- The AI Daily Brief
- Posts
- Rate Of Change
Rate Of Change
We probably shouldn't trust our own predictions on the speed of AI evolution
An incredibly post went viral on Reddit yesterday. It showed a series of comparison images created with Midjourney a year apart with identical prompts - one in April 2022, one in April 2023.
Yup. And there’s this one:
And this one:
One year. These are one year of evolution of text-to-image.
At Midjourney’s “Office Hours” earlier this week, they said they were training v6 and already working on v7 as well (we’re on v5). Their goal is to release every 30 days. Some of their forthcoming updates include the system being able to read prompt words/statements in context and researching character consistency.
And then of course there is text-to-video. This week Runway released an iPhone app for its Gen 1 tool that allows users to video-to-video modify up to 15 seconds at a time. It also started rolling out Gen2 text-to-video in beta and already users are stringing together complete movie trailers and short films.
This entire video was made by a guy typing some words into a box (a bunch of times). The degree to which that this will unleash human creativity is stunning.
— Nathaniel Whittemore (@nlw)
10:54 PM • Apr 27, 2023
The number of reply-guy comments I got on that one, by the way, that were some version of “Stunning? Really? Look at the hands!” was fairly staggering. Something something, forest trees.
Let’s hold aside a moment for the discussions of human creativity, or what these technologies mean for artists, or entire industries really.
Let’s just focus on the rate of change. The tiny, infinitesimal sliver between those janky third-grader art shots to utter photorealism. The fact that a thing that didn’t exist 5 days ago - true consumer text-to-video - now exists, even if only in 5-second bursts and with wonky fingers.
It’s not just that the rate of change is fast. It’s that effectively no one thought it was going to be as fast as it is.
Extrapolating that out, is one of the natural conclusions that we sort of have to reject all previous predictions for the pace at which we get to seminal milestones like AGI? This is certainly what has caused Eliezer Yudkowsky and others like him to get so glum lately.
And even if none of the AGI worst case scenarios happen, we’re still talking about technology that is going to radically transform huge swaths of culture and the economy, also at a faster rate than anyone thought.
There are some who see this rate of change and see nothing but wonder; pure joy at possibility incarnate. There are others who see the rate of change and feel only dread.
For most of the rest of us paying attention, it’s everything all at once. A feeling of acceleration and free fall - of endless opportunity and chaotic unknown.
It’s the age of anxitement.
What We Talked About Today:
Thanks for reading! If you want more AI Breakdown:
The AI Breakdown podcast - https://pod.link/1680633614
The AI Breakdown YouTube - https://www.youtube.com/@theaibreakdown
Signing off from the future’s past - NLW