OpenAI failed to deliver on promise to allocate computing power for AI safety research

From Fortune: 2024-05-21 06:53:00

In July 2023, OpenAI promised 20% of its computing resources for a team dedicated to AI safety. Less than a year later, the team was disbanded amid staff resignations and accusations that OpenAI had not fulfilled its commitment, raising doubts about the company’s credibility. The team’s work had been seen as vital for the future of civilization, researching solutions to steer and control potentially superintelligent AI.

The disbanding of OpenAI’s Superalignment team comes after a turbulent period within the company, which saw staff resignations and allegations of inadequate support for critical research. The loss of key researchers and growing doubts about the company’s commitment to AI safety have raised questions about OpenAI’s credibility and its ability to responsibly handle the development of future, more powerful AI models.

OpenAI faced controversy over the use of a voice for its AI speech generation technology that bore striking similarities to actress Scarlett Johansson’s voice. Johansson claims she declined permission for her voice to be used but OpenAI denies using her voice without consent. Questions and speculation around the creation of the AI voice, named “Sky,” have further eroded trust in OpenAI’s public statements and actions.

The Superalignment team, led by former OpenAI chief scientist Ilya Sutskever, had been tasked with researching solutions to control potentially superintelligent AI systems. However, the team faced challenges in accessing the computing resources promised by OpenAI, hindering their work. Concerns over the company’s prioritization of product launches over AI safety have led to staff resignations and a loss of trust in the company’s ability to handle advanced AI development responsibly.



Read more at Fortune: OpenAI promised 20% of its computing power to combat the most dangerous kind of AI—but never delivered