Chat GPTom

What does Tom know about ___ ?

Responses generated by Open AI GPT-3 API using Babbage model.

If you're anything like me, your internet feeds have been full of chat about GPT lately.

Like any other reasonable person, I fretted about how this might impact the industry at large, and specifically my place in it.

So I decided to flip-reverse the situation and use the OpenAI API to help visitors to my site discover more about me, my experience and what I bring to the table: presenting Chat GPTom!

I encouraged users to click through and fill in the blank to discover what I "know" about their topic of choice.

Some of the answers provided were… interesting. But who was I to question the all-knowing wisdom of GPT-3?

I took a single-word input ${topic} and used it to form a prompt under the hood:

"Explain succinctly but humorously the relation between ${topic} and remote-working web developer Tom Corbett."

This is the actual prompt sent to the GPT-3 API, priming the response to include a few details about me - my full name, industry, lifestyle - and to do so in a certain style.

By asking the language model to explain a relation that likely does not exist (e.g. me to sausages) it is encouraged to get "creative" and offer non-factual information. By including more details in the prompt than are in the original question posed, the response also seems as if the AI "knows me" specifically.

The combination of these twists and the inspired user input made for excellent results, the best of which are showcased here.