Inside Science: Hot Years, Viral Origins, Move 37 and Moon-Navigating Ants

In this episode of Inside Science, Tom Whipple guides listeners through a sequence of science stories across climate, virology, artificial intelligence, and biology. The opening discussion with Laura Wilcox examines evidence that recent years are hotter and what could drive a continuing acceleration in warming, including air-cleaning effects, methane emissions, and clouds in the Pacific. The programme then delves into SARS-CoV-2 evolution, explaining how genomic data suggest the virus was bat-originating and that claims of a lab-adapted prehistory are not supported by the evidence. A Go and AI segment revisits Move 37 from AlphaGo, highlighting a moment of novel strategy and broader implications for creativity in AI. The show ends with ants using a time compensated lunar compass to navigate, illustrating surprising forms of intelligence in nature.

Overview

This Inside Science episode stitches together climate science, genomic epidemiology, game AI, mathematics and animal navigation, exploring how new data and methods reshape our understanding of each field.

Climate change: Are recent years hotter than expected?

Tom Whipple speaks with atmospheric scientist Laura Wilcox about a new analysis that reports warming at about 0.35 degrees Celsius per decade, faster than some past expectations. Wilcox explains how the study filters out year-to-year variability caused by sea surface temperatures, volcanic eruptions, and solar changes to reveal a statistically significant change in the warming rate. The discussion covers why this could be permanent or temporary, with cleaner air potentially accelerating near-term warming by removing cooling particles that reflect sunlight. The segment also notes methane emissions and cloud changes in the Pacific as contributing factors, and describes climate model experiments that test these hypotheses across multiple models. "they're able to show a statistically significant change in the rate of warming" - Laura Wilcox, atmospheric scientist.

Genomics: SARS-CoV-2 origins and evolution

Roland Pease interviews Joel Wertheim about genomic sleuthing on SARS-CoV-2 before and after the outbreak. The analysis tests whether the lineage shows pre-adaptation in animals or laboratories and uses comparisons to gain-of-function studies to calibrate sensitivity. Wertheim explains how the pace and tempo of mutations shift when the virus jumps from bats to humans, and how the approach can identify whether a lab-constructed sequence would bear detectable signatures. A key takeaway is that the outbreak most closely resembles a natural bat lineage up to the moment of human infection, after which human adaptation accelerates. "the viral genomes can test that by measuring the pace of evolution before and after the outbreak" - Joel Wertheim, UC San Diego.

AI and Go: Move 37 and beyond

Tory Grapel from DeepMind explains the difference between Deep Blue’s brute-force chess approach and AlphaGo’s hybrid architecture that blends fast, intuitive reasoning with deliberate search. The Move 37 moment is highlighted as evidence of a computer’s creative capacity, producing a move that seemed counterintuitive yet effective. The conversation broadens to discuss implications for scientific discovery, from medicine to materials science, and considers how AI could expand human creativity rather than replace it. Grapel emphasizes the rapid pace of AI progress and the need to adapt human approaches to these new tools. "Move 37 opened doors for us" - Torre Grapel, DeepMind.

AI in mathematics: formal verification and new frontiers

The discussion shifts to how formal verification languages can ensure the validity of computer-generated proofs, addressing concerns about AI “cheating” on proofs. The segment highlights a collaboration with the AI reasoning model Gauss to formalize Marina Wazovska's sphere-packing theorem, demonstrating how translating complex proofs into machine-checkable steps can drastically reduce human labor. The sphere packing problem is explained at a high level, including the breakthrough work on 8- and 24-dimensional packings and what those dimensions mean for geometry. "formal verification languages... Boil down a piece of work and verifying it" - discussion of formal verification in mathematics.

Ant navigation: lunar compass and time compensation

In a lighter closing note, scientists studying bull ants describe a time compensated lunar compass, showing how ants derive heading by tracking the moon’s movement and computing the relative angle to their nest over several hours. The segment ends on a note about the diversity of intelligence in nature, contrasting human-level AI progress with surprising natural strategies.