Sam Altman: OpenAI to Slow AI Development Without Waiting for Congress
Sam Altman, OpenAI’s CEO, has announced that the company will start pacing AI development, not waiting for Congress to act. In a post on X, he emphasized the need for a federal framework ensuring AI safety but noted that they will begin this work regardless of antitrust exemptions or legislation.
OpenAI’s Approach to AI Safety
Altman highlighted OpenAI’s shift in focusing on safety cases before significant training runs, a change from the traditional practice of releasing finished models. This includes:
- Explicit Safety Cases: OpenAI now creates detailed safety plans before training powerful AI models.
- Industry Standards: They advocate for shared standards for misalignment, monitoring, and safety.
- Independent Auditors: Altman expressed interest in ideas like independent auditors to enhance safety.
Avoiding Risks
He emphasized two primary risks to be mitigated:
- Loss of Control: Altman believes it’s unacceptable for humanity to lose control of the future to AI.
- Power Concentration: He warns against excessive power concentration in the hands of one person or entity, which could lead to a dystopian outcome.
Recent Developments
This stance follows OpenAI’s calls for mandatory national AI safety rules in the US and their support for binding AI regulations in the UK.
In an interview with Fortune, Altman further emphasized the potential risks and OpenAI’s commitment to safety, stating they would not go public this year.