The Artificial Intelligence Show

Episode #188: AI Trends for 2026, Google DeepMind AI Predictions, Gemini 3 Flash, AI World Models & Are AI Job Losses Overblown?

Hosts: Paul Roetzer and Mike Kaput
Date: December 23, 2025

Episode Overview

In this final episode of 2025, Paul and Mike share their informal, instinct-driven outlook on where AI is heading into 2026, discussing technology, business, and societal trends. The episode features in-depth breakdowns of recent interviews with DeepMind leaders Demis Hassabis and Shane Legg, focusing on the road to AGI, world models, and economic transformation. The hosts also analyze timely news, such as the Gemini 3 Flash release, AI job disruption debates, major fundraising moves in the sector, and mounting political conversations around AI infrastructure and regulation. The tone throughout is candid, thoughtful, and focused on both the opportunities and looming challenges AI presents.

Key Discussion Points and Insights

1. AI Trends to Watch in 2026

(Timestamps 07:05–31:21)

A. Technology Trends

• Agent-to-Agent Communication & Commerce
  • AI agents will start making decisions and transactions on behalf of users and companies.
  • "People or things visiting your website might not be humans, they might be people's agents...we're going to start to move into this phase where agent-to-agent becomes more commonplace." (09:00, Paul)
• Personalization of AI Assistants
  • Assistants will deeply integrate user preferences, tone, and context, leading to differentiated user experiences.
  • "You're just going to get so attached to whatever tool has the most context and personal info about you…the only thing that differentiates between the models." (11:02, Mike)
  • Concerns around the “stickiness” and potential lock-in effect due to this personalization.
• Reliability of Agents for Long-Horizon Tasks
  • Major progress expected in agents successfully performing increasingly complex, multi-step tasks across industries.
  • "Meter's research showed AI models had a 50% chance of completing an expert’s one-hour task, doubling every seven months." (14:10, Paul)
• Verification & AI Evaluation
  • Growing need for AI output verification as agents take on more work, spawning new roles.
  • "AI verification is going to be such an in-demand need and skill...maybe even job." (15:18, Mike)
• Progress Across Key Dimensions
  • Rapid advances in multimodality, reasoning, world models, continual learning, expanding context windows, and memory.
  • Increasing indistinguishability between AI-generated and real content (image, video, audio, etc).
  • Robust efforts in robotics and consumer hardware (phones, glasses).
  • Efficiency: Frontier models moving on-device (e.g., Apple’s likely coming leap).
  • The relentless march toward AGI and eventual superintelligence.

B. AI in Business

• From Piloting to Scaling
  • 2026 will see more companies move from experiments to deep, enterprise-wide integration of AI.
• Personalized AI Use Cases
  • Shift from "here’s your co-pilot license, figure it out" to targeted, workflow-tied copilots and GPTs for each department.
• AI Reasoning in Workflows
  • Underutilization of AI for higher-order reasoning and more complex tasks is starting to change.
• Elevating AI Literacy
  • AI literacy is now the #1 sought skill on LinkedIn and essential for employability.
  • “By this time next year it'll be assumed that if you want a job, you better be AI literate." (23:03, Paul)
• Evolving Evaluation Metrics
  • Move away from generic benchmarks (e.g., IQ tests) to custom, workflow-tied evals for organizations and individuals.
• Future of Work Hypothesis
  • AI will trigger a “radical transformation of talent, teams, and organizational structures.”
  • Fewer people needed for the same work, more automation, distributed entrepreneurship, and premium on data.

C. Societal Impact and Backlash

• Mounting Political Friction
  • Over 1,200+ pieces of AI legislation in the US alone; political fights on state vs. federal policy are intensifying.
• Jobs and Economy
  • While data shows little short-term disruption, hosts warn of “significant job disruption” by early 2026.
  • Entry-level workers and new graduates may face unprecedented challenges.
• Booming Tech Sector and IPOs
  • Anticipation of blockbuster IPOs: OpenAI, SpaceX, Anthropic, and others.
• Societal Reckoning
  • Parents, schools, and businesses must all grapple with AI’s new realities; 2026 is likely the year "societal backlash becomes real."
  • “I just have this nagging feeling. 2026 is the year where the societal backlash against AI becomes real.” (30:59, Mike)
  • Paul’s optimism: Breakthroughs in science, medicine, and creativity will also accelerate.

