The chapter explores the struggles faced by a former OpenAI researcher in the governance division, where safety concerns were overlooked in favor of AI progress prediction. It highlights a whistleblower incident regarding Microsoft bypassing the Deployment Safety Board (DSP) to deploy the GPT-4 model in India, questioning tech companies' accountability and self-regulation. Furthermore, it addresses the disappointment in compromising AI safety for compute power advancement, contradicting earlier stances on managing the transition to advanced AI capabilities.
This week, we host a cultural exchange. Kevin and Casey show off their Canadian paraphernalia to Prime Minister Justin Trudeau, and he shows off what he’s doing to position Canada as a leader in A.I. Then, the OpenAI whistle-blower Daniel Kokotajlo speaks in one of his first public interviews about why he risked almost $2 million in equity to warn of what he calls the reckless culture inside that company.
Guests:
- Justin Trudeau, Prime Minister of Canada
- Daniel Kokotajlo, a former researcher in OpenAI’s governance division
Additional Reading:
We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.