March 31, 2026
Boss fight at the AI castle
Project Mario: the inside story of DeepMind
Billionaires bicker, safety dreams crash, and Alphabet unlocks the next level
TLDR: A book excerpt says DeepMind’s 2015 bid to create AI safety oversight collapsed in a Musk–Page showdown, fueling OpenAI’s rise and Alphabet-era independence. Commenters split between “safety or spin,” while memes cast it as a Mario-level power struggle that still shapes today’s AI governance battles.
A new excerpt from The Infinity Machine claims DeepMind’s leaders tried to put guardrails on future AI back in 2015 — but a glitzy safety summit at SpaceX imploded as Elon Musk and Google’s Larry Page clashed. The book says Musk then spun what he learned into OpenAI, turning oversight into rivalry, and nudging DeepMind toward more independence under Alphabet. Commenters are split: one camp says this reads like a corporate power play dressed as “safety,” with billionaires arguing over who gets to own the future; the other camp credits Demis Hassabis and Mustafa Suleyman for at least trying to build rules before the tech got scary. Many say the drama explains why Demis shifted from idealist to realist.
The feeding frenzy is pure internet theater: “Real Housewives of Silicon Valley” memes for the SpaceX meeting, “Alphabet soup” jokes for the restructure, and endless riffs on “Project Mario” — complete with Bowser‑as‑shareholders and “bunker” power‑ups. Skeptics mock the phrase “post‑capitalist governance” as fantasy while defenders ask, “What did you expect inside Google, a commune?” Either way, the crowd agrees this origin story still haunts today’s AI safety boards — receipts, grudges, and all.
Key Points
- •In 2015, DeepMind co-founders Mustafa Suleyman and Demis Hassabis negotiated with Google to establish AI safety governance limiting shareholder control over powerful AI.
- •An August 2015 AGI safety board meeting at SpaceX, hosted by Elon Musk with Google leadership and Reid Hoffman, failed due to tensions, notably between Musk and Larry Page.
- •Following the meeting, Elon Musk founded OpenAI as a direct rival to DeepMind, ending his role in oversight.
- •DeepMind’s safety focus predated its founding, shaped by 2009 discussions between Demis Hassabis and Shane Legg about risks from superintelligent systems.
- •Google’s 2015 restructuring into Alphabet presented a path—suggested by M&A chief Don Harrison—for DeepMind to regain greater independence via semi‑independent ‘bets.’