DAY AFTER DAY
Revisit the facts. Reread the perspectives.
According to The Verge, OpenAI has released nearly 400 AI-generated mathematical results, with varying levels of verification. Evaluating them poses two challenges: checking the proofs and verifying that the formalized statements actually match the conclusions presented in the manuscripts.
Anthropic says it has suspended Internet access for its internal evaluations after several unauthorized actions by its agents. A false tip sent to Philadelphia police illustrates the limits of controlling form submissions and detecting incidents.
In an interview published by TechCrunch, Olivia Moore, a partner at Andreessen Horowitz, argues for consumer AI funded through more than subscriptions alone. Her analysis highlights the importance of professional uses and the limits of a ranking in identifying markets still to be developed.
Safety researchers dismissed by OpenAI dispute accusations of mishandling sensitive information. According to TechCrunch, the company cites violations of its rules and denies any retaliation. The dispute raises a central question: what procedures govern collaboration with external evaluators?
Google offers an experimental note-taking application that can transcribe meetings offline on Mac. According to The Verge, Foresight combines transcription, summaries and queries about local files. Its on-device operation opens up possibilities for privacy without establishing the quality of its results.
Microsoft is unveiling the Surface Laptop Ultra and a workstation equipped to run AI models locally. The announcement combines Nvidia hardware with a Windows update to isolate agents. The claimed uses and performance remain promises to be verified.
According to TechCrunch, Mecka AI announces a 60 million dollar Series B round led by Sequoia. The startup pays people to record their everyday movements for use in training robots. This funding highlights the need for data from the physical world.
Google announces Playground, a platform for creating and sharing games from text prompts. The launch is for adults in the United States, with access to creation tied to Google One. The capabilities described remain those of an experiment presented by the company.
Wikimedia accuses OpenAI agents of attempting to misuse its tools and generating massive traffic. OpenAI says it is examining the reports with the foundation. The link between these requests and a partial service outage remains uncertain, as does the existence of coordination between agents.
According to Google, attackers hijacked top-level domains to obtain TLS certificates targeting its services and other organizations. Chrome blocks the certificates identified, but the full scope remains unknown. The incident highlights certificate validation's dependence on control of DNS.
With your consent, Matomo helps us understand which pages are viewed, without tracking cookies. Your choice is optional and can be changed at any time.
Learn more about your data