Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

CICERO: An AI agent that negotiates, persuades, and cooperates with people

ai.facebook.com

Post reply on HN
Print view View original

241–250 of 300 posts

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#241

Post by theptip »

toxicFork wrote: post #218
EGreg wrote: post #49

Earlier quoted context omitted.

Just wait until we have AI swarms. So far all this has been with one player, amid others, no collusion. You’re going to be surrounded very shortly by sleeper bots, including on HN. Relying on dang and others to root out bots will be futile. A swarm could easily collude to downvote people or get them ostrasized by their own friend group, as we have already seen when it came to crypto, metoo, BLM, lockdowns, vaccines a…

Obviously you need good ai swarms to combat the bad ones

And here we have the early side-plot describing the end of the internet in the book Fall by Neal Stephenson.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#242

Post by theptip »

altruios wrote: post #55
tuzongyu wrote: post #37

Earlier quoted context omitted.

While many players do lie in the game, the best players do so very infrequently because it destroys the trust they’ve built with other players - see e.g. interview with world champion here: https://www.popularmechanics.com/culture/gaming/a34043608/wi... The agent generates plans for itself as well as for other players that could benefit them / that they are likely to do, and it tries to have discussions based on thos…

>> It doesn’t always follow through with what it previously discussed with a player because it may change its mind about what moves to make, but it does not intentionally lie in an effort to mislead opponents. The problem with this statement is it assigns intention to a AI model. It does not 'intend' to lie... but still may effectively do so. Lying may be the wrong word (as it presumes intent)... it's hard to express…

But you can tell what the plan is at a given turn in a model like this, because you can extract its representation of the plan directly.

By analogy think of how Stockfish can evaluate multiple positions; in this case it’s coming up with a plan then serializing that plan to the language model. There is no room for deception between the AI and a researcher that is probing the model activations directly.

(This sort of AI->researcher deception is the scary scenario for AI risk researchers though, it comes into play when the models are so smart/complex that you can’t extract their internal representation directly. See the Eliciting Latent Knowledge paper for a deep dive https://www.lesswrong.com/tag/eliciting-latent-knowledge-elk)

I think “intention” is meaningful here, understood in the lay sense of “has formed a plan of action”. It’s a simple and bounded plan, but a plan nonetheless.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#243

Post by smaudet »

stavros wrote: post #80
mtlmtlmtlmtl wrote: post #76

Earlier quoted context omitted.

Sure, people do that. But people who only do that are sociopaths.

I'm sure the AI hangs out at the pub with its friends after hours though.

If only, like a sociopath it hangs out with no one...

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#244

Post by smaudet »

EGreg wrote: post #162
worldsayshi wrote: post #105

Earlier quoted context omitted.

There are ways to ensure that people are people on the internet though. Not good ways but there are ways. You could for example limit users to those that log in with electronic id's issued by a government or other organisation that you trust to assert that the user is human and then force real names or a single user name for that e-id.

But when bots do a lot of things better than humans, including comment, answer questions, make jokes, say romantic things, etc. many would start to prefer them to humans in their online communities, and give them even more social capital. This already happened in other areas of life. Both fathers and mothers now neglect their own children and elderly parents so they can work for corporations. They often prefer this a…

Exactly, it's not for us at all.

This tech is not a quality of life enhancer at all, it's our competition as a species...but then again if we are really this stupid maybe we (well, you lot) deserve to die at your robot overlords hands.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#245

Post by mach1ne »

falcor84 wrote: post #45
mabbo wrote: post #28

I would love to see this kind of thing applied to an RPG. Randomly generate a city full of people. Make a few dozen of them the important NPCs. Give them situations and goals, problems they need to solve and potential ways to solve them. Certain NPC's goals are opposite others'. Then drop the player into that world and have the 'quests' the player is performing be generated based on the NPCs needing their help. Updat…

This sounds very close to AGI capabilities. Maybe I've just been thinking too much about Westworld, but I'm starting to get very concerned that people can be gradually replaced by AI agents with their own agendas (focus on this last part).

