In this week’s AI Breakdown, I cover the biggest stories shaping the world of AI. OpenAI doubles down on the enterprise with new partnerships and ambitions to make ChatGPT the front door to work. Meta faces backlash over plans to use AI chats for ad targeting. OpenAI teams up with AMD in a massive GPU deal as US–China tensions reshape the global chip race. Anthropic lands major wins with IBM and Deloitte, and new data shows how AI is transforming jobs, from “workslop” cleanup work to declining entry-level roles. All that and more, in under 10 minutes.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Hi everyone.
(00:00):
I'm Andy and this is the AI Breakdown.
Welcome to your weekly news edition where I'll cover what happened in AI last week, why it matters all in less than 10 minutes. 4 00:00:13,848.89137483 --> 00:00:20,658.89137483 First up, as I covered in my midweek episode last week, open AI used its dev conference to make a very loud statement. 5 00:00:21,48.89137483 --> 00:00:22,968.89137483 It's going all in on the enterprise. 6 00:00:23,388.89137483 --> 00:00:32,88.89137483 Reuters reports new partnerships with Spotify, Zillow, Mattel, and others, plus developer tools that let chat GPT plug directly into business apps. 7 00:00:32,418.89137483 --> 00:00:37,938.89137483 Think generating playlists inside Spotify, or filtering property searches on Zillow via chat. 8 00:00:38,688.89137483 --> 00:00:47,658.89137483 Nick Turley, who leads chat, GPT, said the product will evolve into something that feels a little bit more like an operating system, acting as a central hub for services. 9 00:00:48,663.89137483 --> 00:00:54,453.89137483 Sam Altman reinforced the shift with a huge focus on serving businesses and brushed off bubble chatter. 10 00:00:54,903.89137483 --> 00:01:00,123.89137483 Admitting whilst parts of AI are kind of bubbly, he insisted real value will get created. 11 00:01:00,783.89137483 --> 00:01:06,393.89137483 Stocks of several partners briefly popped on the news, which is always a nice sentiment check even if it fades. 12 00:01:07,83.89137483 --> 00:01:10,173.89137483 The bigger question everyone's asking is about sustainability. 13 00:01:10,593.89137483 --> 00:01:16,563.89137483 Open AI's ambitions from locking in massive compute to launching splashy apps requires serious spending. 14 00:01:16,998.89137483 --> 00:01:19,848.89137483 Sooner or later, the enterprise side has to pay the bills. 15 00:01:20,478.89137483 --> 00:01:22,68.89137483 It certainly looks like OpenAI. 16 00:01:22,98.89137483 --> 00:01:35,478.89137483 Want to make chat GPT the front door to work? One interface to search, summarize, draft query internal systems, and invoke third party tools they want to be your go-to for AI agents and assistance to. 17 00:01:35,748.89137483 --> 00:01:44,538.89137483 But the burden is on them to prove ROI at scale ensure that plugins, security and governance are robust enough for mission critical use. 18 00:01:45,453.89137483 --> 00:01:49,233.89137483 If they nail it, they stop being just a chat app and become a platform. 19 00:01:49,323.89137483 --> 00:01:59,412.36055526 Other products build on Meta is changing how it uses your AI chats and the internet, unsurprisingly, is not thrilled from mid-December. 20 00:01:59,442.36055526 --> 00:02:07,302.36055526 Conversations with Meta's AI across Facebook, Instagram, and threads will feed into add and feed personalization. 21 00:02:07,752.36055526 --> 00:02:12,762.36055526 There's no opt out if you use meta AI regions with stricter laws, including the eu. 22 00:02:13,92.36055526 --> 00:02:15,912.36055526 UK and South Korea are excluded for now. 23 00:02:16,782.36055526 --> 00:02:23,262.36055526 Met as Kristi Harris said that people's interactions are another piece of the input informing ads and recommendations. 