2. DeepMind Leaders’ Predictions & the Road to AGI

(Timestamps 31:59–47:53)

Demis Hassabis Interview (DeepMind Podcast)

• Urges a pivot beyond language models to “world models” capable of simulating physical cause and effect.
• Advocates for AGI as a “comparative simulation” to illuminate what makes us uniquely human—such as creativity and emotion.
• Warning: AGI's transition will be “10 times bigger and happen 10 times faster than the industrial revolution.”
  • “We’re about to experience...100 years of change in the next 10 years.” (40:47, Paul citing Demis)
• Demis emphasizes the importance of consistency, reasoning, and especially “continual learning”—where models update based on experience, like humans.
• Points to Google’s strength: deep research investment and sustainable product integration (e.g., Gemini in Workspace, YouTube, Search).
• Cautions that the “AI bubble” may be overhyped short-term, but underhyped in the medium and long term.

Shane Legg Interview (DeepMind Podcast)

• Reaffirms prediction: 50/50 chance for “minimal AGI” by 2028 (performs cognitive tasks like a typical human).
  • “In a few years, where you needed 100 software engineers, maybe you need 20 using advanced AI tools.” (45:10, Shane via Paul)
• Emphasizes “generality”—a true AGI must perform across a wide array of human-level tasks.
• Discusses adversarial testing as a possible AGI benchmark (humans try to find a cognitive task the AI can't do).
• Underscores that economic and social disruption will be “uneven” but profound, starting in coding, soon affecting all knowledge work.
• Mike and Paul repeatedly agree that listening to Demis and Shane makes them want these researchers “at the front edge” for the sake of humanity.

3. Rapid-Fire AI News & Analysis

AI & Job Loss Debate (Brooking Study, David Sacks, Molly Kinder)

(Timestamp 47:53–56:03)

• Brookings report: No significant AI-driven layoffs yet; backed by venture capitalist/White House AI czar David Sacks.
• Molly Kinder (Brookings) clarifies: short-term fears overblown, but mid-long-term disruption "must be taken seriously.”
  • “The public isn’t fretting about headlines today; they’re nervous about what lies ahead... The task isn’t to sell AI to a skeptical public—it’s to shape AI’s deployment and our policy infrastructure so it actually serves them.” (Excerpt from Molly Kinder, 51:44)
• Hosts warn that political figures selectively use data to serve agendas; true impact will be clearer heading into 2026.

Gemini 3 Flash Release

(Timestamp 56:03–59:36)

• Google’s faster, lighter Gemini model launches globally, delivering high-level reasoning at triple the speed and 30% fewer tokens.
• Becomes the default in free Gemini globally, deployed in AI-powered search.
• “What was the frontier today will be available way cheaper, way faster, and eventually, on devices.” (57:26, Paul)
• Models are being stratified for speed vs depth—most users don’t need “pro” versions.

OpenAI Fundraising & Image Model Refresh

(Timestamps 59:37–64:12)

• OpenAI pursues a $100B raise at up-to-$830B valuation; Amazon possibly investing $10B and providing Trainium chip power.
• Hosts speculate on OpenAI’s and SpaceX’s valuation “race," both likely headed to mega-IPOs in 2026.
• OpenAI’s new GPT Image 1.5 improves visual generation but still trails Google’s Nano Banana Pro.
• Paul: “Momentum has certainly swung to Google on basically every front—the model, the image model, the video model… the world models.”

Data Center Water Usage Correction

(Timestamp 64:12–68:18)

• Author Karen Hao issues a correction in her book about Google data centers’ water usage, after a statistical unit error is found.
• Debate politicized, with some accusing environmental critiques of “hoaxes.”
• Paul empathizes with Hao, emphasizes journalistic integrity, and condemns bad-faith Internet pile-ons.

Politics & AI: Regulation, Data Centers, and Roadmaps

(Timestamp 68:18–72:51)

• Bernie Sanders calls for a moratorium on new data center construction; Florida’s Governor DeSantis proposes an AI Bill of Rights.
• Andreessen Horowitz (A16Z) pushes for federal AI standardization.
• Paul: “Anytime it’s political, you have to understand why people are doing what they’re doing and saying what they’re saying... Read the extremes, then look for the middle.”

