AI Summary
5 min read"Biology is messy," says Abhi Mahajan, echoing a sentiment that cuts through the hype around AI-powered biotech. In this episode of Core Memory, guest host Erin A. Mirogi interviews Mahajan—now at the OpenAI Foundation, previously at Dyno Therapeutics and Novetic—to test the central claim of our era: that throwing vast intelligence at biology will dramatically accelerate cures and therapies. Their conversation is a careful, sober tour of where AI actually helps, where it hits walls, and why the bottlenecks are often physical, not intellectual.
Intelligence vs. Physical Grounding
Mahajan argues that the bottleneck in drug discovery depends entirely on the stage. In preclinical work, intelligence can help design better experiments, but the experiments themselves still require time, people, and physical materials. "At the clinical stage, probably intelligence is not all that useful," he says. "It's primarily like a human coordination problem." Even superintelligence can't speed up the logistics of recruiting patients or running trials across multiple sites. Post-approval, AI might be valuable for mining self-reported side-effect databases to catch problems like Vioxx, but that's a narrow use case.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Core Memory
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (03:26) **Intelligence vs. Physical Experimentation as the Bottleneck** - Abhi argues the bottleneck depends on the stage of drug discovery.
- 2 (07:46) **Why AI's Success in Math Doesn't Transfer to Biology** - The key difference is the speed of feedback loops.
- 3 (10:07) **The Real Data Acceleration: Next-Generation Sequencing** - A historical example of a technological leap that wasn't just automation.
- 4 (12:04) **The TIGIT Failure: Would Superintelligence Have Predicted It?** - Abhi tested a frontier model on a historical drug failure and found it unconvincing.
- 5 (17:20) **What Data is Missing to Predict Human Outcomes?** - The simple, uncomfortable answer is human data itself.
- 6 (21:00) **The Noetic Thesis: Using AI to Stratify Clinical Trials** - A detailed look at a specific, practical application of AI in biology.
- 7 (23:17) **Connecting Mouse Data to Human Reality** - Abhi explains a novel use of Noetic's model to make mouse experiments more relevant.
+ Full timestamped outline available in the app
Show Notes
Every couple of days, a new AI-meets-biotech startup sends me a pitch, and the pitches all sound the same. AI is amazing, and it can now tame biology. We’re on the cusp of producing new cures for diseases at a rapid rate, and these startups know just how to get the job done. To which you should wonder – is this at all true?
On this week’s Core Memory episode, we have brought in the big brains to answer – or at least debate – this question. Our resident science whiz and medical professional Eryney Marrogi takes on the hosting duties to interview Abhishaike Mahajan, he of Owl Posting and now OpenAI Foundation fame.
These two men of science take hold of this AI/biotech moment to explore where we’re seeing the most promise from this new wave of technology and where the messiness of biology continues to confound AI. They also get into some lighter topics like our collective death from AI bioweapons and China’s impending biotech dominance. Good times.
You’re all probably buzzing over Moderna’s stock price doubling and the end of cancer. So ride that enthusiasm right into this podcast and see how you’re feeling on the other side.
If you can spare a minute, please do us a favor by filling out this ever so brief survey, so we can learn a bit more about our subscribers. Help us be better for you. Thanks!!
OUR SPONSORS
SendCutSend
You know who else makes stuff for America? That would be SendCutSend. If you want to celebrate our great nation by building a metal part, then head on over to SendCutSend where you’ll get a 15 percent discount thanks to Core Memory on whatever you’re trying to build. We believe in you.
Brex
The Core Memory podcast is also sponsored by Brex, the intelligent finance platform built to help companies spend smarter and move faster.
Did we go to Texas, find a telescope ranch and then obtain an entire nebula in Brex’s honor? Oh yes, we did.
We run on Brex and so should you. Learn more about Brex right here.
Timestamps (they link out to YouTube)
0:00 Intro4:24 Is Intelligence the Bottleneck? More from this podcast