Shoulder to the Boulder
We’ve all done things that seem endless. There’s a task that makes your eyeballs roll backwards every time, and you know there’s pretty much no point in even doing the thing.
You do it anyway. What other alternative is there?
Maybe you’ll push just a little harder and get that boulder over the top of that hill once and for all, but what’s more likely to happen is that you do it all over again tomorrow.
As payback for trying to become one of the gods, Sisyphus ends up literally doing exactly this: pushing the same boulder up the same slope every day.
Ever feel like this? I’m starting to feel like our old friend Sisyphus while trying to keep up with AI. A small handful of people are using AI tools to understand other AI tools, and it’s turtles all the way down. You can certainly choose not to play the game if you want to, but I would not anticipate being able to keep up if that’s what you choose.
Also: the people who choose to keep up will end up with a disproportionate amount of power, so it’s ultimately up to them whether or not those who don’t keep up are treated well.
Anyway, I’m trying to keep up by leveraging current AI tools so I can understand the capacity of what’s coming next just a little better. I’m starting to get the sense that, by the time I begin to understand tool A, tool B will already be released. Then, tool C will come out even sooner, before I’ve had time to adapt to tool B.
While the myth of Sisyphus does a serviceable job of illustrating the seemingly futile task of trying to keep up, my modified Frogger analogy is useful in describing just how fast everything is beginning to move these days:
If you lived somewhere without electricity, this was another log to jump onto as the world turned itself on. If you already knew about machines, you were on a moving log, and this made the jump to the faster log of electrification.
and
Today’s fast-moving log is very clearly AI. This is true for anyone who works at an office, but it’s also true for the rest of us. That’s because this is changing everyone’s lives, just as surely as the industrial revolution did the same thing about 200 years ago.
Starting to use AI tools is considerably easier if you’ve been on the internet for a while, but if you’ve never worked on a computer, it’s easy to imagine why this might never happen.
Getting off of one fast-moving platform you’ve just started getting used to, in order to jump onto an even faster-moving platform? This sounds like a Sisyphean task if ever there was one.
Like Sisyphus, you’re never going to get to the top of that hill—never going to be able to rest on your laurels/platform. The alternative—to allow the boulder to crush you, simply being left behind for every important decision that gets made from now on.
Make no mistake about this: all of the important decisions will be made shockingly quickly from here on out, but the time compression due to AI will make it seem like a thousand years to the people thinking about these things.
If you had a thousand years to think about a single problem, how good of an answer could you come up with? AI with a human in the loop compresses time, in a manner of speaking, by running a billion times more calculations per second than we can without AI.
If you can think fast enough, maybe you can understand how to get on the next platform before the one you’re on becomes irrelevant and useless.
Time to put your shoulder to the boulder. Time to learn1 about these tools.
Recommendations include
, , and - all three are doing slightly different approaches related to AI. Go dive into some of the specifics with them!
A counter-point could be that AI is advancing so fast that, when it comes to a certain type of model or product, you might as well wait until all the weird quirks and bugs and half-functioning elements are ironed out and the full, polished product comes out later. I've witnessed lots of AI models go from barely working and requiring lots of "hacks" and workarounds to suddenly being solved. For instance, in the early days of Stable Diffusion, you had all those voodoo ways to try and make it produce better images (e.g. the infamous "splatterprompting" I wrote about once) - and then the newer image models just "get" what you're saying and make perfect images almost every time.
So in some areas, you can probably afford to let the boulder roll on without you for a bit, only to catch a better boulder later. Uh...that analogy rolled away from me.
I have been repeatedly told and shown that using AI to write is bad, so I don't do it. Likewise I'd rather support human artists rather than subsidize AI programs, but I feel like if I'm going to offer any merchandise based on my characters somewhere, I might have to use it for that.