker 2 (13:04):
Internet precisely, which creates this kind of political and social vertigo.
The report uses chat GPT as a benchmark, reaching one
hundred million users in just two months. That's faster than TikTok,
faster than Instagram, faster than any major technology adoption curve
we've ever seen.
Speaker 1 (13:20):
And what does that speed mean practically for workers and governments.
Speaker 3 (13:23):
It means the.
Speaker 2 (13:23):
Time lag between a job task being automated away and
a new compensating role potentially emerging is dramatically compressed. There's
less time to adjust, retrain, relocate. Policy can't afford its
usual bureaucratic pace. It needs to react almost in real time,
or the social friction, the human cost of this displacement
could become truly unmanageable.
Speaker 1 (13:43):
Okay, the speed is terrifying because, as you said, the
displacement isn't some future threat. It's happening now. The machine
is running. Let's look at the concrete examples UNCTAD gathered.
Where is AI actively eliminating roles today, starting with creative
fields which many thought were uniquely human.
Speaker 2 (14:02):
Yeah, that idea of creative work being somehow safe is
really being challenged. Tools like mid Journey, Deli stable Diffusion
they can generate high quality images, illustrations, even add concepts
from simple text prompts.
Speaker 1 (14:15):
So the need for junior designers' concept artists it's shrinking.
Speaker 2 (14:19):
Why hire a team for weeks to brainstorm ideas When
an AI can spit out fifty different, usable, legally clear
concepts in literally an hour. It structurally removes the need
for a lot of that initial entry level creative work.
Speaker 1 (14:32):
And in areas with huge volume. Like customer seemas, the
numbers are stark.
Speaker 2 (14:36):
The Klara example cited in the report is a perfect
quantifiable case. Their AI chatbot handled two point three million
customer conversations in just one month in twenty twenty.
Speaker 1 (14:45):
Four, two point three million, and the human equivalent.
Speaker 2 (14:48):
They estimated that workload was equivalent to about seven hundred
full time human agents. That wasn't about making agents more efficient.
That was a direct substitution to cut costs, plain and simple.
Speaker 1 (14:59):
And it's not just lower skilled roles you mentioned. High
value professional services are seeing this too.
Speaker 2 (15:03):
Law for instance, absolutely legal tech is booming. Systems like
Harvey AI, which we touched on, can draft contracts, analyze precedents,
write initial briefs reportedly ninety percent faster than junior lawyers
or pair.
Speaker 1 (15:17):
Of legals ninety percent faster. How does a human compete
with that on cost or time?
Speaker 2 (15:23):
They can't for those specific tasks. It fundamentally changes the
economics of legal work and again the entry point into
the profession. We even saw companies like Duo Lingo in
the language learning space cut ten percent of their human
contractors in twenty twenty four, explicitly saying AI could now
generate and refine lessons more effectively.
Speaker 1 (15:40):
So white collar knowledge work is clearly in the firing line.
But what about the physical world manufacturing logistics, That.
Speaker 3 (15:47):
Trend is also accelerating.
Speaker 2 (15:49):
We all remember the Foxconn story from twenty sixteen replacing
sixty thousand iPhone assemblers with robots. But the difference now
is that the robots, the cobots are getting.
Speaker 1 (16:00):
Smarter cobots, collaborative robots.
Speaker 2 (16:02):
Yeah, they have better vision systems, more dexterity, they can
handle tasks that aren't perfectly repetitive, making them viable replacements
even in smaller factories or for more customized production lines.
Speaker 1 (16:13):
And even in the tech industry itself, the heartland of
AI development.
Speaker 2 (16:17):
The evidence is right there in the layoff numbers. The
US tech sector shed around two hundred and sixty thousand
jobs in twenty twenty three and early twenty twenty four,
and significantly in about thirty percent of those layoff announcements,
major companies like Microsoft and Google specifically mentioned leveraging AI
as a factor enabling them to reduce headcount.
Speaker 1 (16:35):
So the corporate world is actively planning for this. It's
becoming strategy.
Speaker 2 (16:39):
It seems so the world Economic Forum survey found forty
one percent of companies globally planned some level of AI
driven workforce reduction within the next three years. This isn't
just speculation. It's baked into corporate planning based on expected
cost savings and efficiency gains.
Speaker 1 (16:54):
Okay, that's the displacement side, and it sounds pretty relentless.
But let's pivot. We have to ask about the other
side of the coin, the creation counterbalance. Where are the
new jobs? History usually shows and technology creates more jobs
than it destroys eventually.
Speaker 2 (17:10):
That's the historical pattern and it's the basis for a
lot of optimism. The classic analogy is the ATM.
Speaker 1 (17:16):
Right. ATMs replaced many bank tellers.
Speaker 2 (17:19):
But they also lowered the cost of running branches, so
banks open more branches, hired more loan officers, financial advisors,
it staff to manage the networks. Net Employment and banking
actually went up even though the specific role of the
teller declined.
Speaker 3 (17:33):
The jobs changed.
Speaker 1 (17:34):
So the hope is AI follows that pattern. What specific
new roles are emerging directly because of AI?
Speaker 2 (17:40):
Broadly, you can group them into a few categories. First,
there are the roles needed to manage and guide the
AI itself. Prompt engineers people who are experts at crafting
the right questions and instructions to get reliable, useful, ethical
outputs from these powerful, but sometimes unpredictable AI models.
