August 16, 2025 • 28 mins
What if the smartest minds in AI just told us the future… and it’s coming way faster than anyone expected? 🤯
In this electrifying deep dive, we break down a powerful discussion between former OpenAI Chief Scientist Ilya Sutskever and former Google CEO Eric Schmidt — two insiders with front-row seats to the AI revolution.
Sutskever pulls no punches, warning that AI will eventually perform every human task, and that we’re standing at the edge of a transformation unlike anything in history. Schmidt takes it further — predicting:
Within 1 year: AI will replace most programmers 💻
Within 3–5 years: We’ll see Artificial General Intelligence matching the smartest humans across every field 🧠
Within 6 years: Artificial Super Intelligence will surpass all human intelligence combined 🌍
Along the way, we explore agentic AI solutions — self-directed systems that can automate multi-step, complex processes without human help. Both experts warn: this pace of change could reshape jobs, economies, and society itself — and waiting to “deal with it later” is no longer an option.
Packed with insider insights, jaw-dropping timelines, and unsettling truths, this is the AI future you need to understand — before it arrives at your doorstep.
🎧 Listen now, share with someone who still thinks AI is “years away,” and be ready for the most disruptive decade in human history.
Become a supporter of this podcast: https://www.spreaker.com/podcast/tech-threads-sci-tech-future-tech-ai--5976276/support.
You May also Like:
🎁The Perfect Gift App
Find the Perfect Gift in Seconds
⭐Sky Near Me
Live map of stars & planets visible near you
✔Debt Planner App
Your Path to Debt-Free Living
In this electrifying deep dive, we break down a powerful discussion between former OpenAI Chief Scientist Ilya Sutskever and former Google CEO Eric Schmidt — two insiders with front-row seats to the AI revolution.
Sutskever pulls no punches, warning that AI will eventually perform every human task, and that we’re standing at the edge of a transformation unlike anything in history. Schmidt takes it further — predicting:
Within 1 year: AI will replace most programmers 💻
Within 3–5 years: We’ll see Artificial General Intelligence matching the smartest humans across every field 🧠
Within 6 years: Artificial Super Intelligence will surpass all human intelligence combined 🌍
Along the way, we explore agentic AI solutions — self-directed systems that can automate multi-step, complex processes without human help. Both experts warn: this pace of change could reshape jobs, economies, and society itself — and waiting to “deal with it later” is no longer an option.
Packed with insider insights, jaw-dropping timelines, and unsettling truths, this is the AI future you need to understand — before it arrives at your doorstep.
🎧 Listen now, share with someone who still thinks AI is “years away,” and be ready for the most disruptive decade in human history.
Become a supporter of this podcast: https://www.spreaker.com/podcast/tech-threads-sci-tech-future-tech-ai--5976276/support.
You May also Like:
🎁The Perfect Gift App
Find the Perfect Gift in Seconds
⭐Sky Near Me
Live map of stars & planets visible near you
✔Debt Planner App
Your Path to Debt-Free Living
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
You may not take interest in politics, but politics will
take interest in you. It's a classic observation, right it is.
But what if we told you that the exact same
principle applies maybe many times over to something far less tangible.
Yet it's increasingly woven into well everything, every aspect of
our lives day by day.
Speaker 2 (00:19):
You're talking about AI, artificial.
Speaker 1 (00:21):
Intelligence exactly AI.
Speaker 2 (00:23):
And it's a bold claim people make, isn't it that
we are living in the most unusual time ever? People
will say that sure about their own era.
Speaker 1 (00:30):
Right, happens all the time.
Speaker 2 (00:31):
But this time, it genuinely feels different. And the reason,
like you said, is profoundly tied to artificial intelligence. We're
already seeing it, aren't we.
Speaker 1 (00:40):
Oh absolutely?
Speaker 2 (00:41):
I mean from what we hear, what we observe, AI
has already fundamentally shifted what it means to be, say
a student. It's impacting work careers in ways that are
still unknown, honestly and largely unpredictable.
Speaker 1 (00:53):
I bet many people listening right now are already asking themselves,
maybe with a bit of anxiety, okay, which skills are
going to remain useful, which ones are going to be obsolete?
How do I even adapt That's the core question for
so many, and.
Speaker 2 (01:06):
That's exactly where we are right now, in this kind
of powerful moment of questioning of transformation. You can help
on to social media, see the latest AI demos, or
just look at how your own work might be shifting,
and you get that feeling, like sort of visceral sense
that something monumental is unfolding. We've all interacted with AI
in some way, probably.
Speaker 1 (01:26):
Sure, spoken to a voice assistant, yeah, had it speak.
Speaker 2 (01:29):
Back, seeing it generate text, maybe even write code. It's
already pretty wild, you know, it is, even with its
current quirks and deficiencies, because it sparks this kind of
undeniable intuition about what's coming next. It's future potential. But
here's the kicker, and this is what we really want
to dig into today. Okay, the real challenge that AI
(01:50):
isn't necessarily what we're seeing day to day right now.
It's something yeah, well, the experts say it's unprecedented, extreme
and vastly different from its current form, hints at of
future that's kind of hard to fully grasp precisely, And
that's really the mission for this deep dive today. We're
not here to just skim the surface or throw a
(02:10):
bunch of technical jargon at you. Instead, we want to
take you on a journey through the insights and frankly,
even the stark warnings from some of the leading minds
right at the forefront of AI.
Speaker 1 (02:21):
Research, people who are building this stuff exactly.
Speaker 2 (02:24):
The goal isn't to overwhelm you. It's to equip you
with a more nuanced understanding of these really profound future
possibilities and crucially, the critical questions they raise for all
of us. Think of it like your essential shortcut to
getting informed on a topic that isn't just going to.
Speaker 1 (02:42):
Shape your world, it already is.
Speaker 2 (02:44):
It's actively shaping it right now.
Speaker 1 (02:45):
Okay, so let's unpack this foundational idea, this sort of
core logic that underpins so much of the thinking in
the AI space. It starts with a deceptively simple argument,
but it's incredibly powerful. The human brain is at its
core a biologue computer, right It processes information, learns, adapts.
Speaker 2 (03:04):
Creates, It has amazing things.
Speaker 1 (03:05):
Amazing things We learn languages, solve super complex problems, create art,
develop theories. So if a biological computer, our brain can
achieve all this, every cognitive function, every creative leap every
bit of logic, then why logically couldn't a really advance
digital computer, a digital brain do the same, and not
just some of it exactly, not just some but anything
(03:28):
we can learn to do.
Speaker 2 (03:29):
It really is the one sentence summary, isn't it Huh?
If we with our organic wetwear can learn and do
this incredible range of tasks, right, then a digital system
free from biologies limits should theoretically be able to do
the same and probably eventually surpass us.
Speaker 1 (03:46):
And that premise, that fundamental idea, immediately brings us to
this really dramatic, almost unsettling question, one that many people
find I think incredibly difficult to fully internalize or even
just emotionally believe. Which is what happens when computers can
do all of our jobs, not just automate a few
tasks here and there, but truly take on every cognitive
function we perform in our professional lives.
Speaker 2 (04:07):
That's the big one. Isn't it feels almost yeah, dystopian
like you said to even say it out loud.
Speaker 1 (04:12):
I remember hearing someone compare it to asking a horse
what it'll do when cars take over the horse just
they can't even conceive of it.
Speaker 2 (04:20):
Right, But the logic, according to these experts holds. And
then what happens after that? What do we as a
society want to use these incredibly capable AIS for good question? Well,
the answer, at least initially seems to be more work,
more economic growth, vastly accelerated r ND, and maybe most importantly,
(04:42):
doing AI research itself.
