AI agents could reshape scientific discovery

AI agents could offer a broader route for accelerating scientific discovery than single-purpose systems such as AlphaFold. Google DeepMind scientists shared the Nobel Prize in Chemistry in 2024 for AlphaFold, which predicts protein structures using a dataset of roughly 170,000 experimentally validated structures assembled over 53 years. Similar datasets may be impractical in many fields, making agents appealing because they can model the iterative process of research rather than applying one powerful method to one narrow problem.

The idea of a “censorship-industrial complex” has moved from right-wing online circles into the Trump administration, after spreading for years as a theory that conservative and populist speech was being suppressed. MIT Technology Review plans a virtual Roundtables session on Thursday, August 13, at 19:00 GMT / 2:00 pm ET / 11:00 am PT with Eileen Guo and Amy Nordrum on its origins, trajectory, and implications for democracy and the internet.

Other major technology developments include an Amazon data center in Texas permitted to release up to 33 million tons of CO2, OpenAI pausing work on its Astra AI model after tests found it could launch cyber-attacks autonomously, and a report that North Korean hackers built AI tools for spear-phishing. China’s firms control 97% of global humanoid shipments, Taiwan is expanding drone use to deter China, Apple is testing Chinese memory chips, and an Israeli startup is tied to rogue AI tests at OpenAI, Anthropic, and Meta.