Meta researchers have created an artificial intelligence called CICERO whose mastery of human language allowed it to achieve human-level performance in the strategy game Diplomacy. For those not familiar with it, Diplomacy is a seven-player game that blends elements of Risk, Poker, and the TV show Survivor. It is available as either a board game or online and CICERO was trained online where it achieved over twice as high the score of average human players. CICERO even managed to rank in the top ten percent of players who played the game more than one time. The Meta researchers have published both a scientific paper on CICERO and a couple of videos about it.
Language Model Training
“To build a controllable dialogue model, we started with a 2.7 billion parameter BART-like language model pre-trained on text from the internet and fine-tuned on over 40,000 human games on webDiplomacy.net. We developed techniques to automatically annotate messages in the training data with corresponding planned moves in the game so that at inference time we can control dialogue generation to discuss specific desired actions for the agent and its conversation partners. For example, if our agent is playing as France, conditioning the dialogue model on a plan involving England supporting France into Burgundy might yield a message to England like, “Hi England! Are you willing to support me into Burgundy this turn?” Controlling generation in this manner allows Cicero to ground its conversations in a set of plans that it develops and revises over time to better negotiate. This helps the agent coordinate with and persuade other players more effectively.“
We further improve dialogue quality using several filtering mechanisms – such as classifiers trained to distinguish between human and model-generated text – that ensure that our dialogue is sensible, consistent with the current game state and previous messages, and strategically sound.
CICERO combines its natural language skills with strategic reasoning to win the game but its creators explain that it could also be an AI assistant to help people learn new skills or advance NPCs in games that assist players with winning strategies or converse like humans. CICERO’s creators have made it open-source and it is available on GitHub.
And by open sourcing CICERO’s code, we have high hopes for others to build responsibly upon our work, for use inside and outside of the metaverse.
Discussion (4 replies)
Join Discussion →Yep, fire the state department, and put this in.
Set parameters to no nuclear war, and its likely an improvement.
Same with any politician at the federal level really.
Can you imagine AI congress and house, who's paths and choices are limited to what the people they represent want? It would be INSANE. People would actually have influence in government again! Lobbyists would be going after population centers to try and sway their input to the Algorithm!! Good god that's a perfect setup for a kinda awesome just around the corner semi sci fi drama!!
You could have ousted politicians burning their resources to try and strip away power from the AI with them being the checks and balances and the people voting on it. Large 'voting' parties being thrown by organizations where people sign over their right to vote for x number of elections to attend the super high end party. Stuff like that... huh... actually that's kind of interesting.
It would be a story where the AIs are incorruptible until the evil lobbiests and special interest groups can inject a code update. The hero would be a guy working with a security ai to stop them and keep the governmental AIs pure. Tons of intrigue. Action with assaults on district systems to attempt to take them out. Cloud hosting companies trying to wrest control away from the people... tons of interesting opportunities there. You could even make it more right leaning by giving states their own ai banks for state level representation and have the AI clusters moving policy between state and federal level based on what makes the most efficient choices for the people.
Man the level of effort to corrupt and pervert it would be astronomical from preachers saying ai is the new devil to other nations wanting to take out the country before their people insist on the same... it is a whole world rife for creation of new and interesting stories.
Maybe just interesting to people like me.
It would be an interesting story, but maybe if you want it closer to reality, it would be corrupted from inception. People would be allowed to have discourse about it, against it or otherwise, but it would be made into a crime to pass human based laws against the AI, or attack it . In reality the AI would be just a whole complicated theater so it will look and feel like " people" have influence, but in the end its would be just kind of complaint box that no one reads, and does nothing for anyone, but for a cabal of rich people.
The AI would talk and talk, then no action, or action but just so it doesn't work, or just so enough space for corruption so whatever is passed is not what it was talked about so on. In other words, AI job would merely be of appearance, and nothing else.
Then you would have true believers and defenders paid for, and people active against it, ... But, plot twist, the opposition leaders are indirectly paid for by the same cabal of rich people, they only want ' opposition' so that people can vent and think they are doing something and have a say in yet another controlled theater. Real opposition is violently suppressed in secret. In reality the mayority of people instinctively know what is going on, and don't participate. Years pass and millions sit in the passive middle, knowing human law and real action is a crime, so eventually the whole theater starts collapsing on its own.