@emollick | 99,933 followers
AI has been destroying humans at adversarial games, like Chess or Go, for a bit.

But now it is successfully outperforming humans at Diplomacy, the classic multiplayer game that requires natural language negotiations over chat, as well as strategic lying. science.org/doi/10.1126/sc…

Tweet Engagement Stats

Stats are based upon replies and quotes of this tweet

Engagement Map

Quotes
Replies
Retweets
Trendsmap

Replies and Quotes

Total of 33 replies and 57 quotes found
In reply to @emollick
For more on the legendary friendship-destroying game that is Diplomacy. foreignpolicy.com/2020/10/23/the…
 
In reply to @emollick
Yes AI might be better at Diplomacy, but can it gather in a house for 10 hours to actually *play* Diplomacy?
 
 
In reply to @emollick
It turns out it doesn't actually require lying. This system plays at the 90th percentile without it: twitter.com/ml_perception/…
 
In reply to @emollick
Has this been done for Settlers of Catan? Asking for a friend..
 
Oh good good, the AI is better at us at lying, super great.
 
This is super interesting!

We believe closed game like diplomacy are easier than open-ended games like ours but we would love AI researcher to look at conquest.

There is currently 2 humans controlling the conquest universe with hundreds of $ to grab.

Ready for the challenge?
 
Una de esas cosas que son a la vez chulas y escalofriantes.
 
Yes, finally! Strategic lying is a good thing for adversarial AI to develop.

Applications include ethics-washing large web-scale models:
"No, my original dataset is definitely non-infringing."
"Ah, well... we don't consider models derived works."

(It's a joke, lighten up!)
 
 
Because obviously what the world REALLY needs is an artificial liar which is better at political deception and "strategic lying" than humans.
 
 
 
 
In reply to @emollick
Pretty amazing. And kinda scary. In this example, Cicero is the AI... pic.twitter.com/TV8JVPEy2x
 
humans: we sure hope AI won't destroy us!

also humans:
 
 
Cool, now it just needs to be able to make a cup of coffee with fluent sensorimotor interaction and we can kill all humans
 
More ominous than you might think 😳
 
 
This is cool, and scary. And pragmatically I’d give the first assignment to negotiate cost savings in the state bureaucracy
 
ai gonna replace diplomats before it replaces truck drivers
 
In reply to @emollick
Interesting. I wonder if a game could be created where AI has to run a society and optimize for a high standard of living, high quality of life for citizens while reducing carbon footprint & reducing inequality for example. Would be super interesting the AI solutions.
 
In reply to @emollick
 
🎵 And nothing can go wrong… 🎵
 
In reply to @emollick
This is really cool! I wonder how much of the AI’s success can be attributed to its negotiation module vs. the strategic analysis module. I.e. it might be far superior to any human at the realpolitik stuff and just adequate at the communication stuff.
 
this is actually the major sign of the end timea
 
Oh great, we taught AI to lie successfully to us to accomplish its goals. This definitely won't be a regrettable development, no way
 
צעד נוסף קדימה בתחום של בינה מלאכותית, מנצחת בני-אדם גם במשחק "דיפלומטיה", משחק מרובה משתתפים שאינו משחק סכום אפס. יש בו צורך לשכנע 6 משתתפים נוספים באמצעות יכולות מילוליות גבוהות.
#פידטק
(בקרוב הפרק על AI ומקומו בגיאופוליטיקה בפודקאסט "מגרש עולמי")
 
The real test is how well it schmoozes at diplomatic cocktail parties.
 
When #AI surpasses human in negotiation skills
 
In reply to @emollick
May someone never entrust the world "peace keeping " role to a supposedly smart AI network. This look a lot like some of the bad sci-fi scenario
 
 
When AIs can negotiate more effectively than humans…
#AI
 
In reply to @emollick
Yeah but “over chat” is key. Get back to us when it can do the same “in-person”.
 
