Listen To The Show
Transcript
Welcome to The Prompt by Kuro House, your daily AI update. Today, we’re diving into the latest twists in the Elon Musk and OpenAI saga, plus some surprising moves in the AI infrastructure world. Let’s get started.
First up, Mira Murati, former OpenAI CTO, just dropped a bombshell in court. According to a report from The Verge, Murati testified under oath that CEO Sam Altman lied to her about safety protocols for a new AI model. She said Altman falsely claimed the legal team had cleared the model from review by the deployment safety board, which wasn’t true. Murati described a toxic management style, accusing Altman of undermining her work and pitting executives against each other. She even confirmed a misalignment between what Altman and OpenAI’s general counsel were saying about the model’s safety. After Altman’s firing, Murati was briefly interim CEO but warned the company was at catastrophic risk before she left to start her own rival lab.
Next, Shivon Zilis, a key Musk loyalist, testified in the Musk versus Altman trial, and her story is wild. The Verge detailed how Zilis, who has four children with Musk, worked across Tesla, Neuralink, and OpenAI starting in 2017. She was the only person taking detailed notes during critical OpenAI meetings that shaped the company’s future and allegedly facilitated communication between Musk and Altman. Despite denying funneling info to Musk, her emails and texts suggest otherwise, revealing Musk’s attempts to control OpenAI by recruiting board members and even poaching top talent. Zilis resigned from OpenAI’s board when she learned about Musk’s competing AI venture, xAI, but texts showed she knew about it earlier. Her primary allegiance appears to be to Musk, raising questions about her credibility on the stand.
Speaking of Musk’s AI ambitions, Wired revealed a previously unknown recruitment attempt. Before leaving OpenAI’s board in 2018, Musk tried to lure Sam Altman to lead a “world-class AI lab” inside Tesla, even offering him a Tesla board seat. The plan was to rival Google DeepMind and Facebook AI Research, with Musk and other Tesla execs at the helm. Emails from Shivon Zilis showed Musk’s desire to absorb OpenAI into Tesla or spin off parts of it to maintain control. The AI lab never launched, and Altman stayed with OpenAI, but the court is now seeing this as evidence of Musk’s efforts to subvert OpenAI’s original mission.
Now, on the infrastructure front, Anthropic and Musk’s SpaceXAI just struck a major deal. According to Wired, Anthropic will use compute power from SpaceXAI’s Colossus 1 supercomputer, gaining access to roughly 300 megawatts and about 220,000 Nvidia GPUs. This partnership is huge, helping Anthropic scale its Claude AI services amid growing demand and rate limits. It also boosts SpaceXAI’s IPO prospects by showing it can monetize its massive data centers, including plans for orbital AI compute capacity. Interestingly, Musk recently praised Anthropic’s safety efforts, a sharp turnaround from his earlier criticisms.
Finally, TechCrunch analyzed what this Anthropic and xAI partnership means for Musk’s strategy. They argue xAI is transforming into a neocloud business, focusing on selling compute resources rather than primarily developing AI models. Musk’s data centers, including future space-based ones, position xAI as a GPU rental powerhouse, competing with giants like Google and Meta. This approach is risky and capital intensive, but it could pay off if Musk’s orbital data center vision succeeds. However, selling compute to competitors might limit xAI’s own software ambitions in the near term.
That’s a wrap for today’s AI update. From courtroom drama to billion-dollar infrastructure deals, the AI landscape is as complex and fascinating as ever. We’ll keep watching how these stories unfold and what they mean for the future of AI. Thanks for tuning in to The Prompt by Kuro House. Catch you tomorrow.