At some point they probably will be, but it seems unlikely to happen within the next 7 years at least. Deep Learning still has the same deep-seated issues it has had since its inception.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#246

Post by morgante »

gcanyon wrote: post #97
bee_rider wrote: post #17

Earlier quoted context omitted.

> Cicero participated anonymously in 40 games of Diplomacy in a “blitz” league on webDiplomacy.net > Cicero ranked in the top 10% of participants who played more than one game and 2 nd out of 19 participants in the league that played 5 or more games. > As part of the league, Cicero participated in an 8-game tournament involving 21 participants, 6 of whom played at least 5 games. Participants could play a maximum of 6…

There's an interesting question of "how much do the literal best humans suck at this?" For example, in chess Magnus Carlsen might be able to beat Stockfish given a handicap of just a pawn or two. An even better computer player than Stockfish might give up three or more pawns, but even a perfect player would likely lose to Carlsen if giving up a rook. -- I'm making this up, I don't think anyone knows the real values,…

> All to say, it seems likely that Diplomacy is a game where the difference between "the best human play" and "the best possible play" is much larger than either Go or Chess.

Definitely, Diplomacy in general is substantially understudied compared to Go or Chess (largely because it's a tiny community). You can play for less than a year and get to top-level performance and much of the established wisdom/strategy of players is fairly bad.

Even the best Diplomacy players are only scratching the performance of how good someone could be.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#247

Post by ImHereToVote »

mach1ne wrote: post #245
falcor84 wrote: post #45

Earlier quoted context omitted.

This sounds very close to AGI capabilities. Maybe I've just been thinking too much about Westworld, but I'm starting to get very concerned that people can be gradually replaced by AI agents with their own agendas (focus on this last part).

At some point they probably will be, but it seems unlikely to happen within the next 7 years at least. Deep Learning still has the same deep-seated issues it has had since its inception.

https://intelligence.org/2017/10/13/fire-alarm/

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#248

Post by charcircuit »

nopinsight wrote: post #235
pesenti wrote: post #20

Paper: https://www.science.org/doi/10.1126/science.ade9097 Code: https://github.com/facebookresearch/diplomacy_cicero Site: https://ai.facebook.com/research/cicero/ Expert player vs. Cicero AI: https://www.youtube.com/watch?v=u5192bvUS7k RFP: https://ai.facebook.com/research/request-for-proposal/toward... The most interesting anecdote I heard from the team: "during the tournament dozens of human players never even su…

"Having read the paper & supplementary materials, watched narrated game & spoken to one of the human players I'm pretty concerned. The @ScienceMagazine paper centres 'human-AI cooperation' & the bot is not supposed to lie. However, videos clearly show deception/manipulation" "Screenshots of the stab below. The human player said: "The bot is supposed to never lie [...] I doubt this was the case here" "I was definitely…

I don't see anything in the papers that say the bot isn't supposed to lie. Lying and being deceptive is a part of the game.

Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people

#249

Post by EGreg »

mycall wrote: post #228
EGreg wrote: post #166

Earlier quoted context omitted.

Pretty much. Able to write vastly more content, that on first blush makes more sense, and is able to support any arbitrary thesis. Eagle Eye was a meh movie but the concept that people's own friends could be made to ostracize them and coerce them to do things, is a major concept in that movie. You don't need violence to make it happen. You just need a swarm of AI bots to coordinate to reputationally outcompete others…

> the concept that people's own friends could be made to ostracize them and coerce them to do things, is a major concept in that movie Can you accurately pigeonhole that thought onto one of these? I'm curious how unique a storyline it is. https://en.wikipedia.org/wiki/The_Thirty-Six_Dramatic_Situat...

I dunno

Have you seen the TV tropes site? Chock full of awesome stuff :)

Post reply on HN