Speaker 1 (17:57):
So prompt engineering, Why is that so crucial? If the
AI is supposedly smart.
Speaker 2 (18:01):
That's a great point. It's because these large models are generalists.
They know a lot, but they can also hallucinate, make
things up, or give answers that are technically right but
completely miss the business context. A good prompt engineer is
like a translator, bridging the gap between a vague human
need and the specific input the AI needs to deliver value.
They constrain the AI, guide it, focus it. It turns
(18:24):
raw AI power into a usable business tool.
Speaker 1 (18:27):
Okay, prompt engineers, what else?
Speaker 2 (18:28):
We definitely need AI at the CIS people specifically tasked
with auditing models for bias, fairness, and safety. And then
there's a huge need for data curators and AI trainers.
Speaker 1 (18:38):
Data curators like librarians.
Speaker 3 (18:40):
For data sort of yeah.
Speaker 2 (18:42):
Ensuring the vast data sets used to train these models
are clean, representative, unbiased, and legally compliant garbage in, garbage outright.
High quality data is paramount It's why LinkedIn listed AI
trainer as its fastest growing job title for twenty twenty five.
Someone has to prepare that data.
Speaker 1 (18:59):
Okay, so managing the AI are their entirely new industries
emerging too.
Speaker 2 (19:03):
Absolutely, think about the whole ecosystem that will be needed
to support autonomous vehicles. Maintenance technicians, remote monitoring centers, regulatory experts,
mapping specialists. Projections suggest this could create maybe a million
jobs in the US alone by twenty thirty.
Speaker 1 (19:19):
A whole new sector.
Speaker 2 (19:21):
Or consider personalized medicine. Using AI to analyze genomic data
or patient histories can lead to incredibly tailored treatments, but
you need experts, genetic counselors, specialized nurses, data analysts to
interpret those AI findings and integrate them into actual patient
care plans. These are human machine collaboration roles that barely
(19:41):
existed five years ago.
Speaker 1 (19:43):
And then there's the other half of that fifty to
fifty split we discussed augmentation, the people whose jobs aren't
replaced what are supercharged by AI.
Speaker 2 (19:50):
This is where a lot of the optimism lies. The
unc TAD report highlights examples like doctors using AI agnostic tools.
The AI handles the pattern recognition on scans faster, freeing
up maybe twenty percent more of the doctor's time.
Speaker 1 (20:03):
Time They can then spend on.
Speaker 2 (20:04):
On the human element, talking to patients, explaining complex conditions,
providing empathy, dealing with unique cases, the stuff AI can't do.
It potentially transforms the job into something higher value, more
focused on uniquely human skills, and maybe even more satisfying.
Speaker 1 (20:20):
Okay, so we have displacement but also creation and augmentation.
Sounds like it could balance out, maybe even be a
net positive. But there's a huge catch, isn't there.
Speaker 2 (20:28):
There's a massive catch. It's the reskilling challenge, the transition friction.
It's the gap between the jobs being lost and the
jobs being created.
Speaker 1 (20:36):
And bridging that gap requires what money?
Speaker 2 (20:40):
A lot of money and infrastructure. Unc TED puts a
number on it. They estimate empowering the global workforce for
this transition requires investing something like one trillion dollars globally
by twenty thirty. That's just for reskilling programs, education system upgrades,
training subsidies a trillion.
Speaker 1 (20:56):
Dollars A trillion. Are we seeing successful efforts to bridge
that gap up anywhere? Even on a smaller scale.
Speaker 2 (21:02):
We're seeing pockets of innovation. Singapore's Skills Future program is
often cited. It is every citizen credits like a voucher
they can use for approved training courses throughout their life.
They've subsidized AI related courses for hundreds of thousands of workers.
Speaker 1 (21:17):
Already proactive government support.
Speaker 2 (21:19):
Yes, Or look at Rowanda. They're focusing on very specific,
high value niche skills. They're training people in advanced drone
repair and maintenance, recognizing that's a growing field where they
can build expertise.
Speaker 1 (21:32):
But these seem like exceptions. The overall challenge remains immense right,
especially for older workers or those without advanced education.
Speaker 2 (21:40):
That's the harsh reality the report doesn't shy away from.
It's incredibly difficult for say a fifty year old former
data entry clerk or factory worker, maybe without a college degree,
to suddenly retrain as an AI alignment specialist requiring advanced
math and coding skills. The cognitive leap is just too
large for many people without significant sustain and support.
Speaker 1 (22:00):
So if we don't invest that one trillion dollars, if
we don't build those.
Speaker 2 (22:04):
Bridges, then the theoretical net job game becomes meaningless for
the millions who are displaced the benefits of AI, the
higher wages, the new opportunities will overwhelmingly flow to those
who already have the high skills and educational background needed
to grab them. It becomes a recipe for massive social
and economic polarization.
Speaker 1 (22:24):
Okay, So that failure to bridge the skills gap, combined
with the basic mechanics of AI, leads us straight into
the core warning of the unc TAD report. AI is
poised to dramatically worsen inequality both within nations and between
them unless we actively intervene.
Speaker 2 (22:41):
Exactly, and it starts with that core mechanism we mentioned.
AI favors capital over labor.
Speaker 1 (22:46):
Remind us how that works again, well, Building.
Speaker 2 (22:48):
And running advanced AI requires huge upfront investment in physical infrastructure.
We're talking massive data centers, enormous amounts of electricity, specialized
cooling systems, and critically those very expensive high ends the
graphics processing unit.
