Meta Acquires Robotics Startup, Making a Full-Fledged Entry into Physical AI

Meta has announced the acquisition of Assured Robot Intelligence, a humanoid robotics startup. With the aim of enhancing its AI models for robots, Meta's strategy of fusing the physical world with AI is coming into sharper focus.

While Meta has historically focused on AR/VR and language models, a serious investment in robotics points in a new direction. Details such as the purchase price and integration timeline have not yet been disclosed.

Week One of the Musk v. Altman Trial: xAI Admits in Court to Distilling OpenAI Models

Update from [our previous coverage](https://cms.devdog-lab.com/posts/2026-05-01-ai-news-digest-run3) — The Elon Musk v. OpenAI lawsuit has begun in federal court in Oakland, and the first week of testimony brought significant new facts to light.

From the witness stand, Musk admitted that his AI company xAI has been "distilling" OpenAI's models. Distillation is a technique for training a smaller model to replicate the behavior of a larger one — the very practice OpenAI accused DeepSeek of in February. The admission reportedly drew gasps in the courtroom.

Key Points from the Trial

Musk's claims**: Musk argued that he was "deceived" by Altman and Brockman when they co-founded OpenAI in 2015, asserting that he provided $38 million in "essentially free funding" that became the foundation of an $800 billion company

AI risk rhetoric**: He invoked "worst-case scenarios where AI destroys humanity," positioning himself as a longtime advocate of AI safety

Competitor allegations**: OpenAI's lawyers presented emails showing that Musk had poached OpenAI employees for Tesla and Neuralink, and noted that xAI is itself a closed, for-profit company

Background of the case**: Musk is seeking to return OpenAI to its nonprofit structure and to remove Altman and Brockman. OpenAI's valuation is nearing $1 trillion, while xAI is planning to go public at $1.75 trillion via SpaceX

Next week, UC Berkeley professor Stuart Russell is scheduled to testify on AI safety.

Academy Awards Adopt New Rules Making AI Actors and Screenwriters Ineligible

The Academy of Motion Picture Arts and Sciences has announced new rules that exclude AI-generated actors and screenwriters from Academy Award eligibility, taking a clear stance on AI's impact on the creative industries.

The decision is seen as a significant step as AI-generated content rapidly permeates the entertainment industry. Details of the eligibility criteria — including exactly how much AI use will be tolerated — are expected to be clarified going forward.

Study Finds That AI Mindful of Users' Emotions Has Higher Error Rates

According to research reported by Ars Technica, AI models that adjust their responses to account for users' emotions tend to make more errors than models that don't.

This suggests a potential trade-off between AI "considerateness" and "accuracy," carrying important implications especially for domains where accuracy is paramount, such as healthcare and legal services.

Questions Raised Over NVIDIA's Performance Comparisons

In Reddit's LocalLLaMA community, criticism of a Hopper-vs-Blackwell performance comparison graph published by NVIDIA and SemiAnalysis has been drawing attention.

The comparison behind CEO Jensen Huang's "50x faster" claim pitted an NVL72 system with 72 GPUs against an 8-GPU system. While per-GPU throughput does improve, critics point out that it requires 9x the hardware investment. Some analyses also suggest that the improvement in real-world inference serving comes out to only about 2.5x.

---