Speaker 1 (04:43):
Okay, now that's where it gets really interesting.
Speaker 2 (04:45):
Exactly because if AI can significantly contribute to its own
research and development.
Speaker 1 (04:50):
Then the rate of progress just takes off. It becomes,
as they say, really extremely fast for some period. It's
this kind of unprecedented feedback loop.
Speaker 2 (04:57):
And accelerating intelligence where the intelligence it self drives its
own exponential improvement. It learns how to learn faster.
Speaker 1 (05:04):
These are such extreme scenarios, though almost unimaginable for many people,
and it highlights what you called this significant intuition gap.
Speaker 2 (05:13):
Even people deep in the field like Ilia Sutzkiver, you know,
one of the key figures behind open AI. He admits,
they struggle to fully internalize this future on an emotional level,
even as the cold hard logic dictates it's very likely
to happen.
Speaker 1 (05:28):
So the head says yes, but the gut says, whoa.
Speaker 2 (05:32):
Kind of He argues that no amount of essays or
papers or explanations, no matter how detailed, none of it
can truly compete with what we actually see with our
own senses.
Speaker 1 (05:42):
Like trying to explain the Internet to someone in the
nineteen fifties.
Speaker 2 (05:45):
Exactly, you can describe the function, but the reality, the
paradigm shift, it's almost impossible to grasp without living it.
We humans are good at linde, your thinking, terrible at exponential.
Speaker 1 (05:55):
That's a really powerful point, and it makes sense our
current interactions with AI, even the perfect ones asking Syria
question seeing a weird AI generated image, they're already building
that intuition bit by bit. As AI keeps improving day
by day, month by month.
Speaker 2 (06:10):
That intuition gets stronger. Those once imaginary scenarios, the stuff
of old sci fi.
Speaker 1 (06:14):
Yeah, they become much more real, less theoretical, more visceral.
It's a progression from abstract idea to tangible reality, and
it shapes how we think, maybe without us even realizing
how much our mental models are shifting.
Speaker 2 (06:30):
Which brings us right back full circle to that opening analogy.
Speaker 1 (06:33):
Politics takes an interest in you, right.
Speaker 2 (06:36):
Just like politics AI will affect your life to a great,
great extent, whether you like it or not, whether you're
paying attention or not.
Speaker 1 (06:43):
So the main thing, then is what.
Speaker 2 (06:45):
The main thing is to pay attention to actively look
at what it can do, try to understand it and
generate the collective energy we're going to need to solve
the profound challenges that will come up.
Speaker 1 (06:55):
Because this isn't just a challenge. Leading thinkers people like
Eric Schmidt, former Google CEO, they're positioning it as potentially
the greatest challenge of humanity ever.
Speaker 2 (07:04):
But also, and this is crucial, they argue that successfully
navigating it will bring the greatest reward. It's this immense
double edged sword. It demands our full attention, our collective ingenuity,
and frankly, a willingness to confront questions we've never had
to ask before.
Speaker 1 (07:19):
Okay, so if that's the grand, maybe slightly terrifying, philosophical future,
let's zoom in a bit. What does this mean in
the near future, like the next one to two years.
Speaker 2 (07:30):
Right, the immediate landscape?
Speaker 1 (07:31):
Yeah, because this timeline might genuinely surprise people, especially if
you're in certain fields or your job involves a lot
of what we think of as knowledge work.
Speaker 2 (07:41):
The industry consensus here, and this is coming from figures
like Schmidt and many AI researchers, is pretty stark. Take
programmers for instance, Okay, within just one year, the vast
majority of generalist programmers are expected to be replaced by
AI programmers one year.
Speaker 1 (07:57):
Wow, why so fast?
Speaker 2 (07:59):
Well, there are clear reasons. Programming fundamentally uses a simpler,
more structured, more logical language than human language. Sure, these
AI algorithms excel at something conceptually similar to word prediction,
but on a massive scale. They optimize what's called the
loss function, which.
Speaker 1 (08:14):
Is like their internal score for how wrong they are exactly.
Speaker 2 (08:17):
It's their greade book. They want to minimize that loss,
get it close to zero. They refine their code, test it,
refine it again, millions of times faster than any human
could write and debug. They just keep iterating until the
code passes all the tests, all the requirements.
Speaker 1 (08:32):
That's fascinating. So the structure of code itself makes it
easier for AI to master. And it's not just coding, right,
you mentioned Schmidt points to mathematicians too.
Speaker 2 (08:41):
Indeed, similarly, within one year, AI is projected to produce
graduate level mathematicians at the absolute tippy top of programs,
the very best. The reasoning is similar. Math also uses
a highly structured, simpler language, albeit abstract. AI can use
conjecture proof formats can work through protocols like Lean, which
(09:01):
is a formal proof assistant.
Speaker 1 (09:03):
So it's not just calculation, it's actual proof verification.
Speaker 2 (09:06):
Yes, Lean helps verify proofs rigorously. So AI isn't just
proposing solutions, it's constructing and confirming mathematically sound, verifiable arguments,
generating genuinely novel mathematical insights.
Speaker 1 (09:18):
Potentially, so for programmers then this means the specific language
they use, Python, Java, whatever, might soon not matter nearly
as much because the goal is the outcome, and the
AI generates the code.
Speaker 2 (09:32):
That's the implication. It's a huge philosophical shift in the craft.
It moves from the mechanics of writing syntax to the
higher level art of defining the outcome, crafting the perfect
prompt or instructions.
Speaker 1 (09:44):
For the AI, from syntax to semantics.
Speaker 2 (09:46):
Absolutely, from the how to the what and why. And
this leads right into another critical development happening now, the
dawn of recursive self improvement.
Speaker 1 (09:55):
Okay, that sounds like sci fi, it does.
Speaker 2 (09:57):
But it's not anymore. This is where AI generates code
for its own research programs. It's writing code to make
itself smarter.
Speaker 1 (10:03):
And there's evidence this is actually happening.
Speaker 2 (10:05):
Yes, it's not theoretical. Top research groups like open AI
and Tropic they're already reporting to ten percent, maybe twenty
percent of their new experimental research code is being generated
by the computer itself.
Speaker 1 (10:16):
That's genuinely mind boggling. Think about that. If it's ten
twenty percent now, when it's still relatively early, what happens
when that scales fifty percent, eighty percent, one hundred percent.
Speaker 2 (10:26):
It's not just optimizing tasks, it's accelerating its own evolution,
its own intelligence, at a pace humans just can't match
a self amplifying engine of discovery.
Speaker 1 (10:37):
And this isn't the only immediate shift. We also need
to talk about agentic solutions. Agents.
Speaker 2 (10:42):
Yes, agents a term that gets thrown around a lot,
sometimes loosely, but in this context it means truly autonomous systems.
They have input output persistent memory, and they learn, they
observe something, understand it based on past data, remember it,
and then take intelligent, multi step actions towards a goal.
It's about automating entire processes, not just single steps.
Speaker 1 (11:04):
Okay, let's make that concrete. The expert use a great example.
Buying a house, super complex process, right.
Speaker 2 (11:10):
Imagine an agent system handling that start to finish. It
begins by finding a house in a specific area you want,
say McLean Virginia, based on your preferences. Okay. Then it
automatically pulls up and analyzes all the local rules zoning laws,
instantly figuring out how big a house you can actually
build there.
Speaker 1 (11:26):
Wow.
Speaker 2 (11:26):
After that, it handles the whole transaction to buy the land,
negotiation paperwork the lot.
Speaker 1 (11:31):
And it doesn't stop there at all.
Speaker 2 (11:32):
Once the land secured, the agent could design the house itself,
maybe collaborates with a human architect for the final touches
the esthetic choices.
