Episode

Where Is All the A.I.-Driven Scientific Progress?

Podcast: Hard Fork
Published: Dec 26, 2025
Duration seconds: 2368
Processing state: processed
Canonical source: https://www.nytimes.com/column/hard-fork
Audio: https://dts.podtrac.com/redirect.mp3/pdst.fm/e/pfx.vpixl.com/6qj4J/pscrb.fm/rss/p/nyt.simplecastaudio.com/3e43d072-f8a5-430f-bc8e-4c70aafdf3c7/episodes/2adf7272-2fc8-4749-abd7-8aca14809182/audio/128/default.mp3?aid=rss_feed&awCollectionId=3e43d072-f8a5-430f-bc8e-4c70aafdf3c7&awEpisodeId=2adf7272-2fc8-4749-abd7-8aca14809182&feed=l2i9YnTd
JSON: /v1/public/podcasts/hard-fork/episodes/where-is-all-the-a-i-driven-scientific-progress
Markdown: /podcast/hard-fork/where-is-all-the-a-i-driven-scientific-progress.md

Actions

POST https://stenobird.com/v1/public/podcasts/hard-fork/episodes/where-is-all-the-a-i-driven-scientific-progress/transcription-requests
Idempotently request low-priority transcript generation for this episode.
GET https://stenobird.com/podcast/hard-fork/where-is-all-the-a-i-driven-scientific-progress.md
Read the agent-friendly Markdown representation of this episode resource.

Summary

AI is currently a powerful tool for accelerating data analysis and protein modeling, but it lacks the fundamental knowledge required to solve complex biological mysteries like Alzheimer's. The true frontier lies in using generative models to scale scientific hypotheses without losing the serendipity of human error.

Topics

Artificial Intelligence
Scientific Discovery
Biotechnology
Generative Models
Protein Folding
Machine Learning
Computational Biology
Scientific Automation

Highlights

Main idea: AI excels at generative modeling for proteins and antibodies but cannot invent knowledge that doesn't yet exist
Practical takeaway: AI agents are expected to begin infiltrating laboratory workflows and automating hypothesis generation by 2026
Failure mode: High compute costs, such as $200 per prompt, currently limit the scalability of advanced scientific models
Main idea: The value of AI in science lies in its ability to accelerate the verification of hypotheses rather than just generating them
Critical nuance: Scientific progress requires 'noise' and serendipity, which can be intentionally integrated into generative models to mimic biological evolution

Chapters

4:40 The Scientist's Perspective: An introduction to Sam Rodriguez's background in physics and biotech and the current state of AI in research.
8:25 Scaling Science with AI: Exploring the concept of 'AI scientists' and the potential to expand scientific output beyond human capacity.
11:30 The High Cost of Intelligence: A discussion on the massive computational expenses and the economic barriers to running advanced scientific models.
14:50 Data, Analysis, and Reliability: Evaluating how AI handles the core scientific loop of gathering data and drawing conclusions.
18:15 AI as a Research Assistant: Analyzing the utility of AI for analyzing existing experimental datasets and the necessity of human verification.
21:25 The Power of Generative Models: The excitement surrounding generative models in creating novel proteins and antibodies from scratch.
27:50 The Limits of Infinite Intelligence: Why even infinite computing power cannot solve scientific problems where the underlying biological knowledge is missing.