Podcast Summary: No Priors – “AI Consolidation, Biotech Opportunities, and World Models with Sarah and Elad”
Date: May 29, 2025 | Hosts: Sarah Guo and Elad Gil

Overview

In this episode, Sarah Guo and Elad Gil take stock of the current landscape of artificial intelligence, focusing on trends of market consolidation, unexploited commercial opportunities (particularly in biotech), and the future of “world models” and reinforcement learning in AI research. The conversation weaves through the dynamics that determine market winners, missed opportunities in biotech innovation, and the challenges of training AI agents to reason and act, not just predict.

Key Discussion Points & Insights

1. State of the AI Market: Crystallization and Ongoing Fluidity

(00:08–04:56)

• Elad observes that after years of explosive, chaotic innovation, “a bunch of markets have sort of consolidated and it’s kind of clear now who are the likely players or winners in like two or three big areas.” (00:18)
  • For LLMs (Large Language Models), coding tools, and medical scribing, a handful of companies like Cursor, Codeium, Cognition, and Microsoft’s Copilot are now front runners.
  • In domains like sales, productivity, and finance, the field remains wide open, possibly due to insufficient product approaches or models not being capable enough yet.

“For the first time in like two years... finally we're hitting a period where at least a subset of things are consolidating back down. And again, these may not be the winners five years from now, but they definitely seem to be emerging as the winners for the next two years.”
— Elad Gil (02:35)

• Sarah stresses the evolving “race to find the verticals of relevance” and sees clear pathways for applications leveraging proprietary data and knowledge creation.

2. Market Entry and The Likelihood of New Entrants Succeeding

(04:56–06:05)

• They discuss open-sourcing trends in coding agents (e.g., Microsoft open-sourcing Copilot) and new workflows potentially enabling further shifts in market leadership.
• The asynchronous vs. synchronous approach to coding automation is dissected—how quickly can async agents catch up in quality?

3. Consolidation Mechanics: Mergers, Acquisitions, and Competitive Mindsets

(06:05–10:08)

• Elad proposes that two main types of consolidation will play out:
  • Product consolidation (feature convergence)
  • Actual mergers and acquisitions
• He encourages startup founders to consider merging if they’re #1 and #2 in a new field to better compete with incumbents, likening it to the historic X.com/PayPal merger.
  • Typical obstacles: ego (“who will run it?”), overthinking culture fit, valuation disagreements.

“Put that aside and just go win… like, who cares? …Do you want to fight it out for the next five years or do you want to go win and then your battleground shifts to the incumbents versus another startup?”
— Elad Gil (07:04)

• Sarah adds that founders/investors often see this as capitulation, missing that it is “capitulating in service of winning.” (08:56)

4. Missed Opportunities in Biotech: Science Outpacing Commercialization

(10:08–16:54)

• Elad shares biotech trends he finds fascinating but notes he’s not an active biotech investor.
• Examples:
  • Recent Japanese fertility advances (turning body cells into sperm/egg, allowing for new forms of parenthood)
  • Vast untapped work in rejuvenation science (e.g., anti-aging drugs, reversing hearing loss, regrowing teeth)
  • Existing science enabling large markets is available, but few entrepreneurs working on them.

“You could potentially replace them without permission. The ramifications of this stuff is pretty crazy if you think about it… but also societally, it’s so impactful in terms of what you could do with that.”
— Elad Gil (12:23)

• Four reasons for biotech stagnation:
  1. Structure of biotech industry (very large, old companies; few new $50B+ companies since the 1980s)
  2. Funding model designed for “flipping” startups to pharma giants, not building independent giants
  3. Regulatory hurdles (FDA, endpoints)
  4. “Prudishness” around working on “commercial,” “low-status” problems (e.g., fixing wrinkles vs. “pure” science)

5. World Models, Reinforcement Learning, and the AGI Path

(16:54–22:56)

• Sarah offers an accessible primer on the next phase of AI: moving beyond prediction to reasoning and action (e.g., planning, reading, taking action, tool use).
• Foundational AI research is stuck between brittle behavior cloning (“monkey see, monkey do”) and the challenge of building rich simulated environments and reward functions for reinforcement learning.
  • RL works well in games (clear rules, clear rewards), but the “real-world” lacks that structure.
  • Agents need universality/diversity or they “overfit instead of adapting.”
• Active research is focused on building better “world models” or collecting richer data traces for training.

“Everybody talks about agents, what people want to do from here. The way people think about AGI is not just predicting text, right? They want to go to broader intelligence into taking actions… it could be like planning or reading documents and drawing conclusions, using tools, receiving feedback… taking a series of actions in pursuit of a goal beyond just sequential text generation.”
— Sarah Guo (17:42)

• Elad expands, likening it to how AlphaGo discovered unprecedented solutions, or biological systems that “evolve” unexpected mechanisms that can teach us new approaches.

“If coding shifted from, hey, let’s copy how people write code into let’s just solve this problem. How different is the type of code that’s written… you’ve created a utility function with an unconstrained approach to actually figuring it out.”
— Elad Gil (21:34)

6. Speculation & Philosophy: Sci-Fi, Mind Uploading, and Future Possibilities

(22:56–24:44)

• Sarah brings up the TV show “Pantheon” about AI and mind uploading and how “uploaded beings” can discover new ways of being and thinking, breaking out of old constraints (23:18).
• Elad riff on the philosophical implications of uploading, including profound shifts in human capability, identity, and agency.

“If a human upload were to occur, what does a transhuman species look like and what are the capabilities set that aren’t a priori obvious that you suddenly expose?”
— Elad Gil (23:47)

• Sarah notes that some mind-modulation will arrive “before upload”—e.g., “ultrasound devices” for emotional/attentional tuning are on the near horizon.

Notable Quotes & Memorable Moments

• “AI is the one market in my career where I’ve sort of consistently said the more I learn, the less I know.”
  Elad Gil (00:18)

• “Maybe you have to go get proprietary data sources that you can retrieve against and like get distribution and then ideally have users that can create or derive or extend knowledge from that.”
  Sarah Guo (03:18)

• “It’s low status. So how, how dare you work on fixing wrinkles as a scientist? You need to be doing something that’s much more pure, et cetera, et cetera. So there’s… sort of, I have to call it, prudishness around commerciality.”
  Elad Gil (16:54)

• “The way people think about AGI is... not just predicting text, right? They want to go to broader intelligence into taking actions.”
  Sarah Guo (17:42)

• “You could potentially replace them without permission. The ramifications of this stuff is pretty crazy... suddenly anybody could become an egg or sperm donor in any capacity.”
  Elad Gil (12:23)

• “Put that aside and just go win... who cares?”
  Elad Gil (07:04)

• “The pie gets bigger if you do that because you’re focused on just winning the market versus competing with each other.”
  Elad Gil (09:24)

Timestamps for Key Segments

• 00:08 – 04:56: AI Market State & Early Consolidation
• 04:56 – 06:05: Market Entry, Open Sourcing, and Code Workflow
• 06:05 – 10:08: Consolidation Mechanics, Mergers, and Competitive Psychology
• 10:08 – 16:54: Biotech: Fertility, Rejuvenation, and Why Innovation Lags
• 16:54 – 22:56: AI World Models, Reinforcement Learning, and Path to AGI
• 22:56 – 24:44: Sci-Fi, Mind Uploading, and Future Human Capabilities

Tone & Language

• The episode is lively, curious, and critical—with both hosts candid about what makes certain markets “crystallize” and others remain stubbornly fragmented. They toggle between high-level technical detail, market insight, and playful speculative thinking.

For further exploration, the transcript and full episodes are available at no-priors.com.