Explosion of World Models and Simulators

(Timestamps 72:51–77:31)

• General Intuition and Runway Labs secure huge investments to develop “universal world models” that simulate physical reality.
• Such models are seen as key to robotics, science, and even replacing traditional laboratories.
• “The AI models to reach human-level intelligence must experience the world the same way we do—world models are that bet.” (74:13, Paul)

US Tech Force: Talent Embedding in Government

(Timestamps 77:31–79:27)

• White House launches initiative to embed top technical talent into federal agencies, in partnership with AWS, Microsoft, Nvidia, and OpenAI (curiously, not Anthropic).
• Paul is skeptical on practicalities (“who’s leaving OpenAI for $150K/year to work in government?”) but supportive of the concept.

Notable Quotes

• "2026 is the year where the societal backlash against AI becomes real. I think we're going to...be having conversations about people that looked at us a weird way because of what we end up doing." — Mike Kaput (00:00, 30:59)

• “We're moving into a phase where agent-to-agent [communication] becomes more commonplace... I think 2026 will be the year this idea becomes real.” — Paul Roetzer (09:00)

• “Personalization is going to be basically at some point the only thing that differentiates between the models.” — Mike Kaput (11:02)

• “AI verification is going to be such an in-demand need and skill...maybe even job.” — Mike Kaput (15:18)

• “The future of work hypothesis is that AI advancements and agent capabilities will force a radical transformation of talent, teams, and structures. Fewer people will be needed to do the work.” — Paul Roetzer (25:15)

• "If you want to have a job, you better be AI literate and you better be investing in that." — Paul Roetzer (23:04)

• “Society is going to have to come to grips with where we are. Businesses, schools, parents—2026 is when these issues we’ve ignored become real.” — Paul Roetzer (29:41)

• “The transition to AGI will be 10 times bigger and 10 times faster than the industrial revolution.” — Demis Hassabis, via Mike Kaput (32:18, cited throughout)

• “We are at the absolute frontier of science...it’s exhilarating, but we understand the enormity of what’s coming… in the next 10 years we’ll experience a hundred years of change.” — Demis Hassabis (40:47, via Paul)

• "In a few years, where you needed 100 software engineers, maybe you need 20 using advanced AI tools." — Shane Legg, via Paul Roetzer (45:10)

• "The public isn't just fretting about headlines today; they are nervous about what lies ahead...The task ahead isn't to sell AI to a skeptical public. It is to shape AI's design and deployment and our policy infrastructure so it actually serves them." — Molly Kinder, Brookings (51:44, read by Paul)

Important Timestamps & Segments

• AI Trends to Watch: Technology, Business, and Society (07:05–31:21)
• DeepMind’s Road to AGI & World Models (31:59–47:53)
• AI Job Losses Debate (Brooking Study) (47:53–56:03)
• Gemini 3 Flash and Model Efficiency (56:03–59:36)
• OpenAI’s Fundraising & Image Model Update (59:37–64:12)
• Karen Hao Book Correction (64:12–68:18)
• Politics of AI: Regulation & Data Center Moratorium (68:18–72:51)
• World Models Investment Frenzy (72:51–77:31)
• US Tech Force Government Program (77:31–79:27)

Memorable Moments

• Paul’s analogy on 5-year business plans being obsolete, needing company reinvention every 6-12 months due to AI change cycles. (42:00)
• Mike’s prediction of impending societal pushback—"People looking at us weird because of what we end up doing." (30:59)
• A candid admission: “Verification of that information is critical... Until we staffed and had a vision for how to do that, we couldn't apply deep research to what we were doing daily.” (15:52, Paul)
• The recurring advice to never take headlines at face value, emphasizing context and motivation behind public statements—especially in politics. (54:47–56:03, Paul & Mike)

Conclusion

The hosts close out 2025 with gratitude, optimism for scientific breakthroughs—and urgency about societal and economic reckoning with AI in 2026. Listeners are urged to build AI literacy, expect rapid disruption (and opportunity), and to stay skeptical and critical of headlines and narratives. For deeper context, listeners are encouraged to check out episode 141 ("The Road to AGI"), which the hosts say is still fully relevant.

Resource Links & Further Listening

• [Episode 141: The Road to AGI] (to be provided in show notes)
• LinkedIn Post by Molly Kinder (as referenced)
• DeepMind Podcast episodes with Demis Hassabis and Shane Legg

Stay Curious & Explore AI!