Diplomacy Gameplay with CICERO | AI at Meta
Вставка
- Опубліковано 18 жов 2024
- Learn more about CICERO gameplay: bit.ly/3gy2oiS
When Meta’s CICERO is one of your opponents, you might want to take notes. Especially when 3x Diplomacy World Champ Andrew Goff and Meta AI Research Engineer Adam Lerer discuss their experience playing with the AI agent. Such as when it coordinates and prompts moves, uses honesty as a tactic or uses its clever strategic planning abilities.
Check out our website: bit.ly/3ETSWji
Check out our blog: bit.ly/3UX2nDT
Read our research paper: bit.ly/3EmCwhM
#ArtificialIntelligence
--
Subscribe: www.youtube.co...
Learn more about our work: ai.facebook.com
Follow us on Twitter: / aiatmeta
Follow us on Facebook: / aiatmeta
Connect with us on LinkedIn: / aiatmeta
Working on AI at Meta, we're bringing the world together by advancing AI, powering meaningful and safe experiences, and conducting open research.
This is a level of skill I wouldn't have expected from an AI within the next 2 or 3 years. Despite the tenor of the other comments here (at the time being), Cicero gains quite a lot of attention in media and science. Speaking of "defining moments" happens far too often in tech journalism, but the step Cicero has taken really seems to be such a moment.
7:54 Top Diplomacy player giving pitch perfect compliment to the team regarding the application of these systems outside games is the best part of this video to me.
Great to see the analysis here , an amazing accomplishment. 2nd-level effects are going to be very interesting over the next few years!
Interesting. My memory of playing Diplomacy was that it was like 99% beneficial to be honest and gregarious, but that reserving your treachery for endgame and win conditions could be super powerful. One has to wonder if 100% honesty is actually a drawback.
I mean it did say it was a top 10% player on the site, not top 1%. Its strong but not world champ strong yet.
It is honest, but it lies by what it doesn't include in its messages. That's an important thing to note.
@@Baset_
Yeah but if you removing lying from your toolbox entirely you will miss wins and better results you could have had. It's not good to lie all the time but there are moments where it is worth it.
Ie: if you're doing a sea lion it's probably bet to lie and ask for a DMZ of english channel as france rather than being honest as france because getting an army onto the english mainland in 01 is worth the price of angering a player who will be out of the game soon anyways.
Great video. Hope to see more llama related videos from Meta
Incredible work!
im not surprised but definitely impressed, great job Meta AI team. but can we replace (Hey Facebook) on quest with an embodied ai avatar that actually works pls 🙏😁
I wonder if they are going to create a general demo system like ChatGPT.
Waiting,Let it speak strategy of alpha zero.
This is interesting!
140 views in 4 hours, shows how people less interested people are in Meta lol
that's a good thing, more shares to buy at a discount
Imagine applying something like this to a video game that’s similar like civilization, would be the best video game of all time, if it was only for quest id buy a quest
Hi I'm Ariful Islam leeton im software engineer and members of the international organization and development open A. I And A. I mata
An AI that lies and externalizes harm. Just goes to show why we need ethical standards and regulation rather than leaving AI development to billionaire psychos
But, Cierto is more honest that human players?
Like, honest to a fault even. Diplodocus is a gunboat (no press) diplomacy bot which is very good at deception but it can't actually lie to you because it plays without talking.
Nope