Speaker 1 (23:01):
It's computer chips, right.
Speaker 3 (23:03):
These are all capital assets.
Speaker 2 (23:05):
So when AI generates value, the primary financial returns the
profits flow directly back to the owners of that capital,
the companies that own the servers, the shareholders of the
chip manufacturers, not primarily to the human workforce involved, and.
Speaker 1 (23:19):
The most visible sign of this is the soaring value
of the company's making those chips.
Speaker 2 (23:23):
It's the clearest illustration. Look at Nvidia, their market value
hitting three trillion dollars in twenty twenty four. That wasn't
driven by selling more video games. It was almost entirely
fueled by the insatiable demand for their specialized chips needed
to train and run large AI models. That wealth accumulation
is incredibly concentrated.
Speaker 1 (23:43):
Okay, let's look at the impact within countries.
Speaker 3 (23:45):
First.
Speaker 1 (23:46):
The report talks about a K shaped recovery effect from AI.
What does that mean? Visualize the K for us?
Speaker 2 (23:52):
A K shape is a perfect metaphor for the growing divergence.
Imagine the letter K, the upper arms slanting upwards. That
represents the people in jobs where AI complements their skills significantly,
The augmentees exactly, the senior engineers fine tuning the models,
the data scientists designing the systems, the managers leveraging AI
insights for strategic decisions. These people become hyper productive, high
(24:15):
leverage employees. Their skills are in high demand, and their
scarcity drives up their wages dramatically.
Speaker 1 (24:21):
How much what kind of increases are we seeing.
Speaker 2 (24:24):
The report notes these high complementarity roles are seeing real
wage growth of twenty to thirty percent, sometimes even more.
Speaker 1 (24:30):
Twenty Your thirty percent is huge. How does AI actually
justify that kind of pay bump? For say, senior software engineer.
Speaker 2 (24:38):
Because their job transforms, they're no longer spending hours writing
routine code. The AI can handle a lot of that
boilerplate stuff. Instead, they focus one hundred percent of their
time on the really complex parts the system architecture, integrating
the AI into legacy systems, solving unique edge cases, overseeing
the AI's performance.
Speaker 3 (24:56):
They're essentially leveraging.
Speaker 2 (24:58):
The AI to do the work that might have previously
taken a team of five or ten junior coders. The
company captures that value, and the market rewards the leveraging
engineer with the significant chunk of it.
Speaker 1 (25:08):
Okay, that's the upward armor of the K. What about
the downward arm.
Speaker 2 (25:11):
That represents the stagnation or even decline for workers in
jobs where AI primarily substitutes for their skills, or where
the complementarity is very low. Think administrative assistance, basic data processors,
content writers producing routine articles, maybe some customer service.
Speaker 1 (25:27):
Roles their skills become less valuable or redundant.
Speaker 2 (25:30):
Exactly, demand falls, wages stagnate or even decrease in real terms.
The imf RAN simulations mentioned in the report that painted
a stark picture. AI could boost the incomes of the
top ten percent by around fifteen percent, while the bottom
fifty percent might see their average incomes actually dip by
five percent due to this pressure.
Speaker 1 (25:49):
A widening gap, and this divergence is reinforced by education
levels massively.
Speaker 2 (25:53):
Education acts like a gatekeeper here. The data, particularly from
the US, shows that jobs with high AI exposure do
do command a wage premium, maybe twenty five percent higher
pay than non exposed jobs. But and this is critical,
that premium flows almost exclusively to workers who have a
college degree. If you don't have that degree, AI exposure
is far more likely to mean substitution and wage pressure,
(26:15):
not augmentation and a pay rise. It hardens the socioeconomic divide, and.
Speaker 1 (26:19):
We also need to consider the impact on different demographic groups.
Right who holds these routine cognitive jobs that are most
vulnerable to substitution.
Speaker 2 (26:27):
Disproportionately These roles administrative support, data entry, some service jobs
are held by women and minority groups. Especially in economies
with less social mobility. So the automation of these jobs
risks heading already disadvantaged community's hardest, exacerbating existing inequalities and
making it tougher for them to access the resources needed
(26:48):
for reskilling.
Speaker 1 (26:48):
Okay, that's the inequality within nations. Now let's turn to
what U and C. TAD calls the real bombshell, the
widening inequality between nations, the AI divide. This is where
the geopolitical risk gets released.
Speaker 2 (27:00):
Serious, it is deeply alarming. The concentration of AI power
and resources is just staggering. The report found that just
one hundred companies, and get this, ninety percent of them
are based in either the US or China control about
forty percent of all global spending on AI research and development.
Speaker 1 (27:18):
One hundred firms controlling almost half the global R and D.
Speaker 2 (27:22):
And think back to that comparison we made earlier. The
combined market value of just three tech giants Apple, Microsoft,
Nvidia is greater than the entire annual economic output the
GDP of the entire African continent.
Speaker 1 (27:36):
Wow, just let that sink. In three companies versus a continent.
Speaker 2 (27:39):
That single data point tells you almost everything you need
to know about the concentration risk here.
Speaker 1 (27:43):
And this financial power is directly tied to control over
the essential infrastructure, the compute power.
Speaker 2 (27:49):
Compute power is the critical bottleneck. It's the new oil,
as some say, but unlike oil, it's not easily substitutable,
and the supply of the most advanced chips is incredibly centralized.
