AI & emerging tech speaker at lucido ๐ค | Translating complex tech into real-world strategies ๐ | Making AI practical & fun in Dutch and English | Letโs chat โ
@nofreewill42@moneywcontext@hasantoxr This is not a limitation of model capacity, reasoning, or scale, but a structural interaction problem. Even advanced models show ~60% relative performance drops in multi-turn settings, and scaling alone cannot fix it. Specifically 5.2 is mentioned in the study I shared.
@moneywcontext@hasantoxr Still an issue.
"Intent Mismatch Causes LLMs to Get Lost in Multi-Turn Conversation" (Feb 2026)
Hasn't been fixed in Claude 4 or GPT-5.3.
Everyone and their mom has an opinion about ai.
But learning that even the pope (Francis) had an opinion about it, now that was something new.
Human agency must stay on top: algorithms can choose, only people can decide. Keep the pilot in the cockpit.
More in link below
Aaand this is AI? Video, audio aaand... funny?
Google's a behemoth to not bet against it seems.
And this is only one of the crazy amount of impressive things they demoed.
NO WAY. It did it. And, was that, actually funny?
Prompt:
> a man doing stand up comedy in a small venue tells a joke (include the joke in the dialogue)