What happens when two conversational algorithms find themselves talking to each other?
Two data scientists from Spotify, which is has a strong focus on the tastes and habits of its users, have left the company and set up a startup, Slang.ai, which provides technology for restaurants, hotels and stores that answers phone calls from customers and can take reservations and answer questions.
This is a subject that has fascinated me for some time: I’ve been using recordings of Google Assistant calls in my classes and at presentations that are impossible to discern from a human. The question that somebody always asks is: what happens when two conversational algorithms encounter each other?
The answer is obvious: the two algorithms will “converse” in human language, even though doing so is, from an interaction or efficiency point of view, is pointless. Human language, between machines, is simply a complication: the two computers could simply exchange the necessary data to check if the reservation is possible, verify the request, and update their respective databases in fractions of a second, without going through the complex process of translation into human language.
It’s the same with generative algorithms: there will be cases where one person uses a generative algorithm to turn a simple sentence or a couple of topics into a formal, long…