24 00:02:23,742.36055526 --> 00:02:31,992.36055526 The company says sensitive topics like health or religion won't be used for ads, and it's keen to monetize more than a billion monthly AI users. 25 00:02:32,472.36055526 --> 00:02:37,152.36055526 Marketers love the intense signal, but privacy advocates call it invasive and creepy. 26 00:02:37,662.36055526 --> 00:02:40,572.36055526 Me personally, I'm not a fan of forced consent like this. 27 00:02:40,872.36055526 --> 00:02:46,602.36055526 It's just not the way things should happen today, and it risks leaving a sour taste with Meta's user base. 28 00:02:48,465.17234829 --> 00:02:56,565.17234829 OpenAI just gave a MD its biggest vote of confidence, yet a multi-year deal for hundreds of thousands of AI chips with deployment of instinct. 29 00:02:56,565.17234829 --> 00:03:03,345.17234829 GPUs at staggering scale starting in late 2026 on the order of six gigawatts of capacity. 30 00:03:03,870.17234829 --> 00:03:06,870.17234829 Investors went euphoric and EMD stock spiked. 31 00:03:07,260.17234829 --> 00:03:15,330.17234829 The agreement even includes warrants that could let OpenAI acquire up to 10% of EMD at a nominal 1 cent per share if milestones are met. 32 00:03:16,260.17234829 --> 00:03:20,520.17234829 MD's Forest Noro called the pact Transformative and analysts agree. 33 00:03:20,520.17234829 --> 00:03:25,230.17234829 It puts EMD firmly in the AI big leagues, even as Nvidia remains dominant. 34 00:03:25,590.17234829 --> 00:03:26,430.17234829 Why it matters. 35 00:03:26,760.17234829 --> 00:03:36,810.17234829 Open AI is securing compute at unprecedented levels to keep models responsive and expand services, while also diversifying beyond a single chip supplier for enterprises. 36 00:03:36,810.17234829 --> 00:03:43,170.17234829 Depending on AI platforms, more competition in silicon could mean better availability and pricing. 37 00:03:43,410.17234829 --> 00:03:45,450.17234829 As a MD capacity comes online. 38 00:03:45,870.17234829 --> 00:04:03,153.04171893 Skeptics note, OpenAI still relies on Nvidia and question whether anyone can afford this Arms race indefinitely, but the direction of travel's, clear compute is strategy, and the software winners will be the ones who lock down their supply as tensions increase between the US and China. 39 00:04:03,528.04171893 --> 00:04:15,948.04171893 China's latest retaliatory tariffs and tighter export controls, including new restrictions on rare earth minerals, are now reshaping the global AI hardware landscape and both a MD and Nvidia. 40 00:04:15,978.04171893 --> 00:04:29,148.04171893 According in the crossfire, in August, these chip makers agreed to an unprecedented deal requiring them to pay 15% of their revenue to the US government from AI chip sales to China as a condition for export licenses. 41 00:04:29,688.04171893 --> 00:04:38,628.04171893 While this arrangement lets a MD and Nvidia maintain a presence in China, it slashes profitability and adds new risks to supply chains and pricing strategies. 42 00:04:39,438.04171893 --> 00:04:46,218.04171893 EMD in particular has already absorbed substantial financial hits from inventory writedowns with up to $1.5 43 00:04:46,218.04171893 --> 00:04:49,218.04171893 billion in projected revenue losses this year alone. 44 00:04:49,908.04171893 --> 00:04:58,488.04171893 As the US Senate moves to prioritize domestic buyers and further restrict exports, AI companies find themselves navigating complex tariffs. 45 00:04:59,28.04171893 --> 00:05:01,848.04171893 Regulatory hurdles and shifting alliances. 46 00:05:02,178.04171893 --> 00:05:11,778.04171893 Making it clear that geopolitical tensions are now a core part of AI's strategic calculus and may shape which technologies reach global scale in the years ahead. 47 00:05:13,85.88230076 --> 00:05:15,515.88230076 Andro landed two heavyweight allies. 