

Autonomous Intelligent Agents for Web3

Semiotic AI is building automated decision-making tools for decentralized markets on the blockchain.
We believe that the convergence of AI and cryptography will propel Web3, the decentralized internet of the future.
Deep RL
Multi-agent Deep Reinforcement Learning for Automated Negotiations
Simulations
Realistic environments to train negotiation agents and test decentralized protocols
Security
Homomorphic Encryption and Zero-Knowledge Proofs for verifiable and private computations.
Deep RL
Deep Reinforcement Learning (RL) is a field of AI that was used in mastering games like Go and in scientific discovery.
Using Deep RL, we develop autonomous agents that can negotiate price and other terms in bilateral or auction type deals.
Agents with prosocial behaviors can be trained to improve the welfare of the decentralized network.

Simulations
Simulation environments are essential in training AI agents efficiently and testing their performance against benchmarks.
We develop realistic simulators to capture economics of decentralized markets and train multiple agents that participate in the network.
Simulation environments are also useful in stress testing new protocols to find weaknesses (e.g. adversarial attacks) before real world failures occur.

Security
Homomorphic Encryption makes it possible to perform computations on encrypted data.
Applications include:
-
Private negotiations and sealed-bid auctions on the blockchain.
-
Decentralized applications with data privacy.
-
Confidential executions and transfers on public blockchains.
