First of all it is abut exchanging information. Web3 is about a decentralized approach. Information is owned and not (mandatory) stored at a central platform. Each network participant keep its own information. If we treat a participant as (software) agent both Web3 and MAS share the same issue, the gossip problem (external link) – in a nutshell “information is only really known if all participants are checked”. Web3 tries to overcome this issue by finding a consensus (e.g., > 50% of participants agree) on transactions – the chain of (transaction) blocks aka block chain or distributed ledger (DLT).
Another notion of a distributed system (Web3 or MAS) is asynchronous communication – or short there is no global clock. To achieve common goals so called “coalitions” are to find. If not each participant (aka agent) might change the objective before the other agent can replay on the previous request. With Web3 those coalitions are established via decentralized executed rules and states(aka Smart Contracts or decentralized executed code).
Next question appears when we consider automation. For Web3 a topic close to automation is called “DApp” (Decentralized App). A smart contract could we one of those DApps, usually running with an self containing (not distributed) state/ knowledge. Sure approaches such as “Oracle” allow to request outside states and make them part of the own state, but this is far from collaborative solutions and intelligence. Agent with more or less intelligent behavior are called “bots”. Due to the following paper (external link) a bot follows a “typical scheme of …
steps:
- agent-subordinator (Head, the central body) decomposes the original problem into
separate tasks; - these tasks are distributed between the agents-executants;
- each agent-executant solves the task, sometimes also dividing it into sub-tasks;
- in order to obtain the overall result a composition, integration of particular results
corresponding to the selected task is produced. Agent-Integrator is responsible for the
overall result (often, this is the same agent-subordinator).”
If so, also a so called decentralized system is execution-wise a centralized system. At the end an intelligent behavior requires to aggregate decentralized information (may be as result of an execution).
Lets have another look on (autonomous) information. If it is autonomous kept, it is an asset. An asset must be unique. Uniqueness also require to have it originated from a central place; otherwise it is divisible. In Web3 we call this unique portion of information a token, which usually is a variable state in a Smart Contract instance (compare the ERC20 external link). Sure cool thinks could be realized if we store a smart contract state in a log, which can’t be changed e.g., block chain. A cool example is the trading of “cat pictures”. Technically we can copy the image information (multiple times), but only the token holder has the original one.
In summary, Web3 is complementary and supportive to MAS (as well as twice versa).