48 00:05:15,935.88230076 --> 00:05:25,115.88230076 IBM is integrated clawed into its enterprise software, starting with a developer environment that helps corporate engineers generate code and automate tasks. 49 00:05:25,400.88230076 --> 00:05:41,30.88230076 And the two companies published a Guide for Building Safe Enterprise Grade AI agents, IBM's software Chief Dinesh Neral Stressed Governance and Reliability, AI that fits how enterprises work, not experimental tools that create new risks. 50 00:05:41,510.88230076 --> 00:05:45,590.88230076 Meanwhile, Deloitte is rolling out Claude to 500,000 employees. 51 00:05:45,875.88230076 --> 00:05:46,655.88230076 Worldwide. 52 00:05:46,925.88230076 --> 00:05:54,185.88230076 Andros largest deployment to date, putting an AI assistant in the hands of auditors, consultants, and back office teams. 53 00:05:54,485.88230076 --> 00:05:58,475.88230076 Claude's reputation for safety and fewer hallucinations is a selling point. 54 00:05:59,75.88230076 --> 00:06:01,655.88230076 This is the shape of enterprise AI adoption. 55 00:06:02,45.88230076 --> 00:06:10,25.88230076 Embed assistance inside trusted tools per scale with training and oversight and measure time saved on research, drafting and coding. 56 00:06:10,715.88230076 --> 00:06:11,765.88230076 The upside is huge. 57 00:06:11,765.88230076 --> 00:06:12,545.88230076 Productivity. 58 00:06:12,890.88230076 --> 00:06:15,500.88230076 The risk is quiet errors slipping into deliverables. 59 00:06:15,800.88230076 --> 00:06:23,930.88230076 If teams treat AI like an Oracle, expect more firms to follow IBM's lead on governance checklists while pushing broad access. 60 00:06:23,930.88230076 --> 00:06:27,740.88230076 Like Deloitte Safety Plus scale is becoming the winning combination. 61 00:06:29,104.95129099 --> 00:06:30,304.95129099 Upworks September. 62 00:06:30,304.95129099 --> 00:06:32,974.95129099 Hiring data reveals something counterintuitive. 63 00:06:33,394.95129099 --> 00:06:34,954.95129099 Companies aren't replacing humans with ai. 64 00:06:36,199.95129099 --> 00:06:38,689.95129099 They're frantically hiring people to clean up after it. 65 00:06:39,169.95129099 --> 00:06:48,949.95129099 The report shows massive surges in translation work up 29% sales copywriting up 12% and quality assurance up 9%. 66 00:06:49,429.95129099 --> 00:06:53,184.95129099 Why they say AI is churning out what's now called work slop. 67 00:06:53,899.95129099 --> 00:06:59,59.95129099 Subpar output that needs human experts to validate, refine, and fix while work. 68 00:06:59,59.95129099 --> 00:07:08,299.95129099 Swap is a real thing, I believe it often stems from poor AI implementation skills, gaps, and lack of proper human oversight rather than technology limitations. 69 00:07:08,989.95129099 --> 00:07:15,589.95129099 According to BetterUp research, 40% of employees have already encountered AI hallucinations or mistakes. 70 00:07:15,919.95129099 --> 00:07:18,949.95129099 With each incident, taken an average of two hours to resolve. 71 00:07:19,579.95129099 --> 00:07:33,109.95129099 Meanwhile, project management hiring among small businesses exploded by 102%, suggesting companies are investing heavily in the human infrastructure needed to coordinate AI implementations properly. 72 00:07:33,889.95129099 --> 00:07:41,569.95129099 My view is this data is genuinely fascinating, but we need to be careful about drawing sweeping conclusions from a single platform. 73 00:07:42,259.95129099 --> 00:07:48,769.95129099 Upwork represents a specific slice of the freelance market, primarily North American and European businesses. 74 00:07:49,84.95129099 --> 00:07:50,404.95129099 Hiring remote workers. 75 00:07:51,64.95129099 --> 00:07:59,404.95129099 We're not seeing what's happening inside large enterprises in different geographic markets or in industries that don't typically use freelance platforms. 