DALL-E: OpenAI's Latest Exponential Reality Check

Last year, I sent you all my post (and erm … follow up post) about how things are starting to feel like they are accelerating far quicker perceptually to me. Between GPT-3 and Alphafold, anything feels possible in 2021 already. It just feels like science and technology are moving a lot quicker than ever before and many of the things we expected a, ‘few years from now’ are here … like, right now!

Well, today’s OpenAI DALL-E announcement is no different:


To some extent, everyone saw some kind of image + text based GPT-3 coming. But like most people, I just didn’t expect it so soon. Privately, I had my bet set on Spring of 2021 but here it is less than 5 days into the new year. The results are already incredible and this is based on a transformer with a parameter count a lot smaller than GPT-3.

My mind has been racing imagining the creative possibilities with such a tool. But at the same time, I think I need to re-evaluate my own internal perceptions on how fast technology is moving, set goals differently, and benchmark myself differently.

I’m trying to come up with some kind of heuristic to navigate the increasing exponentional-ness of the times we are starting to really perceptually live in, but it’s really hard for me to think this way. How do you plan for the future or decide on projects you want to work on if so much will change so quickly? The only thing I can think of is to just work faster and set more concrete goals, on shorter timelines, but I know this is not enough to formulate some framework for navigating the world.

I think we have 2 weeks to ourselves to accomplish something significant, because honestly, at the rate we’re going, I couldn’t comment on 3 week goal timelines at all. Every month or so, it’s like we are entering a completely different world with new rules and game dynamics to play. Savour this moment while you can, it’s already on its way out …