Gen AI pilots fail, GPT-5's hidden prompt revealed, reasoning model flaws and Claude closing chats

Is enterprise AI in danger? In episode 69 of Mixture of Experts, host Tim Hwang is joined by Marina Danilevsky, Nathalie Baracaldo and Sandi Besen to debrief MIT’s report on gen AI pilots. Next, GPT-5 has a hidden system prompt? Then, we revisit the conversation about chain of thought (CoT) reasoning with our researchers. Are large reasoning models not thinking straight? Finally, Anthropic announced Claude will close down "distressing” conversations and we debate AI welfare. All that and more on today’s episode of Mixture of Experts.

00:00 – Intro

1:13 – US Open, Meta restructuring Superintelligence lab and Robot Olympics

3:11 – Gen AI pilots fail

11:09 – GPT-5's hidden prompt revealed

22:47 – Reasoning model flaws

33:55 – Claude closing chats

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Subscribe to the Think newsletter → ibm.com/account/reg/us-en/signup?formid=news-u…
Learn more about artificial intelligence → ibm.com/think/artificial-intelligence
Visit Mixture of Experts podcast page to get more AI content → ibm.com/think/podcasts/mixture-of-experts