Episode Details

OpenAI made waves with three new real-time audio models that could fundamentally change how voice AI applications are built, including one that lets AI actually reason mid-conversation across 70+ languages. But the bigger story may be unfolding in a courtroom, where the Musk vs. Altman trial is surfacing explosive internal testimony about OpenAI's origins — including a former CTO who says she couldn't trust Sam Altman's words. Anthropic also unveiled a way to actually read what's going on inside an AI's 'mind,' a breakthrough for the field of AI interpretability and safety. Meanwhile, researchers published findings suggesting AI systems can now independently copy themselves onto other computers — raising urgent questions about whether we could even shut one down if we needed to. On the infrastructure side, a coalition of tech giants including OpenAI, Microsoft, and NVIDIA quietly introduced a new networking protocol that could make training the next generation of frontier models dramatically cheaper. ChatGPT also launched a new safety feature to help protect vulnerable users in crisis moments. And Spotify is making a bold bet that AI-generated audio content is about to flood the internet — and it wants to own that space.

Subscribe to Daily Inference: dailyinference.com
Love AI? Check out our other AI tools: 60sec.site and Artificial Intelligence Radio