With industry dominating large-scale model training, academia’s comparative advantage has shifted. Its focus should be on exploring high-risk, unconventional concepts like new algorithms and hardware-aligned architectures that commercial labs, focused on near-term ROI, cannot prioritize.

Scientists constrained by limited grant funding often avoid risky but groundbreaking hypotheses. AI can change this by computationally generating and testing high-risk ideas, de-risking them enough for scientists to confidently pursue ambitious "home runs" that could transform their fields.

An AI model disproved a mathematical conjecture not through a flash of creative genius, but by methodically applying a known technique from a different math subfield. This highlights AI's current strength: synthesizing vast, disparate human knowledge rather than generating truly novel, alien ideas. It's an exhaustive librarian, not an intuitive genius.

Early AI models advanced by scraping web text and code. The next revolution, especially in "AI for science," requires overcoming a major hurdle: consolidating and formatting the world's vast but fragmented scientific data across disciplines like chemistry and materials science for model training.

Historically, generating a good hypothesis was the most prestigious part of science. Now, AI can produce theories at near-zero cost, overwhelming traditional validation systems like peer review. The new grand challenge is developing scalable methods to verify and filter this flood of AI-generated ideas.

The internet is an insufficient training ground for scientific AI because most crucial information—including failed experiments, negative data, and nuanced procedural details—is never published. This undocumented knowledge, what scientists call "good hands," represents a major data bottleneck for building truly intelligent scientific models.

The ultimate goal isn't just modeling specific systems (like protein folding), but automating the entire scientific method. This involves AI generating hypotheses, choosing experiments, analyzing results, and updating a 'world model' of a domain, creating a continuous loop of discovery.

A major frontier for AI in science is developing 'taste'—the human ability to discern not just if a research question is solvable, but if it is genuinely interesting and impactful. Models currently struggle to differentiate an exciting result from a boring one.

Current LLMs fail at science because they lack the ability to iterate. True scientific inquiry is a loop: form a hypothesis, conduct an experiment, analyze the result (even if incorrect), and refine. AI needs this same iterative capability with the real world to make genuine discoveries.