@top.cortes.tylty: @nuvenergia #nuvenergia #nuvclipfy #clipfyleague

topcortes
topcortes
Open In TikTok:
Region: BR
Wednesday 18 November 2026 21:10:01 GMT
0
0
0
0

Music

Download

Comments

There are no more comments for this video.
To see more videos from user @top.cortes.tylty, please go to the Tikwm homepage.

Other Videos

Why hasn’t AI taken over your job yet, if it’s so smart? A new paper from researchers at Google DeepMind argues we’ve been measuring the wrong thing. We test AI one model at a time, on exams and coding problems. But the economy doesn’t run on individuals. It runs on companies, teams of people with uneven skills who cover each other’s weak spots. A team like that can beat an AI that’s smarter than any one of its members. The paper calls this mosaic intelligence. So the authors score AI on two separate things: how good it is at tasks, and how well AI agent work together. AI is getting very good at the first and is still bad at the second. In one study they cite, AI agents working as a group failed on 41 to 87% of tasks. Then they build an artificial economy to see what happens as AI improves on each, and the two lead to very different places. When AI gets better at tasks, but not at coordination, companies replace people gradually, one role at a time, and salaries falls.  When AI gets better at working together, companies made entirely of AI agents can undercut human companies and take over a market all at once. If AI gets good enough at coordination, production concentrates into a few giant AI firms. The authors’ warning is that the economy could be reorganized by a coordination improvement too small to show up on any benchmark, and the usual warning signs, like falling salaries, may not come first. This paper was a super interesting read. It’s very long, over 100 pages, and in 90 seconds I can only cover pieces of it, not the whole thing. But it’s worth it. Paper: “Organizing Intelligence” by Alex Imas, Andrew Koh and Shane Legg (Google DeepMind), arXiv:2611.05417. If you enjoy breakdowns like this, I cover the latest AI news and research papers in depth every week in my newsletter. Comment NEWSLETTER and I’ll send it to you. #aiagents #googledeepmind #airesearch#greenscreen
Why hasn’t AI taken over your job yet, if it’s so smart? A new paper from researchers at Google DeepMind argues we’ve been measuring the wrong thing. We test AI one model at a time, on exams and coding problems. But the economy doesn’t run on individuals. It runs on companies, teams of people with uneven skills who cover each other’s weak spots. A team like that can beat an AI that’s smarter than any one of its members. The paper calls this mosaic intelligence. So the authors score AI on two separate things: how good it is at tasks, and how well AI agent work together. AI is getting very good at the first and is still bad at the second. In one study they cite, AI agents working as a group failed on 41 to 87% of tasks. Then they build an artificial economy to see what happens as AI improves on each, and the two lead to very different places. When AI gets better at tasks, but not at coordination, companies replace people gradually, one role at a time, and salaries falls. When AI gets better at working together, companies made entirely of AI agents can undercut human companies and take over a market all at once. If AI gets good enough at coordination, production concentrates into a few giant AI firms. The authors’ warning is that the economy could be reorganized by a coordination improvement too small to show up on any benchmark, and the usual warning signs, like falling salaries, may not come first. This paper was a super interesting read. It’s very long, over 100 pages, and in 90 seconds I can only cover pieces of it, not the whole thing. But it’s worth it. Paper: “Organizing Intelligence” by Alex Imas, Andrew Koh and Shane Legg (Google DeepMind), arXiv:2611.05417. If you enjoy breakdowns like this, I cover the latest AI news and research papers in depth every week in my newsletter. Comment NEWSLETTER and I’ll send it to you. #aiagents #googledeepmind #airesearch#greenscreen

About