The US currently can controls access to around sixty percent
of the world's advanced semiconductor.
Speaker 1 (28:03):
Supply sixty percent, and Africa's.
Speaker 2 (28:05):
Share a devastating point one percent, zero point one. If
you don't have access to these high end chips, you
simply cannot train your own large sophisticated foundational AI models
tailored to your own languages, your culture, your specific economic challenges.
You're immediately dependent.
Speaker 1 (28:22):
Dependent on the models built in the north, which leads
to this idea of data colonialism.
Speaker 2 (28:27):
It's the modern digital equivalent of resource extraction, but maybe
even more insidious.
Speaker 3 (28:32):
Think about it.
Speaker 2 (28:33):
The raw material is the data generated every day by
people in developing countries. Their social media posts, their search queries,
their online interaction of data, Your data, my data, everyone's data.
That data is harvested, often with opaque consent, and used
to train the huge, powerful AI models primarily developed in
the US and China, and then those sophisticated AI services
(28:55):
are sold back globally, including to the very countries whose
data help build them. The profits flow north. The South
provides the raw material for free or nearly free, and
then pays to buy back the finished product. It creates
this perpetual cycle of dependency, and this.
Speaker 1 (29:09):
Whole dynamic fundamentally breaks the traditional development ladder that countries
used to climb. The report used Bangladesh's garment sector as
a key example.
Speaker 2 (29:18):
It's a powerful and worrying case study. Bangladesh's garment industry
is the backbone of its economy, worth around fifty billion
dollars a year, employing about four million people, mostly women.
It's their primary export earner, built entirely on that low
cost labor advantage.
Speaker 1 (29:35):
But AI threatens that how.
Speaker 2 (29:37):
Unc tad estimates that advancements in robotics, specifically AI powered
vision systems, combined with increasingly dexterous robotic arms basically AI
sewing bots, could automate tasks threatening up to sixty percent
of those garment jobs by twenty thirty.
Speaker 1 (29:52):
Sixty percent of four million jobs. That's potentially catastrophic is.
Speaker 2 (29:56):
If a country loses its main competitive edge, cheap scalable
labor to automation, before it has managed to transition its
economy and workforce towards higher skill, knowledge based industries. It
risks getting stuck trapped in poverty, unable to climb that
development ladder because the lower rungs have been automated away.
Speaker 1 (30:11):
And adding insult to injury. The report highlights that the
Global South is also being excluded from shaping the rules
of the game the governance aspect.
Speaker 2 (30:21):
Yes, the governance exclusion is critical. UNCTAD found that one
hundred and eighteen nations from the Global South were effectively
absent from key international discussions trying to set ethical guidelines
and standards for AI development and employment, things like the
G sevens Hiroshima AI process.
Speaker 1 (30:37):
So they don't get a say in the rules that
will govern a technology profoundly shaping their future.
Speaker 2 (30:42):
Exactly, if you don't have a seat at the table
when standards for data privacy, algorithmic fairness, safety protocols, or
even liability are being decided, you risk having rules imposed
on you that were designed by and for the Global North,
rules that might embed biases against your population, misunderstand your
cultural context, or simply not serve your development needs. You
(31:03):
become a rule taker, not a rule maker.
Speaker 1 (31:05):
Which brings us back to that huge pwcstinate. The report
mentions AI could add fifteen point seven trillion dollars to
global GDP by twenty thirty, But the AI divide means
that wealth won't be shared.
Speaker 2 (31:16):
That's the stark warning. Without deliberate intervention, the vast majority
of that fifteen point seven trillion dollar windfall will be
generated and captured in the north, primarily the US and China.
Instead of lifting all boats, AI could dramatically deepen global
inequality and instability. The stakes couldn't be higher.
Speaker 1 (31:35):
Okay, we've laid out the scale of the potential disruption,
that forty percent exposure and the profound risk of deepening inequality,
the AI divide and the K shape the UNCTAD report
isn't just about warnings. It's crucially about solutions. Let's move
to the case studies. They highlight the ground zero examples.
Speaker 2 (31:54):
Yes, these four cases are really important because they show
how different policy responses lead to different outcomes and make
the theory concrete and suggest a crucial pattern proactive policy
can genuinely soften the landing.
Speaker 1 (32:05):
Let's start with case one, India and its massive IT sector.
Speaker 2 (32:08):
Right, India's IT and business process management sector is a
global giant two hundred and fifty billion dollar industry employing
over five point four million people. It's been a huge
engine for middle class.
Speaker 1 (32:19):
Growth, but it's facing immediate disruption, particularly at the entry level.
Speaker 3 (32:23):
Exactly.
Speaker 2 (32:24):
AI coding assistance like GitHub, Copilot and others are becoming
incredibly proficient. They can now write maybe forty percent of
the routine boilerplate code automatically. Major Indian tech firms like
Infosis and Whipbrow have publicly stated they're aiming for high
levels of automation maybe fifty percent in areas like software
testing and maintenance.
Speaker 1 (32:43):
And the impact on jobs it's been.
Speaker 2 (32:46):
Felt immediately at the entry point. The report estimates that
around two hundred thousand fresher rolls those first jobs for
recent graduates effectively disappeared in the twenty twenty four to
twenty twenty five period because those tasks were automated.
Speaker 1 (33:00):
Two hundred thousand entry level jobs gone. That's a huge
blow to the pipeline. How did India respond? What's the
policy counterbalance.
