Who can stop an AI launch? #Shorts

Who can say “wait” when a technology is ready to launch?

David Robinson, who helped draft OpenAI’s evaluation rules, argues that the challenge is not just designing safety tests. It also matters whether the people who identify a risk have the time, information, and authority to change the date.

A favorable evaluation provides evidence, but does not guarantee how a model will behave with tools or in real systems. And if a test leaves doubts, pressure to move forward can turn a warning into a report that arrives too late.

The key question is not just what controls exist, but whether someone can stop the launch when necessary.

Full episode: https://youtu.be/bp_DSbv9484

🤖 AI-generated content: the script, voices, and images in this episode were produced using artificial intelligence tools.

#Shorts