I’ve been thinking and doing some coding to make different AI’s negotiate with each other.
What is the best medium of exchange of information between AI’s?
In a simple bargaigning, a seller tries to sell her product at a given price. Then a potential buyer makes an offer, lower than the asking price and the process goes back and forth until some offer is accepted or not.
What if we want two “machines” to be in charge of doing that kind of negotiations.
The simplest exchange of information would be something like:
> buyer: offer(500) > seller: ask(1000) > buyer: offer(700) > seller: ask(800) > buyer: offer(800) > seller: accept_offer(800)
That’s really simple, there’s no AI or complex logic behind that kind of negotiation. But what if we add some complexity, let’s say the product in the negotiation is a laptop:
Seller: “I want 800 for the laptop, I can’t accept 700. But I can include an additional battery of 3 cells with no extra charge”.
How are we supposed to translate this english message to code?
If the whole domain is entirely known in advance by the two sides of the negotiation, we’d be able to translate it to code.
But in real world negotiations, there’s no knowledge of the whole domain, at least most of the times. It’d be very difficult to even come up with an approximate solution with this approach.
The better medium of exchange between AI’s is human language.
The AI’s are no humans but I think that they will still communicate using human languages.
A big challenge is that the human language is unrestricted, so we’ll have to find ways to restrict the AI’s in order to get (near) predictable workflows.
Leave a Reply