Frame rate, play, and AI

I spend all of my time these days working and building with AI. Mostly LLM systems, but I also read lots of papers on the state of the art, look at new tools and techniques, build out ideas with the teams I work with, and generally am soaking in the events as close to real time as I can. You might say that I have a very high frame refresh rate - I am learning about this as fast as I can because it’s obvious to me that, regardless of the questions and challenges that others point out, there is something real and important here.

Recently though I’ve been talking to more and more people - folks in leadership roles, engineers, friends and family whose only experience of AI is through the news. Most folks have never even had a conversation with GPT-4 or any other model, or generated an image, etc. Their frame rate is very low.

The state of AI is changing rapidly and is likely to do so for some time to come. If you aren’t sampling directly, new capabilities are going to be very surprising to you - they will seem like they’ve come “out of nowhere”. We’ve talked before about the “why not” vs “what if” perspective - being threatened and surprised by something is a good way to get into “why not” mode and miss out.

LLM-based systems are very easy to engage with. Click here and you can talk to a free version of GPT-4. If you haven’t, go ahead! It won’t bite. Ask it for something silly, or something hard. See if you can stump it. Have it tell a joke (it’s bad at that) or draw something (better!). One of my favorite things is to use it as a 'judgement-free explainer’: ask it to explain something that you should understand but don’t, and that you’d be embarrassed to ask a person about.

It’s important to cultivate a sense of play and mess in what you work on. It’s very easy to decide to be too serious, that things have to be perfect before you can try them, to be intimidated. This is a mistake! Innovation and creativity all come from accidents, messes, and open exploration. If you want some more ideas, here’s a great article on some approaches to using AI (by someone I highly recommend following, Ethan Mollick. He often writes things I wish I’d written first).

Read all of the garlic bread example. Trust me, you won’t be intimidated by the time you get to the end of it.

BTW, my friend Reid Hoffman wrote a great article recently about the rule of law and its importance to all of us. Worth a "reid" (couldn't resist).