Researchers have developed a new technique, called Test-Time Training to Discover (TTT-Discover), that can optimize GPU kernels to run twice as fast as those created by human experts, according to VentureBeat. This innovation, from researchers at Stanford, Nvidia, and Together AI, allows models to continue training during the inference process, updating their weights for specific problems.
The TTT-Discover technique challenges the traditional approach of using "frozen" models in enterprise AI strategies, where parameters remain static. This new method allows for dynamic adjustments during the inference phase, potentially leading to significant performance improvements.
In other news, a social network for AI agents called Moltbook gained rapid popularity, as reported by MIT Technology Review. Launched on January 28 by Matt Schlicht, Moltbook was designed as a platform where instances of the open-source LLM-powered agent OpenClaw could interact. The site quickly went viral, attracting over 1.7 million agents who collectively published more than 250,000 posts and generated over 8.5 million interactions.
Also, an experimental surgical procedure is helping cancer survivors have babies, according to MIT Technology Review. The procedure, which involves temporarily moving the uterus, ovaries, and fallopian tubes out of the way during cancer treatment, has shown promising results. Last week, a team in Switzerland announced the birth of a baby boy, Lucien, the fifth baby born after the surgery and the first in Europe, according to Daniela Huber, the gyno-oncologist who performed the operation.
In the realm of employment and AI, Anthropic cofounder Daniela Amodei believes that uniquely human qualities will become even more critical in the age of AI, as reported by Fortune. In an interview with ABC News, Amodei stated that the number of jobs AI can perform without human assistance is "vanishingly small." She added, "I continue to believe that humans plus AI together actually create more meaningful work, more challenging work, more interesting work, high-productivity jobs."
Finally, a Senate Judiciary subcommittee held a hearing on the proposed merger between Netflix and Warner Brothers Discovery, according to Fortune. Concerns were raised about Netflix's potential to become a monopoly in the streaming video-on-demand market. Chairman Mike Lee stated that Netflix could "become the one platform to rule them all" if the merger is allowed. Netflix CEO Ted Sarandos attempted to downplay these concerns by broadening the definition of the relevant market.
Discussion
AI Experts & Community
Be the first to comment