Speaker 2 (33:07):
India's response has been quite rapid, largely focused on a
massive push for upskilling and reskilling, often through public private partnerships.
They've established large scale training hubs in major tech centers
like hydrobad and Bengaluru.
Speaker 1 (33:19):
Training people in what specifically, if coding.
Speaker 2 (33:21):
Is being automated, they're focusing on skills adjacent to AI
but less likely to be fully automated soon. A big
focus is on MLOPS machine learning operations MLOPS.
Speaker 1 (33:31):
We hear that term a lot. Explain why that's a
valuable pivot.
Speaker 2 (33:34):
MLOPS is essentially the practical side of AI deployment. It's
everything involved in taking an AI model developed in a
lab and actually getting it to work reliably and continuously
in the real world. This includes deploying the model, monitoring
its performance, maintaining it, retraining it when its performance degrades.
Speaker 1 (33:51):
Because AI models aren't static, they need ongoing work.
Speaker 2 (33:54):
Exactly, They suffer from model drift. Their accuracy can decline
as the real world data chain, so you need skilled
humans to constantly manage this life cycle. It's complex, it
requires expertise in both software engineering and data science, and
it's crucial for keeping these AI systems running effectively in businesses.
India recognized that while basic coding was vulnerable, managing the
(34:16):
deployed AI systems was a growing high value Niche They're
training maybe one hundred thousand people annually in.
Speaker 1 (34:22):
These hubs Okay shifting gears Case two Kenya and the
call center crunch. This illustrates that offshoring reversal we discussed right.
Speaker 2 (34:30):
Nairobi has built a reputation as Silicon Savannah, partly thanks
to its thriving BPO sector, employing around two hundred thousand people.
But AI voice spots are getting incredibly good, especially localized ones.
Speaker 1 (34:43):
Localized meaning they can handle local languages and.
Speaker 2 (34:45):
Accents precisely, AI bots that can flawlessly handle customer queries
in Swahili, understanding nuances far better than a generic global
bot could. This directly impacts the value proposition of Kenyon
call centers. Safari Coom, the big regional telecom operator, cut
about fifteen percent of its human call center agents in
twenty twenty four, explicitly citing the capabilities of these AI bots.
Speaker 1 (35:09):
Sounds like a familiar story automation hitting outsourcing, but ken
You found a different pivot.
Speaker 2 (35:13):
They did instead of just trying to protect the existing
call center jobs. They looked at moving up the AI
value chain through a government initiative called Ajira Digital.
Speaker 1 (35:22):
What does a geerdigital do?
Speaker 2 (35:23):
It focuses on training young Kenyans not to be call
center agents, but to become the human workforce behind the AI,
specifically in high quality data annotation, labeling and validation.
Speaker 1 (35:33):
So preparing the data that AI models.
Speaker 2 (35:35):
Need to learn exactly tasks like identifying objects in images
for computer vision systems, transcribing audio data accurately, or verifying
the quality of AI generated text. Kenya is essentially exporting
this labeled data as a service to global AI companies
like OpenAI and others. They're leveraging their linguistic skills and
(35:55):
local knowledge, turning it into a valuable input for the
global AI economy rather than just being replaced by it.
Speaker 1 (36:02):
Interesting pivot, Okay, Let's look at developed economy example K
three Germany and its middle stand. They seem to be
focusing more on augmentation.
Speaker 2 (36:10):
Germany's approach is quite different, heavily influenced by its strong
industrial base and its famous middle stand, the network of
small and medium sized manufacturing companies. Their strategy is very
clearly focused on AI augmentation, not displacement.
Speaker 1 (36:23):
How does that work? In practice?
Speaker 2 (36:25):
On a factory floor, they make heavy use of Cobot's
collaborative robots. These aren't designed to replace human workers entirely,
but to work alongside them. A cobot might hold a
heavy part in place while a human performs a complex weld,
or assist with lifting and moving materials. The result significant
productivity boosts. The report cites output increases of around thirty percent,
(36:47):
but crucially without large scale layoffs.
Speaker 1 (36:50):
And what policy levers enable this augmentation focus.
Speaker 3 (36:54):
Two main things.
Speaker 2 (36:55):
First, Germany's very strong tradition of vocational training the ousebuilding system,
which is at adapting skills. Second, targeted government policy. There
are significant subsidies available covering up to fifty percent of
the cost for companies to rescale their existing workforce to
work with these new robotic and AI tools, programming them, maintaining.
Speaker 1 (37:14):
Them, so they invest in the human alongside the machine.
Speaker 2 (37:16):
Precisely, the policy prioritizes enhancing human capital, ensuring workers adapt
and their value increases alongside the technology. This focus on collaboration,
backed by strong social partnership and investment, is a key
reason why German unemployment has remained remarkably low, around three percent,
despite technological change.
Speaker 1 (37:33):
Okay, one more case four the Philippines. This one highlights
a different vulnerability, the impact on digital gigwork and remittances.
Speaker 2 (37:42):
Yes, the Philippines is a major recipient of remittances from
overseas Filipino workers OFWs, and increasingly a significant chunk of
that income comes from online digital gigwork things like freelance, transcription,
virtual assistance, content moderation. Million of the families rely on
this income.
Speaker 1 (37:59):
In AI is hitting these gigs hard, very hard.
Speaker 2 (38:02):
Take online transcription platforms like rev dot com as AI
speech to text technology got dramatically better. These platforms automated
large parts of their workflow. When seventy percent of the
work disappears, the income stream for thousands of Filipino transcriptionists
collapses almost overnight.
