We are currently developing a suite of interactive Bot or Not? tests where participants don’t simply rate a sample that they can read or hear as many times as they like. In our very first text-based Turing Edition we ask pairs of participants to join a messaging platform. They don’t know who (or what) their partner is. They could be two humans, a human and an AI, or even a pair of AIs. Yes, we want to know how good AI is at this task too. The participants have a short text-based Q&A-style conversation with each other, and at the end, each must judge whether they think they’ve been talking to a human or an AI.

How well are our participants doing? We can’t reveal that yet…

This is our latest line of work and it’s under way right now, so feel free to keep checking back in for updates.