🔬 An Oscar, Two Asteroids, and the Algorithm in Your sklearn: John Platt on AI for Science

John Platt, an Oscar‑winning researcher and the mind behind classic algorithms such as Platt‑scaling and SMO, has spent much of his career at the crossroads of artificial intelligence and science. Together with Google, he leads the Empirical Research Assistance (ERA) project, a tool that automates the search for solutions to scientific problems that can be expressed through a scoring function. ERA uses a language model—Gemini or another LLM—to maintain a tree of prior experiments and, through a variant of Monte Carlo Tree Search combined with the Upper Confidence Bound rule, selects and mutates promising notebooks. The process, comparable to an indefatigable research assistant, blends LLM‑generated mutations with shared learning across branches, yielding significant progress from version 2.0 to 2.5.

The system has already produced at least ten publications, including studies aimed at mitigating climate change. A notable example is the reduction of contrail impact, the ice trails that account for roughly 1 % of anthropogenic warming. ERA identified a model that incorporates previously omitted confounding factors, enabling a more accurate estimate of avoided climate impact by adjusting flight altitude.

This development is significant because it demonstrates how AI‑driven automation can accelerate scientific research and provide concrete tools to tackle global challenges such as planetary warming. It also illustrates the potential of optimization algorithms to turn experimental data into practical solutions.

Read the original article on Latent Space

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.