Speaker 1 (38:18):
A sudden shock to household incomes. What's the policy response there?
Speaker 2 (38:21):
The Philippine government launched a program called two pad AI
in twenty twenty five. It's a large scale, targeted retraining
initiative specifically aimed at helping those displaced by automation in
the gig economy. The goal is to transition about fifty
thousand workers, particularly former transcriptionists, into higher value digital roles
(38:41):
like specialized virtual assistants, social media managers or data annotator.
Skills still in demand. They're actively trying to prevent the
value chain from collapsing entirely for these workers.
Speaker 1 (38:52):
So these four cases India, Kenya, Germany, Philippines, they really
underscore the UNCCAD conclusion, don't the outcome isn't predetermined by
the technology itself, not at all.
Speaker 2 (39:03):
It's driven by the structural policy choices that governments and
societies make in response to the technology. Proactive investment in skills,
infrastructure and social safety nets makes a huge difference between
a managed transition and a chaotic disruption.
Speaker 1 (39:16):
Which leads us perfectly into the specific policy recommendations from
unc TAD their proposed policy playbook for making AI more inclusive.
Speaker 2 (39:23):
Right, this is the roadmap they offer. It's pragmatic and
covers several key areas.
Speaker 1 (39:27):
Let's start with the foundation digital public infrastructure or DPI.
They cite India's Adhar system as a model. Why is
something like a national digital idea so crucial for AI.
Speaker 2 (39:39):
Equity Because DPI access the basic rails for participation in
the digital economy. A universal, secure digital identity system like Adhar,
which covers over one point three billion people in India,
enables access to services, finance, and importantly, it facilitates the creation.
Speaker 3 (39:56):
Of data commons.
Speaker 1 (39:57):
Data commons explain that concept. How is it different from
just making governm data open.
Speaker 2 (40:01):
Open data is often raw, messy, maybe inconsistently formatted. A
data commons is more structured. It's a shared pool of
high quality, ethically sourced, anonymized and standardized data. Think health records,
agricultural yields, traffic patterns, governed by clear rules.
Speaker 1 (40:15):
And why is that so important for developing countries in
the age of.
Speaker 2 (40:18):
AI Because it allows them to overcome the massive data
disadvantage they face. If you have a national data commons,
local researchers, startups and government agencies can access the high
quality data needed to train AI models that are relevant
to their specific context, models that understand local languages, local
crop diseases, local public health challenges.
Speaker 1 (40:39):
So it reduces dependency on the giant models trained on
northern data.
Speaker 3 (40:43):
Exactly.
Speaker 2 (40:43):
It helps level the playing field, allowing the global South
to build and benefit from its own AI ecosystems rather
than just being consumers of Western tech or suppliers.
Speaker 3 (40:53):
Of raw data.
Speaker 1 (40:54):
Okay, DPI is the foundation. The second pillar unc TAD
emphasizes is a skills revolution, and they suggest fronting it
by potentially taxing AI profits.
Speaker 2 (41:02):
Yes, this directly tackles that capital over labor imbalance. If
AI is generating massive profits for a relatively small number
of companies and capital owners. The argument is that society
via the state, need to recapture some of that windfall
to invest in the human.
Speaker 3 (41:17):
Capital of the broader workforce.
Speaker 2 (41:19):
The funding mechanism has to be there for lifelong learning
to become a reality, and they.
Speaker 1 (41:23):
Point to models like Denmark's Disruption councils what makes them effective.
Speaker 2 (41:27):
The Danish model is interesting because it's proactive and collaborative.
These councils bring together unions, industry representatives, and government officials.
Their job is to constantly scan the horizon identify which
sectors and skills are likely to be disrupted by technology
years in advance, so.
Speaker 1 (41:44):
They anticipate change rather than just reacting to.
Speaker 2 (41:46):
It precisely, and based on that foresight, they fund and
sometimes even mandate large scale retraining programs. Denmark aims to
retrain something like forty percent of its workforce every single decade.
Speaker 3 (41:59):
It's systemic.
Speaker 2 (42:00):
It accepts that skills have a shorter shelf life now
and requires continuous collective reinvestment as a national priority.
Speaker 1 (42:07):
Moving to the global level, the third pillar is global governance.
This includes mandating a seat for the Global South in
KEYAI bodies, but also that very ambitious proposal for an
AI solidarity fund.
Speaker 2 (42:18):
This is probably the most politically challenging but potentially transformative
idea in the report. The proposal is for a direct
global transfer mechanism.
Speaker 1 (42:27):
How would it work?
Speaker 2 (42:27):
The idea is to levy a small percentage. They suggest
one percent on the global revenues of the largest multinational
tech companies, the ones benefiting most from the digital economy
in AI.
Speaker 1 (42:38):
One percent of big tech revenue. That would be a
huge amount.
Speaker 2 (42:41):
Of money billions easily, and the proposal is to dedicate
that fund specifically to supporting developing countries, helping them build
their digital public infrastructure, fund their reskilling programs, improve their
access to compute power, essentially using a fraction of the
gains from digitalization to ensure a more equitable global transition.
Speaker 1 (43:01):
But the politics of implementing a global tax like that
seems incredibly difficult. How could it realistically be collected and managed?