76 00:07:59,854.95129099 --> 00:08:10,504.95129099 And while Upwork's data reveals companies hiring humans to clean up AI's mistakes, Stanford researchers have uncovered a more troubling pattern in broader employment theater. 77 00:08:11,254.95129099 --> 00:08:16,624.95129099 A comprehensive study published in August, 2025 by economists Eric Bruen. 78 00:08:17,659.95129099 --> 00:08:29,899.95129099 Paula Chand and Ru Chen analyzed payroll data from a DP America's largest payroll processor spanning millions of workers from late 2022 to mid 2025. 79 00:08:30,379.95129099 --> 00:08:47,209.95129099 Their findings reveal that early career workers aged 22 to 25 in AI exposed occupations, experienced a 13% relative decline in employment while employment for older workers in the same fields either remained stable or grew by six to 9%. 80 00:08:47,884.95129099 --> 00:08:53,614.95129099 This age-based displacement pattern could suggest that AI isn't just creating work slop that needs human cleanup. 81 00:08:54,34.95129099 --> 00:08:59,344.95129099 It's systematically replacing the entry level workers who would traditionally handle those tasks. 82 00:08:59,734.95129099 --> 00:09:10,729.95129099 The study found that workers aged 22 to 25 saw a 6% decline in employment in the most AI exposed occupations compared to increases for workers over 35. 83 00:09:11,629.95129099 --> 00:09:17,749.95129099 Significantly, this decline occurred only in rules where AI tends to automate rather than augment human work. 84 00:09:18,319.95129099 --> 00:09:25,429.95129099 So what we're really witnessing isn't just companies hiring humans to fix AI errors as part of an emerging AI cleanup economy. 85 00:09:25,819.95129099 --> 00:09:36,4.95129099 It's a fundamental restructuring of how work gets distributed across age groups and skill levels with profound implications for career progression and economic inequality. 86 00:09:37,764.73524625 --> 00:09:39,774.73524625 That's all for this week's AI roundup. 87 00:09:40,134.73524625 --> 00:09:44,64.73524625 If you found value in this breakdown, please leave a rating and hit subscribe. 88 00:09:44,94.73524625 --> 00:09:45,114.73524625 See you next week.
Hi everyone.
(00:00):
I'm Andy and this is the AI Breakdown.
Welcome to your weekly news edition where I'll cover what happened in AI last week, why it matters all in less than 10 minutes. 4 00:00:13,848.89137483 --> 00:00:20,658.89137483 First up, as I covered in my midweek episode last week, open AI used its dev conference to make a very loud statement. 5 00:00:21,48.89137483 --> 00:00:22,968.89137483 It's going all in on the enterprise. 6 00:00:23,388.89137483 --> 00:00:32,88.89137483 Reuters reports new partnerships with Spotify, Zillow, Mattel, and others, plus developer tools that let chat GPT plug directly into business apps. 7 00:00:32,418.89137483 --> 00:00:37,938.89137483 Think generating playlists inside Spotify, or filtering property searches on Zillow via chat. 8 00:00:38,688.89137483 --> 00:00:47,658.89137483 Nick Turley, who leads chat, GPT, said the product will evolve into something that feels a little bit more like an operating system, acting as a central hub for services. 9 00:00:48,663.89137483 --> 00:00:54,453.89137483 Sam Altman reinforced the shift with a huge focus on serving businesses and brushed off bubble chatter. 10 00:00:54,903.89137483 --> 00:01:00,123.89137483 Admitting whilst parts of AI are kind of bubbly, he insisted real value will get created. 11 00:01:00,783.89137483 --> 00:01:06,393.89137483 Stocks of several partners briefly popped on the news, which is always a nice sentiment check even if it fades. 12 00:01:07,83.89137483 --> 00:01:10,173.89137483 The bigger question everyone's asking is about sustainability. 13 00:01:10,593.89137483 --> 00:01:16,563.89137483 Open AI's ambitions from locking in massive compute to launching splashy apps requires serious spending. 