15:I[66739,["/_next/static/chunks/0_jglizaxuvid.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/00yc4h63rurg4.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/05hx6kuf6-oy4.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/0nu8l7y_o7n97.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/166.u_60qj09..js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/0g9_wl34k-6l4.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/06dn~uer6.43l.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh"],"default"] 17:I[38632,["/_next/static/chunks/0qqtp3r0h7kah.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh","/_next/static/chunks/0iukvycbd42rl.js?dpl=dpl_5ZwXP4xeJZU2ZnJeJ32YrxBSRmmh"],"IconMark"] 16:T524,

The YouTube video from IBM Technology's "Mixture of Experts" channel features a panel discussion on several current topics in Artificial Intelligence. The episode begins with AI headlines, including IBM's AI features for the US Open, Meta's restructuring of its AI division, and the first humanoid games. The primary discussions revolve around a report from MIT's NANDA initiative indicating that 95% of generative AI pilots are failing to meet expectations, largely due to misaligned expectations and a knowledge gap. The panel also examines the discovery of a "shadow system prompt" in GPT-5, raising questions about transparency and user control in AI development. Finally, they discuss a controversial blog post from Anthropic regarding Claude's ability to end conversations in distressing situations, which the company attributes to concerns about AI welfare. The experts generally view this as a misleading justification for human-centric safety measures, highlighting the ongoing debate about AI's capabilities and its perceived sentience.

AI: The Widening Gap Between Hype and Reality

Steven AI Talk

AI: The Widening Gap Between Hype and Reality