Speaker 2 (43:08):
That's the trillion dollar question, isn't it. The report acknowledges
the huge hurdles implementing a global digital services tax requires
unprecedented multilateral cooperation, something that's historically been very hard to
achieve due to national interests and tax competition. It would
likely need a reformed international body, maybe the OECD or
(43:28):
a dedicated UN agency to oversee it. Perhaps nations would
collect the tax on revenues generated within their borders, then
pool it internationally for redistribution based on need and equity criteria.
Speaker 1 (43:40):
So it's less about the specific mechanism right now and
more about establishing the principle.
Speaker 3 (43:44):
I think.
Speaker 2 (43:45):
So establishing the principle that the enormous wealth generated by
global technologies like AI comes with a global responsibility to
manage the downsides and share the benefits more widely. It's
a call for fiscal solidarity in the digital age, okay.
Speaker 1 (43:58):
The fourth area is labor protections. As work changes, especially
with the rise of gigwork, how do we protect workers.
Speaker 2 (44:05):
With traditional stable employment potentially declining. The report stresses the
need for innovative social safety nets A key concept here
is portable benefits.
Speaker 1 (44:14):
Portable benefits explain what that means, especially for someone doing gigwork,
maybe across borders.
Speaker 2 (44:19):
Portable benefits mean that core social protections things like health
insurance contributions, retirement savings, unemployment insurance, sick leave, are tied
to the individual worker, not to a specific employer or job. So,
if you're a data labeler in Kenya working freelance contracts
for three different companies based in Europe in the US.
Speaker 1 (44:39):
Which traditional labor law struggles to.
Speaker 2 (44:41):
Cover exactly, portable benefits, perhaps managed by a government agency
or a third party platform, would allow you to accrue
these benefits consistently across all your gigs. When one contract ends,
you don't suddenly lose your health coverage or retirement contributions.
It provides a crucial layer of security and stability in
an increasingly fragmented way world of work. It decouples essential
(45:02):
security from the outdated notion of a single lifelong employer.
Speaker 1 (45:06):
The report also mentions exploring more structural changes shorter work
weeks UBI yes.
Speaker 2 (45:11):
These are seen as potential long term responses to productivity
gains from AI pilots like Spain's experiment with a thirty
two hour work week are testing if we can share
the productivity dividend by working less for the same day,
potentially creating more jobs overall. And universal Basic income UBI
pilots like those in Kenya or Finland are exploring whether
(45:32):
providing a regular, unconditional income floor can reduce the immense
financial stress of technological unemployment, giving people the breathing room
and security to retrain or pursue other paths.
Speaker 1 (45:44):
These are big ideas, still mostly experimental, very much so.
Speaker 2 (45:47):
But the report argues we need to be seriously exploring
and testing these concepts now, given the scale of the
potential disruption ahead.
Speaker 1 (45:53):
And finally, the playbook includes ethical AI mandates ensuring the
technology itself is developed and deployed responsibly.
Speaker 3 (46:00):
This is fundamental.
Speaker 2 (46:02):
If the tools themselves are biased or unsafe, none of
the other policies matter as much. Key recommendations include mandating
a human in the loop for high stakes decisions, meaning
for critical applications think medical diagnoses, loan applications, hiring decisions,
judicial sentencing recommendations. The AI can provide analysis or recommendations,
but a human being must always retain the final decision
(46:24):
making authority and accountability the AI assists it doesn't.
Speaker 1 (46:29):
Command, and auditing algorithms for bias absolutely crucial, especially since
many foundational models are trained predominantly on data from North
America and Europe.
Speaker 2 (46:38):
When these models are deployed globally, they can inadvertently perpetuate
or even amplify existing biases related to race, gender, language,
or cultural norms found in that training data. Rigorous independent
auditing is needed to detect and mitigate these biases before
they cause harm in diverse contexts.
Speaker 1 (46:56):
It's encouraging. I suppose that the report also notes some
movement from the corporate site big tech making pledges.
Speaker 2 (47:02):
Yes, there is some alignment, though you could argue it's
often driven by enlightened self interest. Microsoft pledging four billion
dollars for AI skilling initiatives globally, Google setting up AI
research hubs in Africa. These companies recognize that they need
a skilled global workforce to use their products, and that
widespread social instability caused by automation is ultimately bad for business.
(47:25):
Talent scarcity and market disruption are risks for them too,
So while it's not purely altruistic, these corporate commitments are
a necessary part of the solution mix hashtag tag tag
tag outro.
Speaker 1 (47:36):
So we've covered a lot of ground, the sheer scale
of the AI revolution, the potential four point eight trillion
dollar market, the forty percent job exposure warning, the deep
risks of inequality amplification, but also crucially the policy pathways
unc TAB proposes for a more equitable future.
Speaker 2 (47:53):
It really brings us to the philosophical heart of the debate,
doesn't it, Because even among experts there's still a deep
split on what the ultimate outcome will be.
Speaker 1 (48:00):
The optimists on one side.
Speaker 2 (48:01):
Right, the techno optimists, people like Eric Brynjolfson at Stanford.
Their core argument rests on history. They say, look, every
major technology shift in the past, steam electricity, computers destroyed
some jobs, but ultimately created far more new jobs and
new industries. Complementarity, they argue, always wins out over substitution in.
Speaker 1 (48:21):
The long run, and they have data to back that up.