Speaker 1 (11:41):
But it does the bulk of the design.
Speaker 2 (11:42):
Work largely independently. Yeah, then you prove the design. After that,
it autonomously finds and hires the best contractor for the job,
manages all the payments for materials, labor, tracks the.
Speaker 1 (11:54):
Invoices, and the slightly funny part, and yes.
Speaker 2 (11:58):
Quite tellingly, it even has the capability to initiate legal
action to sue the contractor for lack of performance. If
things go south and they don't hold up their end
of the bargain.
Speaker 1 (12:09):
I love that detail suing the contractor. It just perfectly
shows how comprehensive these systems are meant to be.
Speaker 2 (12:16):
And to end, it's not just one piece. It's stitching
together a whole chain of complex, interdependent decisions and actions.
Speaker 1 (12:23):
But here's the key insight from that example, something Eric
Schmidt really emphasized. This isn't just about buying a house
or replacing programmers. That single analogy, believe it or not, describes,
in his words, every business process, every government process, and
every academic process in our nation.
Speaker 2 (12:39):
Think about that, finding information, executing complex tasks, financial transactions,
even litigation. It suggests a fundamental shift in how all
tasks get done, not just tech but everywhere.
Speaker 1 (12:53):
It truly is a foundational change. And what's striking is
the timeline locking in. They say within the next year
or two, as.
Speaker 2 (13:00):
The expert stress, we're not going to stop it. This
isn't just about jobs and specific sectors. It's a complete
reimagining of how work broadly defined, gets done. It goes
beyond current automation, which also just streamlines human processes. This
is creating new often self optimizing processes.
Speaker 1 (13:17):
The implications are yeah, profound, far reaching. It could alter
the very base of how our economy and society function.
So let's look further out now, midterm horizon Artificial general
intelligence AGI projected for the next what three to five
years roughly?
Speaker 2 (13:32):
Yeah, No definitions vary.
Speaker 1 (13:33):
Okay, So what's the core idea we should grasp? The
most important characteristic?
Speaker 2 (13:37):
The basic idea discussed for maybe fifteen twenty years in
AI circles is the point where an intelligence system has
the flexibility of a human flexibility. What really distinguishes AGI
from current narrow AI, which is powerful, sure, but always
directed by a human toward the specific goal is that
AGI implies that computer can generate its own objective of function,
(14:00):
its own goals, its own strategies.
Speaker 1 (14:03):
So it's not just executing tasks, it's defining the tasks,
deciding what's important on its.
Speaker 2 (14:08):
Own, exactly, self directed, autonomous in its intellectual pursuits. That's
the game changer.
Speaker 1 (14:13):
And there's this thing called the San Francisco consensus around
this sounds almost like a local.
Speaker 2 (14:17):
Cult, huh. Indeed, it's a specific belief held strongly by
many researchers. There deep in the AI hubs.
Speaker 1 (14:24):
Maybe it is the water.
Speaker 2 (14:25):
Maybe, But seriously, they've largely convinced themselves based on observing
these exponential scaling laws and the rapid progress that AGI
will arrive within two to three cranks cranks. A crank
in their lingo is about eighteen months, So their timeline
for AGI is roughly three to four and a half
years from now.
Speaker 1 (14:41):
That's incredibly soon. And their definition of AGI potent intelligence
greater than the sum of human intelligence.
Speaker 2 (14:49):
Not just as smart as one human, but smarter than
all human brains combined working together.
Speaker 1 (14:55):
Wow, greater than the sum of human intelligence. That's hard
to even wrap your head around. So I hear not
everyone agrees on that specific timeline. Even among the experts.
Speaker 2 (15:04):
You're absolutely right, there is debate. While most agree AGI
is highly likely eventually, the exact timeline is uncertain. Some
think it might take longer, maybe unforeseen scaling issues, the
difficulty of real world interaction, or just achieving genuine self
direction beyond complex pattern matching.
Speaker 1 (15:22):
So the consensus is more like, we don't know exactly when.
Speaker 2 (15:26):
Pretty much, it highlights how unpredictable these huge leaps are.
But regardless of whether it's three five, maybe eight years.
The implication of AGI once it arrives is staggering.
Speaker 1 (15:36):
And that implication is vividly called the smartest human in
your pocket. Explain that because it sounds revolutionary on a
very personal.
Speaker 2 (15:43):
Level, it means quite literally, every single person, regardless of
background or wealth, could have access to the equivalent of
the smartest human being whoever lived or could live, available
instantly for every problem, every problem. Imagine needing an architecture
design boom. You have the best architect imaginable right there,
facing a complex math problem. The greatest mathematician in history
(16:07):
is at your fingertips. Need help with writing science, the
most brilliant writer or scientist on call, on demand. It
completely democratizes expert level intelligence, makes it universally accessible. It
could flatten knowledge hierarchies and ways we can barely conceive,
transforms problem solving from a rare, specialized resource to something ubiquitous.
Speaker 1 (16:28):
Okay, So, if AGI is midterm, what's beyond that? The
long view? Artificial superintelligence ASI projected maybe six years out plus.
Speaker 2 (16:36):
By the more aggressive timelines. Yes, ASI refers to computers
that aren't just smarter than one human or even the
sum of humanity but smarter than the sum of humans combined,
a qualitative leap beyond AGI.
Speaker 1 (16:47):
So it could perform any intellectual task a human can,
but just vastly better, faster, more memory, deeper cognition.
Speaker 2 (16:55):
That's the idea, and the San Francisco Consensus, again based
purely on scaling projects, suggests it could arrive within six years.
Speaker 1 (17:02):
Six years for something that could completely alter human civilization.
But there's a huge real world constraint here, isn't there
The energy required.
Speaker 2 (17:13):
Absolutely, This isn't just abstract code. It needs power, massive
amounts of it. A critical practical requirement for ASI with
current tech is an enormous amount of electricity. How much
are we talking significant gigawatts, like the power consumption of
a major city or even a small country. This likely
means building multiple dedicated nuclear power plans just to fuel
(17:34):
these AI systems.
Speaker 1 (17:35):
Which brings in huge geopolitical issues, environmental questions, infrastructure.
Speaker 2 (17:39):
Challenges exactly, and it highlights a deeper problem, this whole
path leading to this kind of superhuman intelligence. It's just
not understood in our society. We genuinely have no language
in our current politics, economics, or social structures for what
happens when this.
Speaker 1 (17:53):
Arrives, it's completely uncharted territory for our laws, our democracy,
our philosophy. I know Henry Kissinger co authored that book Genesis,
trying to grapple with some of this.
Speaker 2 (18:03):
That's right. Genesis attempts to explore these very questions, how
such intelligence could redefine human experience purpose. But despite these
discussions in some circles, the reality is, as the expert stress,
it's seriously under hyped.
Speaker 1 (18:18):
For the general public, people just don't grasp it.
Speaker 2 (18:21):
People do not understand what happens when intelligence at this level,
effectively free and everywhere, arrives. It's moving faster than society,
faster than democracy, faster than our laws can possibly adapt.
We are just not prepared for the kinds of disruption,
good and bad that ASI could unleash. There's a huge
gap between the labs and the living rooms.
Speaker 1 (18:41):
Which brings us back inevitably to the jobs debate, the
age old question. With every tech revolution historically, automation looms
three hundred years ago, robots last century. Jobs change, yes,
but often more jobs get created than destroyed. Skills shift,
new industries pop up. So the tough question is, you'd
have to convince me this time is different, is it? Why?
