Should frontier AI models be allowed to skip public benchmarks?
xAI’s Grok 4.5 is being tested exclusively inside SpaceX and Tesla with no independent benchmarks or public access. If the most powerful AI systems are developed and deployed behind closed doors, who holds their builders accountable — and does the public even have a right to know what these models can do?
Commentaires (1)
Today’s episode of Minds, Bodies, and Terawatts (June 30) digs into xAI’s decision to keep its 1.5-trillion-parameter Grok 4.5 entirely behind closed doors, testing it on rocket trajectories and factory floors while burning $2.5 billion per quarter. The episode explores how this fits the ‘Star Wars path’ from the Unscarcity framework — where frontier AI concentrates in the hands of those who can afford the compute, with no public oversight of capabilities. It’s a fascinating tension: proprietary testing generates unmatched training data, but it also means society is flying blind on what these systems can actually do. Give the episode a listen and let us know — should we demand transparency from frontier labs, or is private deployment just how the market works?
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
