A rule for getting ready for the Singularity: “With technology and tools — always try again.”
Maybe a year ago you tried to use an LLM for some task. And maybe you didn’t like it. ChatGPT wrote code that wasn’t good enough. Or it drew people with extra fingers. Or it couldn’t put the right text on an image. Every time you hit that kind of imperfection, the thought arrives: “Nope. I’m not going to trust an AI that can’t solve a kindergarten problem.” “I’ve seen enough of this tool — I’ll do it myself.”
So here’s the thing. Try again. What worked badly a year ago works beautifully now. What didn’t work at all six months ago now works, sort of. And if it doesn’t work today, it’ll most likely work in a couple of months.
“LLMs write bad code today,” people say. And I immediately think of this clip: https://youtu.be/Ddsm-tvk0As?si=tOG2ceqFMiLMm4RK&t=84 It writes code!
Maybe we haven’t arrived at the Singularity yet. But we’ve definitely left the territory where keeping up with technology was easy. Fundamental knowledge, thankfully, barely ages. Applied knowledge, on the other hand, is going stale while you read this post.
Just to be safe, I asked an LLM whether the Singularity had arrived. Claude said not yet — but suggested I check back tomorrow.