Speaker 2 (19:01):
That's the core debate, isn't it? And it keeps people
up at night? The argument for this time is different
hinges on the nature of the intelligence being automated. Previous
revolutions automated muscle power or repetitive cognitive tasks. AI, however,
is automating cognition itself, learning, reasoning, creativity. It's not just
a tool, It's potentially an independent problem solver. Okay, And
(19:24):
what's fascinating A direct input on this is the connection
to demographics, especially in Asia. Their reproduction rates are incredibly
low one point zero or even lower in places like
South Korea Japan.
Speaker 1 (19:35):
Meaning populations are shrinking fast.
Speaker 2 (19:37):
Rapidly disappearing. This demographic crisis is directly accelerating their push
for automation in AI, not just for profit, but is
an existential need to keep society functioning. They literally won't
have enough people for basic services or their economies without
radical automation.
Speaker 1 (19:52):
And if these trends AI replacing cognition demographic ships continue,
what's the potential outcome in say, thirty four.
Speaker 2 (20:00):
Years, Well, one projected scenario is stark. A few humans
work incredibly hard, their productivity massively amplified by AI, while
the rest of us will be dependent on those hardworking humans.
Speaker 1 (20:13):
That's yeah, a truly thought provoking, maybe dystopian picture for some,
or maybe just a completely new paradigm. What does that
mean for value contribution, daily life in a post labor world,
universal basic income, something else entirely exactly.
Speaker 2 (20:28):
It forces us to confront fundamental questions about a future
where human labor as we know it might play a
vastly different role, or maybe even a marginalized one. It
challenges our whole concept of a productive life, maybe even
a fulfilling life.
Speaker 1 (20:41):
Okay, let's shift gears slightly and dive into some of
the actual mechanics, the inner workings, and what's happening in
the industry right now. Give people a glimpse under the hood.
We often think of AI as text in text out
language to language, but it's already way beyond that. It's multimodal.
Speaker 2 (20:56):
Multimodality just means these models can handle different time types
of input and output, not just text. You can give
it a picture and ask tell me what's in this picture,
or describe the mood here or video or video? Yeah,
summarize this clip, identify the objects. Technically, this happens through
API's application programming interfaces.
Speaker 1 (21:15):
Like a standardized connector.
Speaker 2 (21:17):
Kind of think of an API like a menu at
a restaurant your app. The customer doesn't need to know
how the kitchen, Open AI system, Google system makes the dish.
It just needs to know what to order from the menu.
The API calls to get what it wants, like image
classification or TECHT summary. It's a clean interface. Allowing complex
systems to talk to each other easily makes the whole
(21:38):
thing much more versatile.
Speaker 1 (21:40):
So AI isn't just processing text. It's starting to understand
and interact with our messy multisensory world. Now beyond multi modality,
Eric Schmidt pointed out three really interesting tangible things happening
this year that are pushing things forward fast. First, infinite
context windows.
Speaker 2 (21:57):
Infinite context windows. This essentially means the AI can remember
and refer back to basically all his previous interactions and
a huge amount of input text. It gets rid of
the old memory limits.
Speaker 1 (22:07):
Okay, and why is that important?
Speaker 2 (22:09):
It allows the AI to keep feeding its own answers
back into the question, creating this continuous loop that's absolutely
essential for complex step by step planning and long form reasoning.
Think back to the housebuilding agent.
Speaker 1 (22:22):
Right.
Speaker 2 (22:22):
Instead of one command, the AI can plan the whole
multi year thing with perfect memory. Okay, Step one, find
a contractor done, Now what exactly do I need to
discuss with them? Then find an architect what how? Okay?
Found one? Tell them what to do. Get the design
review it, approve it, get redesigns based on feedback. It's
a real series of linked steps the AI manages over
(22:44):
time without losing.
Speaker 1 (22:45):
Track, like a project manager with perfect memory and infinite capacity. Okay.
The second interesting development agents. We touched on this, but
it's worth revisiting its specific meaning here.
Speaker 2 (22:56):
Right in this cutting edge sense agents are autonomous memory sources.
They can add. They're designed to watch something a process, data,
user activity, and environment, and when they see a trigger,
they take an appropriate learned at action based on everything
they've seen and remembered.
Speaker 1 (23:10):
And the challenge now is the.
Speaker 2 (23:12):
Fascinating challenge is that the specs for how these powerful
agents should talk to each other are completely undefined. All
the big players, Google, open AI and Thropic want their
own proprietary agents because agents represent huge.
Speaker 1 (23:27):
Control, control over data, the user exactly.
Speaker 2 (23:30):
And they actively resist making their agents easily interact with
others agents. They want to lock you into their ecosystem.
So while people talk about an agent store like an
app store for AI agents.
Speaker 1 (23:41):
Don't hold your breath this year.
Speaker 2 (23:43):
Probably not this year, too much competition, no standards yet,
it's a foundational battle for the future digital economy playing
out right now.
Speaker 1 (23:50):
So it's a power struggle as much as a technical one.
And the third interesting thing this year text to text
to code computers' writing programs from plain English instructions.
Speaker 2 (24:00):
This is a huge breakthrough, especially if you've ever tried
managing programmers. Yeah, imagine just telling a computer in normal language,
write me a program to do X, Y and Z,
and it actually spits out working code.
Speaker 1 (24:10):
Schmidt gave a kind of funny, wild example.
Speaker 2 (24:13):
Yeah, a program an executive might write if they could
search all literature for energy policy experts with tech backgrounds,
identify them, rank them, score them based on some goal,
automatically send personalized invites to a conference. If they accept,
congratulate them.
Speaker 1 (24:29):
And if they say no, well.
Speaker 2 (24:31):
You can program it to ask why not, and then
use a synthetic voice to call them and tell them
they're idiots for not coming.
Speaker 1 (24:37):
Okay, definitely inappropriate, but it makes the point vividly exactly.
Speaker 2 (24:40):
It highlights the incredible ease with which really complex multi
step tasks, things needing human intelligence, coordination, negotiation right now
could be automated just using natural language. It's not just
simple scripts, it's automating huge chunks of what whole industries do.
Speaker 1 (24:57):
And all this rapid evolution is driven by this intense competition,
right the big player.
Speaker 2 (25:01):
Absolutely in the US, You've got the Big three, Anthropic
allied with Amazon's Cloud, Gemini, which is Google's AI powerhouse,
and open Ai backed by Microsoft. They're all making huge strides,
locked in this high stakes race.
Speaker 1 (25:15):
But it's not just them. Facebook Meta took a different
path open source, a.
Speaker 2 (25:20):
Very distinct path. Yes, they released their huge four hundred
billion parameter model Lama as open source. That's strategically massive.
Speaker 1 (25:29):
Why what does that do?
Speaker 2 (25:30):
An open source model like LAMA changes the game. It
allows wider reduction, lets the whole community innovate on top
of it, and could lead to a much faster, more
decentralized spread of advanced AI globally. It could seriously challenge
the dominance of the closed systems for the Big three.
Speaker 1 (25:45):
So all these giants, the Big three, Facebook, they're all
fighting for.
Speaker 2 (25:50):
What supremacy on all fronts, the best reasoning, the most
accurate answers, the sharpest predictions, the best image classifiers, the
richest multimodal understand.
Speaker 1 (26:00):
Everything, and all this super advanced tech.
Speaker 2 (26:02):
Eventually, it eventually diffuses or distills down. That's the technical
term gets refined in the smaller, more specialized, more efficient
models that can run on more normal computers and.
Speaker 1 (26:12):
Devices, and that's where the action will be soon.
Speaker 2 (26:15):
That distillation process is where the real action will be
in the next one to two years, as these advanced
capabilities become more widely available, integrated into the apps and
tools that will impact pretty much every business and every person.
