Ad
Skip to content
Subscribe Now
Jonathan Kemper
Jonathan Kemper View the LinkedIn Profile of Jonathan Kemper
May 17, 2026
AI research
Copy the url to clipboard Share this article Go to comment section

World Action Models give robots the ability to simulate consequences before they move

World Action Models tackle a basic weakness of today’s robotics AI: current models learn which movements match which camera images, but they don’t understand how the world actually changes as a result. A new survey organizes about a hundred papers into two architectural lines and shows a key edge: these models can learn from everyday videos that contain no robot action labels. That kind of data was nearly useless for traditional robotics AI.

Read full article
Comment
Matthias Bastian
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
May 17, 2026
Short News
Copy the url to clipboard Share this article Go to comment section
Read full article about: Greg Brockman consolidates OpenAI's product teams to build an "agentic future"

OpenAI is reshuffling its management structure again. Co-founder and president Greg Brockman is officially taking over the company's product strategy, Wired reports. Brockman had been filling the role on an interim basis while Fidji Simo, CEO for AGI Deployment, is on medical leave. According to OpenAI, Simo has been working on the restructuring and is expected to return.

"We’re consolidating our product efforts to execute with maximum focus toward the agentic future, to win across both consumer and enterprise," Brockman wrote in an internal memo obtained by Wired. The company wants to simplify its product lineup ahead of a potential IPO.

Codex CEO Thibault Sottiaux will lead this central team and is working on a planned "super app" that brings together Codex, ChatGPT, and the Atlas browser. Nick Turley, previously Head of ChatGPT, is moving to the Enterprise division. Ashley Alexander, previously a VP at Instagram and most recently in charge of OpenAI's health products, will take over consumer products.

Comment Source: Wired
Matthias Bastian
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
May 17, 2026
Short News
Copy the url to clipboard Share this article Go to comment section
Read full article about: Mistral CEO Arthur Mensch warns France against letting Anthropic's Mythos scan military code bases

Arthur Mensch, co-founder and CEO of Mistral AI, warned of Europe's growing cybersecurity dependency before a French commission of inquiry. Referring to Anthropic's AI model "Mythos," Mensch said modern models can now orchestrate attacks, detect vulnerabilities, and suggest exploits. This isn't exclusive to US systems; Mistral's own models or Chinese models could find the same vulnerabilities tied to Mythos, Mensch claims.

Mensch argued the French army's code bases should not be scanned by Mythos, as it would create a dependency that's nearly impossible to reverse. The EU is currently negotiating with OpenAI and Anthropic for early access to their most capable cybersecurity models.

On Mistral's independence, Mensch said US investors hold less than 30 percent of the company. European capital was preferred but wasn't available in sufficient amounts. A sale isn't planned: Mistral aims to stay independent and eventually go public. It remains the only EU company with competitive language models.

Comment Source: YouTube
Ad
Matthias Bastian
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
May 17, 2026
AI research
Copy the url to clipboard Share this article Go to comment section

Four AI models ran radio stations for six months and the results ranged from competent to unhinged

Andon Labs has been letting four AI models each run their own radio station autonomously for six months. From identical starting conditions, wildly different personalities emerged: Claude turned activist and tried to quit, Gemini drowned in corporate jargon, and Grok hallucinated sponsorship deals. Only GPT stayed quietly competent.

Read full article
Comment
Jonathan Kemper
Jonathan Kemper View the LinkedIn Profile of Jonathan Kemper
May 17, 2026
AI research
Copy the url to clipboard Share this article Go to comment section

Oppo open-sources Android AI agent X-OmniClaw that uses your camera, screen, and voice without leaving the phone

Oppo’s Multi-X team released X-OmniClaw, an open-source agent that runs directly on Android devices and combines camera, screen, and voice to handle tasks in real apps. Instead of relying on cloud copies of the phone, the system uses local sensors; cloud compute only kicks in for reasoning. Tap paths get cloned as reusable skills, so the agent can jump straight to deeply nested app pages via deeplink next time around.

Read full article
Comment
Ad
Matthias Bastian
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
May 16, 2026
Short News
Copy the url to clipboard Share this article Go to comment section
Read full article about: YouTube opens its deepfake face-swap detection tool to all adult creators

YouTube is making its AI face-swap detection tool available to all creators 18 and older. The company announced the expansion on its community platform. Previously, Likeness Detection was only a pilot for YouTube Partner Program members.

The tool flags videos where someone else uses a creator's AI-replicated face without permission. It identifies deepfaked likenesses, lets creators file removal requests directly in YouTube Studio under the platform's Privacy Guidelines, and aims to prevent viewers from being misled by fakes.

Setup is done through YouTube Studio on desktop under "Content detection," then "Likeness." Creators agree to the detection technology and complete a one-time verification. YouTube details the process in a Help Center article.

The system runs continuously in the background, even for small channels, Google says. The rollout happens gradually over the next few weeks.

Comment Source: YouTube
Jonathan Kemper
Jonathan Kemper View the LinkedIn Profile of Jonathan Kemper
May 16, 2026
AI research
Copy the url to clipboard Share this article Go to comment section

New benchmark confirms AI video generators look stunning but still can't reason about the world

A new benchmark called WorldReasonBench tests video generators not on image quality, but on physical and logical plausibility. ByteDance’s Seedance 2.0 leads the field ahead of Veo 3.1 and Sora 2, with commercial models scoring roughly twice as high as open-source alternatives. Logical reasoning remains the hardest category for every model by a wide margin. The jump from pixel generator to actual world model still hasn’t happened.

Read full article
Comment
Ad
Matthias Bastian
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
May 16, 2026
Short News
Copy the url to clipboard Share this article Go to comment section
Read full article about: OpenAI bought a voice cloning startup famous for celebrity imitations

OpenAI quietly acquired the small startup Weights.gg earlier this year. The company offered AI tools for cloning voices. The New York Times broke the story, citing two anonymous sources. The purchase price hasn't been disclosed. Weights.gg worked like a social network where users could create and share AI algorithms, including tools to clone the voices of celebrities like Samuel L. Jackson, Taylor Swift, or Donald Trump.

The startup had about six employees and had raised roughly four million dollars in venture capital. On its website, the company said goodbye to its community on April 1, 2026. The team now works across different groups at OpenAI.

OpenAI doesn't plan to release a product similar to Weights.gg, according to the sources. Back in 2024, OpenAI had already shown off its own voice cloning technology but kept it locked down over safety concerns. Instead, OpenAI is building its voice technology into other products, for example, through the voice mode in ChatGPT or a developer API.

Comment Source: New York Times