In reply to @emollick
I am unfamiliar with the game, but looking at your description of it makes me wonder if it might be possible to get the artificial intelligence to learn to beat humans at the game while avoiding lying as much as possible in the process. Maybe even avoiding outright lying entirely
 
ICYMI @timhwang, a possible add-on to future team Diplomacy sessions...
 
This is kind of freaking me out: an AI "inferring players' beliefs and intentions from its conversations and generating dialogue in pursuit of its plans". 😬
 
AI driven foreign policy.

el salvador you wanna try that one out also?

perhaps someday it won't matter if we can parse the difference of human influence vs machine, for the game asks ever forward motion, and we were already held to it's whims.
 
 
In reply to @emollick
this is a pretty good real life Turing test
 
Soon we will be able to use AI to deal with timewasters like open source project politics, Mastodon policy disputes etc.
 
So this could be a possible explanation for everything going on in 2022! The dangers of AI is real and is a HUGE threat
 
As a childhood player of this unbelievable game, I cannot like this enough.
 
 
 
I'll only believe this when the AI can repeatedly get solo victories at a high rate playing as Italy or Austria-Hungary. 😂
 
Tenhle týden jenom blitz Diplomacy, ale stejně - vypadá to, že budeme jenom přihlížet válkám navzájem se obelhávajících AI, a nebudeme moc tušit, co a proč se vlastně děje.

Jako doteď? :)
 
 
« Cicero uses a strategic reasoning module to intelligently select intents and actions. This module runs a planning algorithm that predicts the policies of all other players based on the game state and dialogue so far, » pic.twitter.com/EAP2Tp5HOr
 
 
In reply to @emollick
Dave..... Dave...... I'll open the bay door after you help me take Austria on the next turn pic.twitter.com/pxTJIGFAbA
 
 
In reply to @emollick
Looking forward to AI and robots paying taxes, buying products and services.
 
WHAT PART OF “SUFFER NOT THE SILICA ANIMUS TO LIVE” WAS DIFFICULT FOR YOU ASSHOLES TO UNDERSTAND?
 
In reply to @emollick
Fascinating article & game selection! The Meta guys had a paper a few years back on negotiation bots using reinforced deep learning that they trained on crowdsourced sets of mock negotiations. This is a perfect evolution of that!
 
I only read a few screens before panic-wondering if the Swedish forest my grandfather emigrated from might still be available.
 
As someone that used to play Civilization nonstop growing up, this is something
 
 
 
While our attention is focused on consumerism, overconsumption, and brutality, AI systems are learning to negotiate and reason.
 
Again I ask, why do we need this ?
 
In reply to @emollick
AI hasn't come up against the Diplomacy mastermind that is Vertex yet
 
This is impressive…and somewhat frightening. Shows the potential of AI even when the “data” is natural language.

Why is it scary? Well…Wanna know who else makes decisions on natural language data? Financial advisors.
 
Not quite the dramatic moment that #Go was but a new milestone
 
It just keeps getting better. At a scary rate.
 
In reply to @emollick
Doesn’t this neglect the most important part of diplomacy — enabling diplomats to enjoy themselves?
 
In reply to @emollick
what does this mean in practice?
 
 
The power of AI being trained by @JustinTrudeau - strategic lying!
 
 
I for one welcome our new diplomatic overlords
 
In reply to @emollick
Humans are profoundly easy to hack. Anything beyond an antelope and we're useless. $200 million is lost to catfishing each year. That's how easy we are to hack.
 
Over 17 replies and quotes not shown

Retweeters

1,302 retweeters not shown
For access to this functionality a Trendsmap Explore subscription is required.

A Trendsmap Explore subscription provides full access to all available timeframes

Find out more

Thanks for trying our Trendsmap Pro demo.

For continued access, and to utliise the full functionality available, you'll need to subscribe to a Trendsmap Pro subscription.

Find out more

This account is already logged in to Trendsmap.
Your subscription allows access for one user. If you require access for more users, you can create additional subscriptions.
Please Contact us if you are interested in discussing discounts for 3+ users for your organisation, or have any other queries.