Speaker 1 (26:28):
So bringing this all together, what we've unpacked today paints
this vivid, maybe startling picture, a technological shift that feels genuinely,
perhaps unnervingly unprecedented. We're standing at the edge of something
that promises to redefine human capability, work itself, maybe even
society's fundamental structure, and all at a speed that's just
(26:51):
mind boggling faster than our institutions can keep up.
Speaker 2 (26:53):
And it's crucial to reiterate this isn't about drumming up fear.
It's about being prepared, having a clear eyed understanding. This
deep dive was really designed to equip you, the listener,
with the insights you need to navigate this accelerating future,
paying attention, actively building your own intuition about what AI
can do as it evolves, Engaging thoughtfully with these really
complex questions. It's going to be.
Speaker 1 (27:15):
Paramount because your life, whether you actively engage or not,
will be affected by AI. Understanding its trajectory is kind
of becoming non optional pretty much. And so we leave
you with this final lingering question, something to moll over
as we all navigate this future. If artificial superintelligence does
become smarter than the sum of humans, if it can
(27:36):
generate its own goals independent of us, what then becomes
the ultimate objective of humanity itself? How do we ensure
our goals, our values, our purpose aligned with an intelligence
that could surpass our collective understanding in ways we can
barely imagine?
Speaker 2 (27:51):
And how do we prepare, really prepare for a world
where the the very definition of work value, maybe even
our place in the cosmos, might be completely reimagined.
Speaker 1 (28:01):
It feels like this fundamental conversation, maybe the most important
one of our time, has only just begun, and it
absolutely demands our full attention.
You may not take interest in politics, but politics will
take interest in you. It's a classic observation, right it is.
But what if we told you that the exact same
principle applies maybe many times over to something far less tangible.
Yet it's increasingly woven into well everything, every aspect of
our lives day by day.
Speaker 2 (00:19):
You're talking about AI, artificial.
Speaker 1 (00:21):
Intelligence exactly AI.
Speaker 2 (00:23):
And it's a bold claim people make, isn't it that
we are living in the most unusual time ever? People
will say that sure about their own era.
Speaker 1 (00:30):
Right, happens all the time.
Speaker 2 (00:31):
But this time, it genuinely feels different. And the reason,
like you said, is profoundly tied to artificial intelligence. We're
already seeing it, aren't we.
Speaker 1 (00:40):
Oh absolutely?
Speaker 2 (00:41):
I mean from what we hear, what we observe, AI
has already fundamentally shifted what it means to be, say
a student. It's impacting work careers in ways that are
still unknown, honestly and largely unpredictable.
Speaker 1 (00:53):
I bet many people listening right now are already asking themselves,
maybe with a bit of anxiety, okay, which skills are
going to remain useful, which ones are going to be obsolete?
How do I even adapt That's the core question for
so many, and.
Speaker 2 (01:06):
That's exactly where we are right now, in this kind
of powerful moment of questioning of transformation. You can help
on to social media, see the latest AI demos, or
just look at how your own work might be shifting,
and you get that feeling, like sort of visceral sense
that something monumental is unfolding. We've all interacted with AI
in some way, probably.
Speaker 1 (01:26):
Sure, spoken to a voice assistant, yeah, had it speak.
Speaker 2 (01:29):
Back, seeing it generate text, maybe even write code. It's
already pretty wild, you know, it is, even with its
current quirks and deficiencies, because it sparks this kind of
undeniable intuition about what's coming next. It's future potential. But
here's the kicker, and this is what we really want
to dig into today. Okay, the real challenge that AI
(01:50):
isn't necessarily what we're seeing day to day right now.
It's something yeah, well, the experts say it's unprecedented, extreme
and vastly different from its current form, hints at of
future that's kind of hard to fully grasp precisely, And
that's really the mission for this deep dive today. We're
not here to just skim the surface or throw a
(02:10):
bunch of technical jargon at you. Instead, we want to
take you on a journey through the insights and frankly,
even the stark warnings from some of the leading minds
right at the forefront of AI.
Speaker 1 (02:21):
Research, people who are building this stuff exactly.
Speaker 2 (02:24):
The goal isn't to overwhelm you. It's to equip you
with a more nuanced understanding of these really profound future
possibilities and crucially, the critical questions they raise for all
of us. Think of it like your essential shortcut to
getting informed on a topic that isn't just going to.
Speaker 1 (02:42):
Shape your world, it already is.
Speaker 2 (02:44):
It's actively shaping it right now.
Speaker 1 (02:45):
Okay, so let's unpack this foundational idea, this sort of
core logic that underpins so much of the thinking in
the AI space. It starts with a deceptively simple argument,
but it's incredibly powerful. The human brain is at its
core a biologue computer, right It processes information, learns, adapts.
Speaker 2 (03:04):
Creates, It has amazing things.
Speaker 1 (03:05):
Amazing things We learn languages, solve super complex problems, create art,
develop theories. So if a biological computer, our brain can
achieve all this, every cognitive function, every creative leap every
bit of logic, then why logically couldn't a really advance
digital computer, a digital brain do the same, and not
just some of it exactly, not just some but anything
(03:28):
we can learn to do.
Speaker 2 (03:29):
It really is the one sentence summary, isn't it Huh?
If we with our organic wetwear can learn and do
this incredible range of tasks, right, then a digital system
free from biologies limits should theoretically be able to do
the same and probably eventually surpass us.
Speaker 1 (03:46):
And that premise, that fundamental idea, immediately brings us to
this really dramatic, almost unsettling question, one that many people
find I think incredibly difficult to fully internalize or even
just emotionally believe. Which is what happens when computers can
do all of our jobs, not just automate a few
tasks here and there, but truly take on every cognitive
function we perform in our professional lives.
Speaker 2 (04:07):
That's the big one. Isn't it feels almost yeah, dystopian
like you said to even say it out loud.
Speaker 1 (04:12):
I remember hearing someone compare it to asking a horse
what it'll do when cars take over the horse just
they can't even conceive of it.
Speaker 2 (04:20):
Right, But the logic, according to these experts holds. And
then what happens after that? What do we as a
society want to use these incredibly capable AIS for good question? Well,
the answer, at least initially seems to be more work,
more economic growth, vastly accelerated r ND, and maybe most importantly,
(04:42):
doing AI research itself.
Speaker 1 (04:43):
Okay, now that's where it gets really interesting.
Speaker 2 (04:45):
Exactly because if AI can significantly contribute to its own
research and development.
Speaker 1 (04:50):
Then the rate of progress just takes off. It becomes,
as they say, really extremely fast for some period. It's
this kind of unprecedented feedback loop.
Speaker 2 (04:57):
And accelerating intelligence where the intelligence it self drives its
own exponential improvement. It learns how to learn faster.
Speaker 1 (05:04):
These are such extreme scenarios, though almost unimaginable for many people,
and it highlights what you called this significant intuition gap.
Speaker 2 (05:13):
Even people deep in the field like Ilia Sutzkiver, you know,
one of the key figures behind open AI. He admits,
they struggle to fully internalize this future on an emotional level,
even as the cold hard logic dictates it's very likely
to happen.
Speaker 1 (05:28):
So the head says yes, but the gut says, whoa.
Speaker 2 (05:32):
Kind of He argues that no amount of essays or
papers or explanations, no matter how detailed, none of it
can truly compete with what we actually see with our
own senses.
Speaker 1 (05:42):
Like trying to explain the Internet to someone in the
nineteen fifties.
Speaker 2 (05:45):
Exactly, you can describe the function, but the reality, the
paradigm shift, it's almost impossible to grasp without living it.
We humans are good at linde, your thinking, terrible at exponential.
Speaker 1 (05:55):
That's a really powerful point, and it makes sense our
current interactions with AI, even the perfect ones asking Syria
question seeing a weird AI generated image, they're already building
that intuition bit by bit. As AI keeps improving day
by day, month by month.
Speaker 2 (06:10):
That intuition gets stronger. Those once imaginary scenarios, the stuff
of old sci fi.
Speaker 1 (06:14):
Yeah, they become much more real, less theoretical, more visceral.
It's a progression from abstract idea to tangible reality, and
it shapes how we think, maybe without us even realizing
how much our mental models are shifting.
Speaker 2 (06:30):
Which brings us right back full circle to that opening analogy.
Speaker 1 (06:33):
Politics takes an interest in you, right.
Speaker 2 (06:36):
Just like politics AI will affect your life to a great,
great extent, whether you like it or not, whether you're
paying attention or not.
Speaker 1 (06:43):
So the main thing, then is what.
Speaker 2 (06:45):
The main thing is to pay attention to actively look
at what it can do, try to understand it and
generate the collective energy we're going to need to solve
the profound challenges that will come up.
Speaker 1 (06:55):
Because this isn't just a challenge. Leading thinkers people like
Eric Schmidt, former Google CEO, they're positioning it as potentially
the greatest challenge of humanity ever.
Speaker 2 (07:04):
But also, and this is crucial, they argue that successfully
navigating it will bring the greatest reward. It's this immense
double edged sword. It demands our full attention, our collective ingenuity,
and frankly, a willingness to confront questions we've never had
to ask before.
Speaker 1 (07:19):
Okay, so if that's the grand, maybe slightly terrifying, philosophical future,
let's zoom in a bit. What does this mean in
the near future, like the next one to two years.
Speaker 2 (07:30):
Right, the immediate landscape?
Speaker 1 (07:31):
Yeah, because this timeline might genuinely surprise people, especially if
you're in certain fields or your job involves a lot
of what we think of as knowledge work.
Speaker 2 (07:41):
The industry consensus here, and this is coming from figures
like Schmidt and many AI researchers, is pretty stark. Take
programmers for instance, Okay, within just one year, the vast
majority of generalist programmers are expected to be replaced by
AI programmers one year.
Speaker 1 (07:57):
Wow, why so fast?
Speaker 2 (07:59):
Well, there are clear reasons. Programming fundamentally uses a simpler,
more structured, more logical language than human language. Sure, these
AI algorithms excel at something conceptually similar to word prediction,
but on a massive scale. They optimize what's called the
loss function, which.
Speaker 1 (08:14):
Is like their internal score for how wrong they are exactly.
Speaker 2 (08:17):
It's their greade book. They want to minimize that loss,
get it close to zero. They refine their code, test it,
refine it again, millions of times faster than any human
could write and debug. They just keep iterating until the
code passes all the tests, all the requirements.
Speaker 1 (08:32):
That's fascinating. So the structure of code itself makes it
easier for AI to master. And it's not just coding, right,
you mentioned Schmidt points to mathematicians too.
Speaker 2 (08:41):
Indeed, similarly, within one year, AI is projected to produce
graduate level mathematicians at the absolute tippy top of programs,
the very best. The reasoning is similar. Math also uses
a highly structured, simpler language, albeit abstract. AI can use
conjecture proof formats can work through protocols like Lean, which
(09:01):
is a formal proof assistant.
Speaker 1 (09:03):
So it's not just calculation, it's actual proof verification.
Speaker 2 (09:06):
Yes, Lean helps verify proofs rigorously. So AI isn't just
proposing solutions, it's constructing and confirming mathematically sound, verifiable arguments,
generating genuinely novel mathematical insights.
Speaker 1 (09:18):
Potentially, so for programmers then this means the specific language
they use, Python, Java, whatever, might soon not matter nearly
as much because the goal is the outcome, and the
AI generates the code.
Speaker 2 (09:32):
That's the implication. It's a huge philosophical shift in the craft.
It moves from the mechanics of writing syntax to the
higher level art of defining the outcome, crafting the perfect
prompt or instructions.
Speaker 1 (09:44):
For the AI, from syntax to semantics.
Speaker 2 (09:46):
Absolutely, from the how to the what and why. And
this leads right into another critical development happening now, the
dawn of recursive self improvement.
Speaker 1 (09:55):
Okay, that sounds like sci fi, it does.
Speaker 2 (09:57):
But it's not anymore. This is where AI generates code
for its own research programs. It's writing code to make
itself smarter.
Speaker 1 (10:03):
And there's evidence this is actually happening.
Speaker 2 (10:05):
Yes, it's not theoretical. Top research groups like open AI
and Tropic they're already reporting to ten percent, maybe twenty
percent of their new experimental research code is being generated
by the computer itself.
Speaker 1 (10:16):
That's genuinely mind boggling. Think about that. If it's ten
twenty percent now, when it's still relatively early, what happens
when that scales fifty percent, eighty percent, one hundred percent.
Speaker 2 (10:26):
It's not just optimizing tasks, it's accelerating its own evolution,
its own intelligence, at a pace humans just can't match
a self amplifying engine of discovery.
Speaker 1 (10:37):
And this isn't the only immediate shift. We also need
to talk about agentic solutions. Agents.
Speaker 2 (10:42):
Yes, agents a term that gets thrown around a lot,
sometimes loosely, but in this context it means truly autonomous systems.
They have input output persistent memory, and they learn, they
observe something, understand it based on past data, remember it,
and then take intelligent, multi step actions towards a goal.
It's about automating entire processes, not just single steps.
Speaker 1 (11:04):
Okay, let's make that concrete. The expert use a great example.
Buying a house, super complex process, right.
Speaker 2 (11:10):
Imagine an agent system handling that start to finish. It
begins by finding a house in a specific area you want,
say McLean Virginia, based on your preferences. Okay. Then it
automatically pulls up and analyzes all the local rules zoning laws,
instantly figuring out how big a house you can actually
build there.
Speaker 1 (11:26):
Wow.
Speaker 2 (11:26):
After that, it handles the whole transaction to buy the land,
negotiation paperwork the lot.
Speaker 1 (11:31):
And it doesn't stop there at all.
Speaker 2 (11:32):
Once the land secured, the agent could design the house itself,
maybe collaborates with a human architect for the final touches
the esthetic choices.
Speaker 1 (11:41):
But it does the bulk of the design.
Speaker 2 (11:42):
Work largely independently. Yeah, then you prove the design. After that,
it autonomously finds and hires the best contractor for the job,
manages all the payments for materials, labor, tracks the.
Speaker 1 (11:54):
Invoices, and the slightly funny part, and yes.
Speaker 2 (11:58):
Quite tellingly, it even has the capability to initiate legal
action to sue the contractor for lack of performance. If
things go south and they don't hold up their end
of the bargain.
Speaker 1 (12:09):
I love that detail suing the contractor. It just perfectly
shows how comprehensive these systems are meant to be.
Speaker 2 (12:16):
And to end, it's not just one piece. It's stitching
together a whole chain of complex, interdependent decisions and actions.
Speaker 1 (12:23):
But here's the key insight from that example, something Eric
Schmidt really emphasized. This isn't just about buying a house
or replacing programmers. That single analogy, believe it or not, describes,
in his words, every business process, every government process, and
every academic process in our nation.
Speaker 2 (12:39):
Think about that, finding information, executing complex tasks, financial transactions,
even litigation. It suggests a fundamental shift in how all
tasks get done, not just tech but everywhere.
Speaker 1 (12:53):
It truly is a foundational change. And what's striking is
the timeline locking in. They say within the next year
or two, as.
Speaker 2 (13:00):
The expert stress, we're not going to stop it. This
isn't just about jobs and specific sectors. It's a complete
reimagining of how work broadly defined, gets done. It goes
beyond current automation, which also just streamlines human processes. This
is creating new often self optimizing processes.
