The Download: beyond CRISPR, and OpenAI’s superalignment findings

Google DeepMind used a large language model to solve an unsolvable math problem

The news: Google DeepMind has used a large language model to crack a famous unsolved problem in pure mathematics. The researchers say it is the first time a large language model has been used to discover a solution to a long-standing scientific puzzle—producing verifiable and valuable new information that did not previously exist.

Why it matters: Large language models have a reputation for making things up, not for providing new facts. Google DeepMind’s new tool, called FunSearch, could change that. It shows that they can indeed make discoveries—if they are coaxed just so, and if you throw out the majority of what they come up with. Read the full story.

—Will Douglas Heaven

Needle-free covid vaccines are (still) in the works

Covid shots do an admirable job of boosting our immune response enough to protect against serious illness, but they don’t boost immunity in the one spot we’d like them to: our airways.

That’s why researchers have been working on vaccines you breathe into your lungs or spray into your nose. The idea is that these vaccines will elicit an immune response in the mucous membranes of your respiratory tract that might help stave off infection or, if you do become infected, make you less likely to transmit the virus.

These “mucosal” covid vaccines aren’t available in the US or Europe, but they are in other parts of the world. So when will the US get its first mucosal covid vaccine? What will it look like? And will it work as intended? Read the full story.

—Cassandra Willyard