hperrinMar 20

Is it just me or is everyone who absolutely loves AI just painfully below average?

https://lemmy.ca/post/62061142

Is it just me or is everyone who absolutely loves AI just painfully below average? - Lemmy.ca

Like, every AI generated thing I’ve seen, when viewed from the eyes of someone who actually knows what they’re doing, is at best below average. Maybe some things aren’t quite as bad as the general “AI slop”, but of the things I’m actually experienced in (code and art), I just see so many amateur mistakes in everything AI. Regarding art, AI can make really visually appealing things, but it gets the details wrong. That’s something that a below average artist does. And regarding code, it’s the same thing. Overall, it has the appearance of decent code, but it gets the details wrong, just like a below average dev. (Probably about the level of a high school senior or college freshman.) I’m not super experienced at writing, but I can also tell that it’s not very good at that. The stories it writes just aren’t compelling, but I’m not experienced enough to tell you why. And the same with music. It’s just below average, but I couldn’t tell you why. I’m not trying to sound elitist by saying this, but I’ve noticed people who aren’t very good at these things tend to praise how good the AI is. So, is it just me, or are the big fans of AI just below average at whatever the AI is doing?

Show threadjoshthewaster

Imagine an army of high school seniors at your command - what could you do? Very likely you’ll spend a good amount of time babysitting but you could probably mentor one or two into doing something interesting and with all the rest you could get mediocre book reports. That’s not nothing. Is it amazing art, no. Incredible writing, no. Truly new or interesting ideas, also probably no. Can you get a lot of stuff done that is pretty boring and repetitive - yes. And I mean a lot, no sleep or breaks, just boring repetitive tasks.

I think one of the biggest miscalculations right now is how many of those tasks there are that actually need doing. It’s more than a lot of people might guess, so many forms, shuffling of data, and other burecratic BS that we just don’t really need people to do now. On the other hand, a lot of those forms need to be analyzed or data that needs to be understood but is that really all that much? Idk. I do think AI art misses the point but an army of high school seniors, wielded by someone with the skills and understanding to tackle a real problem can be remarkably effective.

All this to say, hype people gonna hype but other actually smart people are quietly learning how to wield “AI”.

Mar 20 at 6:22amWeb
Show threadhperrinMar 20
Well I guarantee you I couldn’t get a project with actually good code. I should know, I started writing code at 10, and by the time I was a high school senior, I was still pretty dog shit at it. That was 20 years ago, so I’m a bit better at it now.
Show threadteslekovaMar 20
I can do better than that: I’m even worse now because I stopped twenty years ago!
Show threadteslekovaMar 20

The problem is, if it’s repetitive enough we can already automate it without AI, and if it isn’t, AI will make enough boneheaded mistakes that it needs proofreading.

I do expect that proofreading will become easier over time as we all get a better idea of what sort of mistakes the AI makes. So we might actually start to improve productivity, in a sheer volume sorta way.

However, proofreading a lot of mediocre work is basically what teachers marking assignments do. If you want to know how bad that would be as a full-time job, ask a teacher how much they like marking assignments compared to the rest of their work, and then look at how much they like their job overall.