Speaker 1 (13:17):
The implications are yeah, profound, far reaching. It could alter
the very base of how our economy and society function.
So let's look further out now, midterm horizon Artificial general
intelligence AGI projected for the next what three to five
years roughly?
Speaker 2 (13:32):
Yeah, No definitions vary.
Speaker 1 (13:33):
Okay, So what's the core idea we should grasp? The
most important characteristic?
Speaker 2 (13:37):
The basic idea discussed for maybe fifteen twenty years in
AI circles is the point where an intelligence system has
the flexibility of a human flexibility. What really distinguishes AGI
from current narrow AI, which is powerful, sure, but always
directed by a human toward the specific goal is that
AGI implies that computer can generate its own objective of function,
(14:00):
its own goals, its own strategies.
Speaker 1 (14:03):
So it's not just executing tasks, it's defining the tasks,
deciding what's important on its.
Speaker 2 (14:08):
Own, exactly, self directed, autonomous in its intellectual pursuits. That's
the game changer.
Speaker 1 (14:13):
And there's this thing called the San Francisco consensus around
this sounds almost like a local.
Speaker 2 (14:17):
Cult, huh. Indeed, it's a specific belief held strongly by
many researchers. There deep in the AI hubs.
Speaker 1 (14:24):
Maybe it is the water.
Speaker 2 (14:25):
Maybe, But seriously, they've largely convinced themselves based on observing
these exponential scaling laws and the rapid progress that AGI
will arrive within two to three cranks cranks. A crank
in their lingo is about eighteen months, So their timeline
for AGI is roughly three to four and a half
years from now.
Speaker 1 (14:41):
That's incredibly soon. And their definition of AGI potent intelligence
greater than the sum of human intelligence.
Speaker 2 (14:49):
Not just as smart as one human, but smarter than
all human brains combined working together.
Speaker 1 (14:55):
Wow, greater than the sum of human intelligence. That's hard
to even wrap your head around. So I hear not
everyone agrees on that specific timeline. Even among the experts.
Speaker 2 (15:04):
You're absolutely right, there is debate. While most agree AGI
is highly likely eventually, the exact timeline is uncertain. Some
think it might take longer, maybe unforeseen scaling issues, the
difficulty of real world interaction, or just achieving genuine self
direction beyond complex pattern matching.
Speaker 1 (15:22):
So the consensus is more like, we don't know exactly when.
Speaker 2 (15:26):
Pretty much, it highlights how unpredictable these huge leaps are.
But regardless of whether it's three five, maybe eight years.
The implication of AGI once it arrives is staggering.
Speaker 1 (15:36):
And that implication is vividly called the smartest human in
your pocket. Explain that because it sounds revolutionary on a
very personal.
Speaker 2 (15:43):
Level, it means quite literally, every single person, regardless of
background or wealth, could have access to the equivalent of
the smartest human being whoever lived or could live, available
instantly for every problem, every problem. Imagine needing an architecture
design boom. You have the best architect imaginable right there,
facing a complex math problem. The greatest mathematician in history
(16:07):
is at your fingertips. Need help with writing science, the
most brilliant writer or scientist on call, on demand. It
completely democratizes expert level intelligence, makes it universally accessible. It
could flatten knowledge hierarchies and ways we can barely conceive,
transforms problem solving from a rare, specialized resource to something ubiquitous.
Speaker 1 (16:28):
Okay, So, if AGI is midterm, what's beyond that? The
long view? Artificial superintelligence ASI projected maybe six years out plus.
Speaker 2 (16:36):
By the more aggressive timelines. Yes, ASI refers to computers
that aren't just smarter than one human or even the
sum of humanity but smarter than the sum of humans combined,
a qualitative leap beyond AGI.
Speaker 1 (16:47):
So it could perform any intellectual task a human can,
but just vastly better, faster, more memory, deeper cognition.
Speaker 2 (16:55):
That's the idea, and the San Francisco Consensus, again based
purely on scaling projects, suggests it could arrive within six years.
Speaker 1 (17:02):
Six years for something that could completely alter human civilization.
But there's a huge real world constraint here, isn't there
The energy required.
Speaker 2 (17:13):
Absolutely, This isn't just abstract code. It needs power, massive
amounts of it. A critical practical requirement for ASI with
current tech is an enormous amount of electricity. How much
are we talking significant gigawatts, like the power consumption of
a major city or even a small country. This likely
means building multiple dedicated nuclear power plans just to fuel
(17:34):
these AI systems.
Speaker 1 (17:35):
Which brings in huge geopolitical issues, environmental questions, infrastructure.
Speaker 2 (17:39):
Challenges exactly, and it highlights a deeper problem, this whole
path leading to this kind of superhuman intelligence. It's just
not understood in our society. We genuinely have no language
in our current politics, economics, or social structures for what
happens when this.
Speaker 1 (17:53):
Arrives, it's completely uncharted territory for our laws, our democracy,
our philosophy. I know Henry Kissinger co authored that book Genesis,
trying to grapple with some of this.
Speaker 2 (18:03):
That's right. Genesis attempts to explore these very questions, how
such intelligence could redefine human experience purpose. But despite these
discussions in some circles, the reality is, as the expert stress,
it's seriously under hyped.
Speaker 1 (18:18):
For the general public, people just don't grasp it.
Speaker 2 (18:21):
People do not understand what happens when intelligence at this level,
effectively free and everywhere, arrives. It's moving faster than society,
faster than democracy, faster than our laws can possibly adapt.
We are just not prepared for the kinds of disruption,
good and bad that ASI could unleash. There's a huge
gap between the labs and the living rooms.
Speaker 1 (18:41):
Which brings us back inevitably to the jobs debate, the
age old question. With every tech revolution historically, automation looms
three hundred years ago, robots last century. Jobs change, yes,
but often more jobs get created than destroyed. Skills shift,
new industries pop up. So the tough question is, you'd
have to convince me this time is different, is it? Why?
Speaker 2 (19:01):
That's the core debate, isn't it? And it keeps people
up at night? The argument for this time is different
hinges on the nature of the intelligence being automated. Previous
revolutions automated muscle power or repetitive cognitive tasks. AI, however,
is automating cognition itself, learning, reasoning, creativity. It's not just
a tool, It's potentially an independent problem solver. Okay, And
(19:24):
what's fascinating A direct input on this is the connection
to demographics, especially in Asia. Their reproduction rates are incredibly
low one point zero or even lower in places like
South Korea Japan.
Speaker 1 (19:35):
Meaning populations are shrinking fast.
Speaker 2 (19:37):
Rapidly disappearing. This demographic crisis is directly accelerating their push
for automation in AI, not just for profit, but is
an existential need to keep society functioning. They literally won't
have enough people for basic services or their economies without
radical automation.
Speaker 1 (19:52):
And if these trends AI replacing cognition demographic ships continue,
what's the potential outcome in say, thirty four.
Speaker 2 (20:00):
Years, Well, one projected scenario is stark. A few humans
work incredibly hard, their productivity massively amplified by AI, while
the rest of us will be dependent on those hardworking humans.
Speaker 1 (20:13):
That's yeah, a truly thought provoking, maybe dystopian picture for some,
or maybe just a completely new paradigm. What does that
mean for value contribution, daily life in a post labor world,
universal basic income, something else entirely exactly.
Speaker 2 (20:28):
It forces us to confront fundamental questions about a future
where human labor as we know it might play a
vastly different role, or maybe even a marginalized one. It
challenges our whole concept of a productive life, maybe even
a fulfilling life.
