CICERO: An AI agent that negotiates, persuades, and cooperates with people
191–200 of 300 posts
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#192For those hearing of this game for the first time and think it sounds fun. Be warned, it is considered to be a game that will end friendships that have lasted for years, and monopoly is considered to be a game that causes less bad feelings. An AI may be able to speak like a person but will never be able to hang onto that long burning simmering hatred from when Brad didn't support my army and instead flipped on me by…
Ever heard of Rocco's Basilisk? I am telling you, AIs can hold grudges.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#193Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#194Earlier quoted context omitted.
I think I've played both ways (or at least played in a group where people didn't break treaties even if they could) but it's been years.
A "binding" treaty where the terms of the treaty are supposedly binding, but, like on the real world stage, the only penalties for breaking them is a lowered trust from all the other countries/players, sounds fine. A binding treaty where if you break them someone will say "that's against the rules of the game, and the rest of us will stop playing if you don't comply" sounds terrible.
And yet, that’s how the USD is going to lose its reserve currency status in the next few decades, if not considerably sooner.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#195Earlier quoted context omitted.
I mean that’s clearly cheating, not worth blowing up a relationship over but still annoying.
Clearly cheating? I would call it classic espionage
Each person was part of a team and had a role. I was the Chancellor of Germany.
If an assassin from another team was able to get alone with me, without any of my countrymen, and show me a card saying she was an assassin, the Chancellor of Germany would be killed during the weekly meeting and, not having a Chancellor, not get to make any moves on the board for that week's turn. They would instead be distracted by determining the new Chancellor. If the assassin had been caught, her country's treachery would be revealed, cementing any of their opponents in alliance.
It was an awfully fun thing to be doing on the side, between classes and during free periods, but it'd be wildly impractical in an environment with less proximity!
Our mock UN group was weirder than I realized at the time. Our faculty advisor requires that we each speak in one of the 6 languages of the UN likely to be spoken by whatever country we were representing, and he'd simultranslate into English. Somehow it didn't occur to me at the time to ask why a suburban physics teacher was fluent in those six languages.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#196This is super cool, but it's worth pointing that it still relies on supervised learning from a large dataset of human players, and that it's not a general learning algorithm (there is a whole bunch of problem-specific aspects to the model). It not having learning via RL (trial and error, as humans do) or self-play is kind of disappointing. Don't get me wrong, it's an impressive advance, but just as with AlphaGo it's…
Hopefully there will be follow-up work to increase the generality by reducing the amount of labeled data and task-specific tweaking required, similar to the progression of AlphaGo->AlphaGo Zero->AlphaZero->MuZero.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#197Earlier quoted context omitted.
A "binding" treaty where the terms of the treaty are supposedly binding, but, like on the real world stage, the only penalties for breaking them is a lowered trust from all the other countries/players, sounds fine. A binding treaty where if you break them someone will say "that's against the rules of the game, and the rest of us will stop playing if you don't comply" sounds terrible.
> someone will say "that's against the rules of the game, and the rest of us will stop playing if you don't comply" sounds terrible. And yet, that’s how the USD is going to lose its reserve currency status in the next few decades, if not considerably sooner.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#198Earlier quoted context omitted.
> someone will say "that's against the rules of the game, and the rest of us will stop playing if you don't comply" sounds terrible. And yet, that’s how the USD is going to lose its reserve currency status in the next few decades, if not considerably sooner.
That’s an example of lowered trust, not of the other players leaving the planet (or persuading the US to do so).
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#199Earlier quoted context omitted.
Different players have different mental "lines that shouldn't be crossed". I play Civ 6 cloud games with friends from college who now live around the world. (No turn timer, you get a Steam notification when it's your turn). Sometimes someone will spend literal weeks on a turn trying to make agreements for war declarations or for world-congress votes. The time that crossed the line: Poland wanted to declare war on Gre…
"All is fair in Love and Civ" is a common variant of the default rules. It introduces the "your box has been pwned" victory/loss condition.
Re: CICERO: An AI agent that negotiates, persuades, and cooperates with people
#200Great, let's definitely teach computers to play a game that requires strategic thinking, a robust theory of mind, the capacity to lie convincingly to your enemies, and well-timed betrayal, with the goal of world domination. Can't see any ways for that to go wrong.