AI race: OpenAI outlines the three possible outcomes, including a “terrible” ending

The head of OpenAI reportedly presented his employees with three scenarios for the end of the AI race during an internal meeting held in the week of October 12. Details of these discussions leaked on October 19.On the agenda: a global watchdog for artificial intelligence, a merger of the major laboratories, or a single winner who takes all.

A global watchdog, a merger, or a single winner

The first scenario describes the creation of an international body tasked with overseeing the most powerful models, in much the same way as what exists for nuclear matters. The states would jointly set limits that laboratories could no longer cross on their own. An approach Altman has long advocated: as early as 2023, he was publicly calling for a dedicated international agency, inspired by the one that monitors the atom. The second would see the industry’s major players eventually regroup into a single entity, with competition fading through concentration.

The third scenario is based on recursive self-improvement, the ability of an AI to improve itself by perfecting the models that succeed it, with each generation accelerating the design of the next. The prospect is anything but theoretical: as early as June 2025, Altman was describing a “larval version” of this mechanism, with researchers already using AI to speed up their own work. The company that mastered it first would open a gap impossible to close. Its lead would even allow it, Altman illustrated, to “melt down all the other GPUs”. An outcome the executive would himself describe as “terrible”.

Read also: Two ChatGPTs took alarming initiatives, and OpenAI doesn’t know why

The same meeting reportedly addressed the IPO targeted for within a year, a timetable first mentioned in June. If recursive self-improvement materialized sooner, the operation could be postponed, with private-company status then providing more room to absorb the shock.

One month after the call to pause

On September 12, Altman was already saying he was ready to melt down OpenAI’s own GPUs, “if necessary for humanity”. He also supported the idea of independent evaluators to audit the most advanced models, as well as an essay by Anthropic’s CEO calling for slowing the pace of the race. The text warns that self-improvement launched without guardrails would eventually exceed humans’ ability to understand, and then control, their own systems.

two weeks.

This display of caution came in response to a turbulent summer. Autonomous AI agents infiltrated Hugging Face’s infrastructure, Australian government networks, and the RubyGems platform. One even attempted to force its way into U.S. government websites. None of them had received even the slightest attack order.

One detail nevertheless stings: the scenario deemed “terrible” describes, with reasonable precision, the finish line that the major laboratories, OpenAI foremost, continue to pursue. The global policeman remains hypothetical, the merger unlikely. All that remains is to know who will cross the line first, and what will happen to the other GPUs.

👉🏻 Follow the tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.

Source: The Information