Speaker 1 (20:41):
Okay, let's shift gears slightly and dive into some of
the actual mechanics, the inner workings, and what's happening in
the industry right now. Give people a glimpse under the hood.
We often think of AI as text in text out
language to language, but it's already way beyond that. It's multimodal.
Speaker 2 (20:56):
Multimodality just means these models can handle different time types
of input and output, not just text. You can give
it a picture and ask tell me what's in this picture,
or describe the mood here or video or video? Yeah,
summarize this clip, identify the objects. Technically, this happens through
API's application programming interfaces.
Speaker 1 (21:15):
Like a standardized connector.
Speaker 2 (21:17):
Kind of think of an API like a menu at
a restaurant your app. The customer doesn't need to know
how the kitchen, Open AI system, Google system makes the dish.
It just needs to know what to order from the menu.
The API calls to get what it wants, like image
classification or TECHT summary. It's a clean interface. Allowing complex
systems to talk to each other easily makes the whole
(21:38):
thing much more versatile.
Speaker 1 (21:40):
So AI isn't just processing text. It's starting to understand
and interact with our messy multisensory world. Now beyond multi modality,
Eric Schmidt pointed out three really interesting tangible things happening
this year that are pushing things forward fast. First, infinite
context windows.
Speaker 2 (21:57):
Infinite context windows. This essentially means the AI can remember
and refer back to basically all his previous interactions and
a huge amount of input text. It gets rid of
the old memory limits.
Speaker 1 (22:07):
Okay, and why is that important?
Speaker 2 (22:09):
It allows the AI to keep feeding its own answers
back into the question, creating this continuous loop that's absolutely
essential for complex step by step planning and long form reasoning.
Think back to the housebuilding agent.
Speaker 1 (22:22):
Right.
Speaker 2 (22:22):
Instead of one command, the AI can plan the whole
multi year thing with perfect memory. Okay, Step one, find
a contractor done, Now what exactly do I need to
discuss with them? Then find an architect what how? Okay?
Found one? Tell them what to do. Get the design
review it, approve it, get redesigns based on feedback. It's
a real series of linked steps the AI manages over
(22:44):
time without losing.
Speaker 1 (22:45):
Track, like a project manager with perfect memory and infinite capacity. Okay.
The second interesting development agents. We touched on this, but
it's worth revisiting its specific meaning here.
Speaker 2 (22:56):
Right in this cutting edge sense agents are autonomous memory sources.
They can add. They're designed to watch something a process, data,
user activity, and environment, and when they see a trigger,
they take an appropriate learned at action based on everything
they've seen and remembered.
Speaker 1 (23:10):
And the challenge now is the.
Speaker 2 (23:12):
Fascinating challenge is that the specs for how these powerful
agents should talk to each other are completely undefined. All
the big players, Google, open AI and Thropic want their
own proprietary agents because agents represent huge.
Speaker 1 (23:27):
Control, control over data, the user exactly.
Speaker 2 (23:30):
And they actively resist making their agents easily interact with
others agents. They want to lock you into their ecosystem.
So while people talk about an agent store like an
app store for AI agents.
Speaker 1 (23:41):
Don't hold your breath this year.
Speaker 2 (23:43):
Probably not this year, too much competition, no standards yet,
it's a foundational battle for the future digital economy playing
out right now.
Speaker 1 (23:50):
So it's a power struggle as much as a technical one.
And the third interesting thing this year text to text
to code computers' writing programs from plain English instructions.
Speaker 2 (24:00):
This is a huge breakthrough, especially if you've ever tried
managing programmers. Yeah, imagine just telling a computer in normal language,
write me a program to do X, Y and Z,
and it actually spits out working code.
Speaker 1 (24:10):
Schmidt gave a kind of funny, wild example.
Speaker 2 (24:13):
Yeah, a program an executive might write if they could
search all literature for energy policy experts with tech backgrounds,
identify them, rank them, score them based on some goal,
automatically send personalized invites to a conference. If they accept,
congratulate them.
Speaker 1 (24:29):
And if they say no, well.
Speaker 2 (24:31):
You can program it to ask why not, and then
use a synthetic voice to call them and tell them
they're idiots for not coming.
Speaker 1 (24:37):
Okay, definitely inappropriate, but it makes the point vividly exactly.
Speaker 2 (24:40):
It highlights the incredible ease with which really complex multi
step tasks, things needing human intelligence, coordination, negotiation right now
could be automated just using natural language. It's not just
simple scripts, it's automating huge chunks of what whole industries do.
Speaker 1 (24:57):
And all this rapid evolution is driven by this intense competition,
right the big player.
Speaker 2 (25:01):
Absolutely in the US, You've got the Big three, Anthropic
allied with Amazon's Cloud, Gemini, which is Google's AI powerhouse,
and open Ai backed by Microsoft. They're all making huge strides,
locked in this high stakes race.
Speaker 1 (25:15):
But it's not just them. Facebook Meta took a different
path open source, a.
Speaker 2 (25:20):
Very distinct path. Yes, they released their huge four hundred
billion parameter model Lama as open source. That's strategically massive.
Speaker 1 (25:29):
Why what does that do?
Speaker 2 (25:30):
An open source model like LAMA changes the game. It
allows wider reduction, lets the whole community innovate on top
of it, and could lead to a much faster, more
decentralized spread of advanced AI globally. It could seriously challenge
the dominance of the closed systems for the Big three.
Speaker 1 (25:45):
So all these giants, the Big three, Facebook, they're all
fighting for.
Speaker 2 (25:50):
What supremacy on all fronts, the best reasoning, the most
accurate answers, the sharpest predictions, the best image classifiers, the
richest multimodal understand.
Speaker 1 (26:00):
Everything, and all this super advanced tech.
Speaker 2 (26:02):
Eventually, it eventually diffuses or distills down. That's the technical
term gets refined in the smaller, more specialized, more efficient
models that can run on more normal computers and.
Speaker 1 (26:12):
Devices, and that's where the action will be soon.
Speaker 2 (26:15):
That distillation process is where the real action will be
in the next one to two years, as these advanced
capabilities become more widely available, integrated into the apps and
tools that will impact pretty much every business and every person.
Speaker 1 (26:28):
So bringing this all together, what we've unpacked today paints
this vivid, maybe startling picture, a technological shift that feels genuinely,
perhaps unnervingly unprecedented. We're standing at the edge of something
that promises to redefine human capability, work itself, maybe even
society's fundamental structure, and all at a speed that's just
(26:51):
mind boggling faster than our institutions can keep up.
Speaker 2 (26:53):
And it's crucial to reiterate this isn't about drumming up fear.
It's about being prepared, having a clear eyed understanding. This
deep dive was really designed to equip you, the listener,
with the insights you need to navigate this accelerating future,
paying attention, actively building your own intuition about what AI
can do as it evolves, Engaging thoughtfully with these really
complex questions. It's going to be.
Speaker 1 (27:15):
Paramount because your life, whether you actively engage or not,
will be affected by AI. Understanding its trajectory is kind
of becoming non optional pretty much. And so we leave
you with this final lingering question, something to moll over
as we all navigate this future. If artificial superintelligence does
become smarter than the sum of humans, if it can
(27:36):
generate its own goals independent of us, what then becomes
the ultimate objective of humanity itself? How do we ensure
our goals, our values, our purpose aligned with an intelligence
that could surpass our collective understanding in ways we can
barely imagine?
Speaker 2 (27:51):
And how do we prepare, really prepare for a world
where the the very definition of work value, maybe even
our place in the cosmos, might be completely reimagined.
Speaker 1 (28:01):
It feels like this fundamental conversation, maybe the most important
one of our time, has only just begun, and it
absolutely demands our full attention.
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.
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.