… Musk: It is standard practice to use other AIs to validate your AI. OpenAI and xAI did not immediately respond to WIRED’s request for comment. OpenAI has been trying to prevent its competitors from distilling its AI models, in particular, the Chinese AI lab DeepSeek. …
… Code as Policy The idea that AI-powered coding could offer a powerful new way to build robots was first highlighted in a research paper from 2022 that dubbed the approach “code as policy.” Since then, AI’s coding skills have advanced at a dizzying pace, and the code-as-policy method has gained trac… …
… In January 2025, DeepSeek released a cutting-edge open model using a more efficient approach that made its training far cheaper. But a variety of other Chinese models from big companies like Alibaba, as well as startups like Moonshot AI, Z.ai, and MiniMax, have also become popular in the West. …
… A startup called Wafer is training AI models to do one of the most difficult and important jobs in AI—optimizing code so that it runs as efficiently as possible on a particular silicon chip. …
… But making custom silicon remains enormously expensive and technically complex, which means Meta will likely continue purchasing the majority of its AI hardware from other firms, at least in the near future. …
… Cursor has also started training in-house AI models that it can cost effectively serve to customers. The startup recently launched Composer 2, an AI model based on an open-source system from the Chinese AI lab Moonshot AI, that Cursor did additional pretraining and post-training on. …
… Mallory Knodel, a cryptography researcher at New York University, says it would be “great for people using chatbots that use Meta AI to have confidentiality and privacy within that exchange.” Crucially, that means Meta would not be able to access AI chat data for training, says Knodel, who along wi… …
… The dextrous, humanlike hand from Singaporean company Sharpa can do everything from card tricks to peeling an apple. Dexterity remains a key unsolved problem in robotics. …
… We trained it to be good at code, but as a side effect of being good at code, it's also good at cyber.” He adds in the video that “more powerful models are going to come from us and from others. …
… He noted that Anthropic trained its AI models not only on difficult coding problems from academic competitions but also on real-world problems from messy code repositories. “That was a lesson that we were delayed on,” Brockman said. …