Global AI News Aggregator
By
–
Meta AI’s @anton_bakhtin discusses how he and his team used Reinforcement Learning to teach CICERO the best moves for the millions of possible actions that could be made on any given board state. pic.twitter.com/W8SJKDz6Vf
— AI at Meta (@AIatMeta) 10 décembre 2022
Meta AI’s @anton_bakhtin discusses how he and his team used Reinforcement Learning to teach CICERO the best moves for the millions of possible actions that could be made on any given board state.
By
–
Meta AI’s @ml_perception reveals what was used to build CICERO: Theoretical game planning component Dialogue model Component filtering out any dialogue mistakes before messages are sent to players
By
–
it is pretty simple at the moment – just keeps a buffer of the past N conversations in memory. a pretty simple improvement would be dynamically adjusting N based on the length of the conversations
By
–
excited to hear your thoughts/ideas! right now its pretty simple: just take the last N lines of conversation. but a very natural extension to determine N dynamically based on their length
By
–
What surprised @polynoamial most about CICERO? Its ability to be a great ally. With built-in strategic reasoning & natural language, #CICERObyMetaAI can understand and help its allies achieve their goals.
By
–
Meta AI’s @em_dinan explains what makes CICERO different from other language models. Its ability to cooperate with humans through natural language, Its grounding in specific intentions and plans for the game board.
By
–
…and ranked in the top 10% of participants who played more than one game in an online league. Learn more about CICERO’s win rate: https://
bit.ly/3HigP5J Do you think you could outplay #CICERObyMetaAI?
By
–
It’s not just that #CICERObyMetaAI can play Diplomacy with humans, it’s also that it often WINS: In online games, CICERO achieved more than double the average score of human players…
By
–
@BruceHammer helped add a requests chain This hits a given URL, and then passes the parsed response as context to the LLM. Very useful interacting with the web! See the example of doing a google search here: https://
langchain.readthedocs.io/en/latest/exam
ples/chains/llm_requests.html
…
By
–
LangChain version 0.0.31 @johnvmcdonnell fixed how Agents handle a missing tool (so no longer errors) Bash utility and a LLMBashChain (
@coyotespike
) a generic transformation chain (
@AkashSamant4
) a requests chain (
@BruceHammer
) on these additions: