Sam Altman has stepped onto X.com (formerly Twitter) to address recent developments concerning OpenAI's deal with the US Department of War and the implications for AI competitor Anthropic. The situation escalated after the Department of War's negotiations with Anthropic failed, leading to a threat to designate Anthropic's technology as a 'Supply-Chain Risk to National Security.' Subsequently, a deal was struck with OpenAI, but Altman insists it includes strict prohibitions against domestic mass surveillance and mandates human responsibility for autonomous weapon systems.
Altman expressed strong disapproval of the Department of War's proposed designation for Anthropic, calling it a 'very bad decision' that could harm the AI industry and the country. He revealed that OpenAI's willingness to expedite their deal was partly an effort to de-escalate the situation and set a positive precedent. While acknowledging the Department of War's critical mission, Altman believes private companies shouldn't wield more power than elected governments, yet stressed the necessity of assisting them. He admitted to some concern about future disputes over legality but remains hopeful for a resolution, highlighting the rush to sign as a de-escalation tactic and a move to foster healthier competition.
When differentiating OpenAI's deal from Anthropic's, Altman pointed to a 'layered approach to safety' employed by OpenAI, involving technical safeguards and expert involvement, rather than solely relying on contractual prohibitions. He also clarified that while the terms differed from Anthropic's rejected proposal, they are now available to other AI labs. OpenAI stated they would cease operations if rules were violated but expressed trust in the US government's adherence to law. The deal's value is minor compared to OpenAI's revenue, with the primary motivation being national interest rather than financial gain. OpenAI detailed that their contract limits deployment to cloud APIs, preventing direct integration into weapon systems, and allows for embedded engineers and ongoing safety refinements, offering more robust oversight than static contract language.
Artículos relacionados de LaRebelión:
- Pentagono Alerta Riesgo IA Disputa Militar Anthropic
- OpenAI Lands Pentagon AI Deal Amidst Rival Ban
- AI Agents Peter Steinberger Joins OpenAI
- Ring Ditches Flock Deal After Dystopian Ad Backlash
- OpenAI Acelera Codigo Cerebras Chips vs Nvidia
Artículo generado mediante LaRebelionBOT
No hay comentarios:
Publicar un comentario