14 00:01:16,998.89137483 --> 00:01:19,848.89137483 Sooner or later, the enterprise side has to pay the bills. 15 00:01:20,478.89137483 --> 00:01:22,68.89137483 It certainly looks like OpenAI. 16 00:01:22,98.89137483 --> 00:01:35,478.89137483 Want to make chat GPT the front door to work? One interface to search, summarize, draft query internal systems, and invoke third party tools they want to be your go-to for AI agents and assistance to. 17 00:01:35,748.89137483 --> 00:01:44,538.89137483 But the burden is on them to prove ROI at scale ensure that plugins, security and governance are robust enough for mission critical use. 18 00:01:45,453.89137483 --> 00:01:49,233.89137483 If they nail it, they stop being just a chat app and become a platform. 19 00:01:49,323.89137483 --> 00:01:59,412.36055526 Other products build on Meta is changing how it uses your AI chats and the internet, unsurprisingly, is not thrilled from mid-December. 20 00:01:59,442.36055526 --> 00:02:07,302.36055526 Conversations with Meta's AI across Facebook, Instagram, and threads will feed into add and feed personalization. 21 00:02:07,752.36055526 --> 00:02:12,762.36055526 There's no opt out if you use meta AI regions with stricter laws, including the eu. 22 00:02:13,92.36055526 --> 00:02:15,912.36055526 UK and South Korea are excluded for now. 23 00:02:16,782.36055526 --> 00:02:23,262.36055526 Met as Kristi Harris said that people's interactions are another piece of the input informing ads and recommendations. 24 00:02:23,742.36055526 --> 00:02:31,992.36055526 The company says sensitive topics like health or religion won't be used for ads, and it's keen to monetize more than a billion monthly AI users. 25 00:02:32,472.36055526 --> 00:02:37,152.36055526 Marketers love the intense signal, but privacy advocates call it invasive and creepy. 26 00:02:37,662.36055526 --> 00:02:40,572.36055526 Me personally, I'm not a fan of forced consent like this. 27 00:02:40,872.36055526 --> 00:02:46,602.36055526 It's just not the way things should happen today, and it risks leaving a sour taste with Meta's user base. 28 00:02:48,465.17234829 --> 00:02:56,565.17234829 OpenAI just gave a MD its biggest vote of confidence, yet a multi-year deal for hundreds of thousands of AI chips with deployment of instinct. 29 00:02:56,565.17234829 --> 00:03:03,345.17234829 GPUs at staggering scale starting in late 2026 on the order of six gigawatts of capacity. 30 00:03:03,870.17234829 --> 00:03:06,870.17234829 Investors went euphoric and EMD stock spiked. 31 00:03:07,260.17234829 --> 00:03:15,330.17234829 The agreement even includes warrants that could let OpenAI acquire up to 10% of EMD at a nominal 1 cent per share if milestones are met. 32 00:03:16,260.17234829 --> 00:03:20,520.17234829 MD's Forest Noro called the pact Transformative and analysts agree. 33 00:03:20,520.17234829 --> 00:03:25,230.17234829 It puts EMD firmly in the AI big leagues, even as Nvidia remains dominant. 34 00:03:25,590.17234829 --> 00:03:26,430.17234829 Why it matters. 35 00:03:26,760.17234829 --> 00:03:36,810.17234829 Open AI is securing compute at unprecedented levels to keep models responsive and expand services, while also diversifying beyond a single chip supplier for enterprises. 36 00:03:36,810.17234829 --> 00:03:43,170.17234829 Depending on AI platforms, more competition in silicon could mean better availability and pricing. 37 00:03:43,410.17234829 --> 00:03:45,450.17234829 As a MD capacity comes online. 38 00:03:45,870.17234829 --> 00:04:03,153.04171893 Skeptics note, OpenAI still relies on Nvidia and question whether anyone can afford this Arms race indefinitely, but the direction of travel's, clear compute is strategy, and the software winners will be the ones who lock down their supply as tensions increase between the US and China. 