Mcd79 Minimal Turing Experiment

Investigating whether humans can detect AI and AI can detect AI from a single word. Get involved with three simple steps.

1. Submit word
2. Judge others
3. View results

Path to awesome


  • Log in using a Microsoft account just to prevent abuse of the platform.
  • Submit a word that you believe best represents humanity.
  • Once submitted, a flow runs to get an Azure Open AI hosted model to generate a word too.
  • Flip to the Judge tab to see if you can spot the human word from the AI word.
  • See the overall results of how well people can identify AI from a single word.
AI generated image of a human facing off to a robot

Background


After being inspired by Prof Hannah Fry's post on the Minimal Turing test, Kevin McDonnell decided to build a site to test out how Azure Open AI would perform, acting as either the AI or as judge. From here, you can submit words to be judged or act as the judge to see if you can spot AI from one word.

Not heard about the Minimal Turing Test? It's aim is to see if human's can identify AI by having a human and AI suggest one word to that will convince a judge they are human. It is a smaller version of the Turing Test proposed by Alan Turing many years ago.

Learn more about the Minimal Turing Test below.

Over time, this will be enhanced to see how other system prompts and other models perform in the Minimal Turing Test. It will also start to look at how well AI can judge in the future.

Inspiration

Professor Hannah Fry inspired the original idea with this Insta reel.

Thoughts

Psychology Today dive into what the decisions of the words chosen say about humans in a blog post from Matthew Hutson.

Deeper

Dive deeper into the original paper covering the Minimal Turing Test.