Toggle light / dark theme

For the first time ever, researchers show how a large language model can help discover novel solutions to long-standing problems in math and computer science.


The card game Set has long inspired mathematicians to create interesting problems.

Now, a technique based on large language models (LLMs) is showing that artificial intelligence (AI) can help mathematicians to generate new solutions.

The AI system, called FunSearch, made progress on Set-inspired problems in combinatorics, a field of mathematics that studies how to count the possible arrangements of sets containing finitely many objects. But its inventors say that the method, described in Nature on 14 December1, could be applied to a variety of questions in maths and computer science.

Researchers at Shanghai Jiao Tong University School of Medicine, China, have discovered that shutting down part of the innate immune system increases anti-tumor activity.

In a paper, “Noncanonical MAVS signaling restrains dendritic cell–driven antitumor immunity by inhibiting IL-12,” published in Science Immunology, the team details how exploring the role of mitochondrial antiviral signaling in tumor immunity uncovered unexpected insights into the relationship with immune responses and potential therapeutic implications.

Mitochondrial antiviral-signaling (MAVS) proteins are part of the innate immune system encoded by the nuclear genome found mainly on the mitochondrial outer membrane. Considered a first line of defense against viral infections, they are rapidly produced upon viral recognition and quickly reduced when a virus is cleared from the system.

RELATED: OpenAI CEO Sam Altman celebrates ChatGPT’s first birthday

If the deal with German publisher Axel Springer pulls through, it’s highly likely that reports citing ChatGPT’s decline in accuracy will reduce significantly. OpenAI will also get access current and archived news articles across the publisher’s brands, which it will use to train its LLMs. The chatbots will also refer back to the original source to provide absolute transparency to users. And since its a mutual partnership, Axel Springer will also leverage OpenAI’s resources to further enhance and improve its products.

What are your thoughts on OpenAI’s first attempt to compensate publishers and websites for using their content to train its AI models? Let us know in the comments.

The valuation of Elon Musk’s SpaceX hit $180 billion based on an ongoing secondary share sale, CNBC confirmed Wednesday.

The company is discussing an agreement with investors to sell stock from insiders in a purchase offer at $97 a share, according to a person familiar with the discussions. The offer does not include raising new capital, as the purchase offer represents a secondary sale of existing shares and is expected to close in January.

SpaceX typically performs these secondary rounds about twice a year, to give employees and other company shareholders a chance to sell stock. The latest valuation represents a 20% increase from SpaceX’s previous high of $150 billion, which the company hit through a July secondary sale at $81 a share.

Hours prior to the secretive spacecraft’s launch, SpaceX stood down from the 7th planned liftoff of the U.S. Space Force’s own X-37B reusable space plane, and even removed the Falcon Heavy rocket containing it from the Kennedy Space Center launch pad. The mission, known as USSF-52, was scrubbed on Wednesday (Dec. 13) to “perform additional system checkouts.” Exact reasons for this delay remain unknown, and a new date has yet to be set for launch.

Related: China’s mysterious space plane returns to Earth after 9-month orbital mission

Much like the X-37B, little is known about China’s reusable space plane, subbed Shenlong, or “Divine Dragon.” From what bits of information are available to the public, though, the spacecraft appears to be used for testing new payloads and orbital operations. It launches vertically atop a rocket, conducts its mission and then lands horizontally on a runway similar to NASA’s space shuttle.