39 00:04:03,528.04171893 --> 00:04:15,948.04171893 China's latest retaliatory tariffs and tighter export controls, including new restrictions on rare earth minerals, are now reshaping the global AI hardware landscape and both a MD and Nvidia. 40 00:04:15,978.04171893 --> 00:04:29,148.04171893 According in the crossfire, in August, these chip makers agreed to an unprecedented deal requiring them to pay 15% of their revenue to the US government from AI chip sales to China as a condition for export licenses. 41 00:04:29,688.04171893 --> 00:04:38,628.04171893 While this arrangement lets a MD and Nvidia maintain a presence in China, it slashes profitability and adds new risks to supply chains and pricing strategies. 42 00:04:39,438.04171893 --> 00:04:46,218.04171893 EMD in particular has already absorbed substantial financial hits from inventory writedowns with up to $1.5 43 00:04:46,218.04171893 --> 00:04:49,218.04171893 billion in projected revenue losses this year alone. 44 00:04:49,908.04171893 --> 00:04:58,488.04171893 As the US Senate moves to prioritize domestic buyers and further restrict exports, AI companies find themselves navigating complex tariffs. 45 00:04:59,28.04171893 --> 00:05:01,848.04171893 Regulatory hurdles and shifting alliances. 46 00:05:02,178.04171893 --> 00:05:11,778.04171893 Making it clear that geopolitical tensions are now a core part of AI's strategic calculus and may shape which technologies reach global scale in the years ahead. 47 00:05:13,85.88230076 --> 00:05:15,515.88230076 Andro landed two heavyweight allies. 48 00:05:15,935.88230076 --> 00:05:25,115.88230076 IBM is integrated clawed into its enterprise software, starting with a developer environment that helps corporate engineers generate code and automate tasks. 49 00:05:25,400.88230076 --> 00:05:41,30.88230076 And the two companies published a Guide for Building Safe Enterprise Grade AI agents, IBM's software Chief Dinesh Neral Stressed Governance and Reliability, AI that fits how enterprises work, not experimental tools that create new risks. 50 00:05:41,510.88230076 --> 00:05:45,590.88230076 Meanwhile, Deloitte is rolling out Claude to 500,000 employees. 51 00:05:45,875.88230076 --> 00:05:46,655.88230076 Worldwide. 52 00:05:46,925.88230076 --> 00:05:54,185.88230076 Andros largest deployment to date, putting an AI assistant in the hands of auditors, consultants, and back office teams. 53 00:05:54,485.88230076 --> 00:05:58,475.88230076 Claude's reputation for safety and fewer hallucinations is a selling point. 54 00:05:59,75.88230076 --> 00:06:01,655.88230076 This is the shape of enterprise AI adoption. 55 00:06:02,45.88230076 --> 00:06:10,25.88230076 Embed assistance inside trusted tools per scale with training and oversight and measure time saved on research, drafting and coding. 56 00:06:10,715.88230076 --> 00:06:11,765.88230076 The upside is huge. 57 00:06:11,765.88230076 --> 00:06:12,545.88230076 Productivity. 58 00:06:12,890.88230076 --> 00:06:15,500.88230076 The risk is quiet errors slipping into deliverables. 59 00:06:15,800.88230076 --> 00:06:23,930.88230076 If teams treat AI like an Oracle, expect more firms to follow IBM's lead on governance checklists while pushing broad access. 60 00:06:23,930.88230076 --> 00:06:27,740.88230076 Like Deloitte Safety Plus scale is becoming the winning combination. 61 00:06:29,104.95129099 --> 00:06:30,304.95129099 Upworks September. 62 00:06:30,304.95129099 --> 00:06:32,974.95129099 Hiring data reveals something counterintuitive. 63 00:06:33,394.95129099 --> 00:06:34,954.95129099 Companies aren't replacing humans with ai. 64 00:06:36,199.95129099 --> 00:06:38,689.95129099 They're frantically hiring people to clean up after it. 65 00:06:39,169.95129099 --> 00:06:48,949.95129099 The report shows massive surges in translation work up 29% sales copywriting up 12% and quality assurance up 9%. 