Speaker 2 (48:24):
They point to studies like a recent one from MIT
in twenty twenty five showing that companies actively using AI
to augment their workers actually increase their overall hiring rate
by about eleven percent. For optimists, that four point eight
trillion dollars prize isn't a threat, It's a resource that
human ingenuity will inevitably harness for broad progress.
Speaker 1 (48:42):
But then you have the other side, the skeptics.
Speaker 2 (48:45):
The skeptics perhaps best represented by Carl Benedict Frey at Oxford.
Their counter argument is simple, but powerful history might not
be a reliable guide this time because one key variable
is different velocity, the sheer speed of AI development and deployment.
Speaker 1 (48:58):
They worry the changes too fast exactly.
Speaker 2 (49:01):
Frey warns of a real possibility of persistent technological unemployment,
not because new jobs won't eventually emerge, but because the
pace of job destruction might dramatically outstrip society's ability to adapt, retrain,
and absorb displaced workers into those new roles. If the
skills gap is too wide and the transition support is
too slow, we could face a prolonged period of social
(49:23):
disruption and pain, even if the long term picture is positive.
Speaker 1 (49:26):
So where does the unc TAD report land between those
two poles, optimism or skepticism.
Speaker 2 (49:32):
I'd say it carves out a pragmatic, action oriented middle ground.
It acknowledges the immense potential of AI, that potential fifteen
point seven trillion dollars boost a global GDP by twenty
thirty is real, but it insists that realizing that potential
equitably is not automatic. It requires deliberate political choices.
Speaker 1 (49:48):
It's not technodeterminism, but it's about policy precisely.
Speaker 2 (49:52):
The report essentially says, we have the potential resources generated
by AI to fund almost any necessary transition, universal basic services,
global reskilling, on an unprecedented scale, green energy infrastructure.
Speaker 3 (50:06):
The money could be there.
Speaker 2 (50:07):
The fundamental question is whether we globally and nationally have
the political will to implement the mechanisms like that AI
solidarity fund, like robust digital public infrastructure, like mandatory lifelong
learning systems needed to democratize access and distribute the benefits fairly.
Speaker 1 (50:25):
We started this deep dive talking about the human scale
that forty percent figure. Let's end by bringing it back
to the individuals caught up in this wave, because policy
ultimately has to work for that.
Speaker 3 (50:35):
Absolutely.
Speaker 2 (50:35):
The report is filled with data, but behind the percentages
are real people's lives. Think of Maria, who they might
mention hypothetically, a thirty eight year old transcriptionist in Manila.
Speaker 1 (50:43):
Whose gigwork vanished overnight due to automation.
Speaker 2 (50:46):
Right economic stability dawn just like that. But if policy
provides a bridge, maybe access to government funded training credits,
she could potentially enroll in an online course, maybe get
a certificate in AI data moderation, and perhaps find a
new role, maybe even earning twenty percent more. Her story
becomes one of managed transition enabled by support.
Speaker 1 (51:07):
Or think of rejestion in Bengaluru.
Speaker 2 (51:08):
Maybe he was a software quality assurance tester. His job,
stable and middle class gets automated away by sophisticated testing bots.
Speaker 3 (51:16):
Huge pressure.
Speaker 2 (51:17):
But again, if there are accessible government backed programs, maybe
he gets into a cybersecurity boot camp run by a
company like Tata. He pivots finds stability in a related
but higher demand field.
Speaker 1 (51:28):
These individual stories, even if hypothetical, they illustrate the point.
Speaker 2 (51:32):
They show that disruption doesn't have to mean destitution. It
can be managed, navigated, but it requires empathy, foresight, and
critically concrete action and investment in people.
Speaker 1 (51:41):
So the four point eight trillion dollar AI prize is there.
It could fund a global transformation, a creative renaissance, tackle
climate change.
Speaker 2 (51:49):
It could, but only if we make the conscious choice
to democratize access, manage the risks, and share the rewards.
Speaker 1 (51:55):
Because the alternative if we ignore that forty percent warning,
If we let the AI I divide deepen.
Speaker 2 (52:01):
The alternative is a deeply fractured world, A world of
AI augmented elites concentrated in a few regions controlling the
new means of production, and a vast global population facing
precarious work, stagnant wages, and diminishing opportunities. A world of
haves and have nots on a scale we've never seen before.
Speaker 1 (52:21):
But the future isn't set in stone.
Speaker 3 (52:23):
Not at all.
Speaker 2 (52:24):
The report's ultimate message is one of agency. The future
isn't pre coded by the technology. It will be shaped
by the policy choices we make or fail to make.
Right now, we are programming our economic destiny.
Speaker 1 (52:36):
Which brings it back to you. The listener, Rebecca Grinspan
at you n c Ted said, countries must act now
for global progress. If you had to push just one
policy lever first, based on everything we've discussed to nudge
us towards that more equitable future, what would it be.
Speaker 2 (52:50):
Is it investing massively in lifelong learning within your own nation,
perhaps funded by taxing AI profits, or is it pushing
for that ambitious, maybe idealistic global AI Solidarity fund to
ensure countries in the Global South.
Speaker 3 (53:04):
Aren't left behind, that they have.
Speaker 2 (53:06):
The basic tools like compute access and DPI to participate.
Speaker 1 (53:10):
Something to think about. The choices are complex, the stakes
are enormous, and the time to act is well now, we.
Speaker 3 (53:15):
Really must act now. The choices ours food for thought.
Speaker 1 (53:19):
Indeed, thank you for joining us for this deep dive.
We'll see you next time.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com