We’re all still hearing about the looming threat of AI, and how important AI is, and how it’s taking over jobs and how we’re (I’m not buying it, but whatever) on the verge of an Artificial General Intelligence revolution and all of that related bullstench, but I think it’s also important to remember that, fundamentally, AI is an idiot. An idiot who is being used to do things like generate images of motorcycle wheel stands on Temu, and screwing that up with genuinely astounding energy.
We’ve covered this sort of thing before, and I always find it fun and perhaps maybe important, because it’s a good reminder that no matter how natural-sounding the answers some AI agent may give you, or how incredible an AI-generated image may look, fundamentally everything AI produces is derivative mush pulled from vast arrays of human creativity and work, and these non-self-aware machines have no idea what is real or not, what is right or wrong, or what is true.
AI systems have never been outside, never felt how water beads on skin, never driven a car, never eaten a hot dog, never laughed or cried or loved or lathered or licked anything. They literally have no idea what anything they output actually is, which is why we get spectacular amalgamations like the images in this Temu ad for a heavy duty motorcycle wheel chock:
We were sent this glorious madness by The Car Respecter on eX-Twitter, and it is really something remarkable. Let’s take a closer look at what’s going on there:
Wow. So, before we really dig into this, here’s a picture of the wheel chock in its actual context and scale:
See? It’s what it says: it’s a chock for one motorcycle wheel. It’s not big. But look back up at that picture there: the AI seems to think the wheel chock is some kind of massive lift-type thing taller than a man, and a motorcycle is a sort of three-wheeled, eight-foot-tall monstrosity with an upper wheel perpendicular to the other two, but in line with the seat, so I guess you’d ride sideways on that thing? Somehow?
Something like this?
I suppose no human at Temu actually looks at these things, which is likely the worst possible way to use AI, period. I could be appalled at Temu’s lack of quality, but even typing those words felt ridiculous.
Remember, what all of these current AI models do is gather and re-organize pieces of already-existing content from their datasets; they don’t actually create. They can produce convincing images, but there’s no awareness about what that image actually is; it’s just comparing images with other ones that have been identified to be a certain category of image. It’s impressive, sure, but it’s also not a replacement for a human. A powerful tool, sure, but things like this only happen when there’s no actual wet, biological brain in the mix.
AI is impressive at so many things, but, fundamentally, still an idiot that knows nothing about anything.