66 00:06:49,429.95129099 --> 00:06:53,184.95129099 Why they say AI is churning out what's now called work slop. 67 00:06:53,899.95129099 --> 00:06:59,59.95129099 Subpar output that needs human experts to validate, refine, and fix while work. 68 00:06:59,59.95129099 --> 00:07:08,299.95129099 Swap is a real thing, I believe it often stems from poor AI implementation skills, gaps, and lack of proper human oversight rather than technology limitations. 69 00:07:08,989.95129099 --> 00:07:15,589.95129099 According to BetterUp research, 40% of employees have already encountered AI hallucinations or mistakes. 70 00:07:15,919.95129099 --> 00:07:18,949.95129099 With each incident, taken an average of two hours to resolve. 71 00:07:19,579.95129099 --> 00:07:33,109.95129099 Meanwhile, project management hiring among small businesses exploded by 102%, suggesting companies are investing heavily in the human infrastructure needed to coordinate AI implementations properly. 72 00:07:33,889.95129099 --> 00:07:41,569.95129099 My view is this data is genuinely fascinating, but we need to be careful about drawing sweeping conclusions from a single platform. 73 00:07:42,259.95129099 --> 00:07:48,769.95129099 Upwork represents a specific slice of the freelance market, primarily North American and European businesses. 74 00:07:49,84.95129099 --> 00:07:50,404.95129099 Hiring remote workers. 75 00:07:51,64.95129099 --> 00:07:59,404.95129099 We're not seeing what's happening inside large enterprises in different geographic markets or in industries that don't typically use freelance platforms. 76 00:07:59,854.95129099 --> 00:08:10,504.95129099 And while Upwork's data reveals companies hiring humans to clean up AI's mistakes, Stanford researchers have uncovered a more troubling pattern in broader employment theater. 77 00:08:11,254.95129099 --> 00:08:16,624.95129099 A comprehensive study published in August, 2025 by economists Eric Bruen. 78 00:08:17,659.95129099 --> 00:08:29,899.95129099 Paula Chand and Ru Chen analyzed payroll data from a DP America's largest payroll processor spanning millions of workers from late 2022 to mid 2025. 79 00:08:30,379.95129099 --> 00:08:47,209.95129099 Their findings reveal that early career workers aged 22 to 25 in AI exposed occupations, experienced a 13% relative decline in employment while employment for older workers in the same fields either remained stable or grew by six to 9%. 80 00:08:47,884.95129099 --> 00:08:53,614.95129099 This age-based displacement pattern could suggest that AI isn't just creating work slop that needs human cleanup. 81 00:08:54,34.95129099 --> 00:08:59,344.95129099 It's systematically replacing the entry level workers who would traditionally handle those tasks. 82 00:08:59,734.95129099 --> 00:09:10,729.95129099 The study found that workers aged 22 to 25 saw a 6% decline in employment in the most AI exposed occupations compared to increases for workers over 35. 83 00:09:11,629.95129099 --> 00:09:17,749.95129099 Significantly, this decline occurred only in rules where AI tends to automate rather than augment human work. 84 00:09:18,319.95129099 --> 00:09:25,429.95129099 So what we're really witnessing isn't just companies hiring humans to fix AI errors as part of an emerging AI cleanup economy. 85 00:09:25,819.95129099 --> 00:09:36,4.95129099 It's a fundamental restructuring of how work gets distributed across age groups and skill levels with profound implications for career progression and economic inequality. 86 00:09:37,764.73524625 --> 00:09:39,774.73524625 That's all for this week's AI roundup. 87 00:09:40,134.73524625 --> 00:09:44,64.73524625 If you found value in this breakdown, please leave a rating and hit subscribe. 88 00:09:44,94.73524625 --> 00:09:45,114.73524625 See you next week.
Popular Podcasts
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
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.
The Bobby Bones Show
Listen to 'The Bobby Bones Show' by downloading the daily full replay.