It Still Can't Do My Job: Four Years of Moving Goalposts (2022

Re: It Still Can't Do My Job: Four Years of Moving Goalposts (2022–2026)

#141

Post by anuramat » Sun, Jul 05, 2026, 2:28 AM UTC

jenniferhooley wrote: post #136

anuramat wrote: post #130

Earlier quoted context omitted.

> 100000x in a year why not? fable 5 (2026) is an inf% improvement over having nothing (2022) in any case, stop pretending that it can be represented by a single scalar value, or at least use log scale units

Oh, you think we are going to have another breakthrough in artificial intelligence rivaling the one we had in 2022 this year or next? What research or studies lead you to believe this? Could I see one? I haven't personally seen the research leading me to believe there's a massive new intelligence breakthrough on the horizon but maybe I missed it! As to the current version - I use AI constantly, as in 12 hours a day 7…

> this year or next

maybe; I meant "in a single year eventually", not this year necessarily

> what research

transformers were published in 2017, and nobody anticipated agentic coding back then; but if I had to guess:

1. lean self play as in https://arxiv.org/abs/2502.00212 and https://arxiv.org/abs/2606.13473

2. "recursive self improvement"/autoresearch

even basic stuff like coding agents removes a major bottleneck -- when finding relevant literature and turning an idea into code takes as much time as writing it down in your notes, researchers from top tier labs can test every single hypothesis they can come up with

> hilarious

"oneshot a PoC -> oneshot a production ready game" improvement sounds like something that will eventually be done in a 1kloc diff in the RL pipeline

IMHO: "genuinely good" might never happen though because of all the non-coding stuff -- you can't really benchmaxx when the target is defined as "pleasant to humans"

Re: It Still Can't Do My Job: Four Years of Moving Goalposts (2022–2026)

#142

Post by solumunus » Sun, Jul 05, 2026, 5:17 AM UTC

cwbuilds wrote: post #26

solumunus wrote: post #19

Earlier quoted context omitted.

It’s not meaningless because AGI = genuinely being able to replace humans in almost any job. Sure, AI is getting better at stuff that humans direct it to do, but the fact that a human is required in the loop is super meaningful.

Maybe I'm wrong, but I think it will always be used as a tool however smart it gets. In software, it hasn't replaced all the software engineers (even though it's better and faster than us), it's just meant that we now have more leverage. We can write more code and work on more projects than ever. I don't see why we would suddenly stop using it like that?

Why do you think you would have a choice? It’s up to the business owner if they want to spend money on AGI rather than hire you, if the AGI is more effective per $ they will choose that.

Re: It Still Can't Do My Job: Four Years of Moving Goalposts (2022–2026)

#143

Post by inigyou » Sun, Jul 05, 2026, 9:55 AM UTC

OtomotO wrote: post #137

inigyou wrote: post #132

Earlier quoted context omitted.

That's true for money, but there are real capital resources whose ownership matters extremely. E.g. every part of food production.

People will not just lay down and wait for death's sweet embrace though. There is a whole lot of people involved in food production and the owners can't own it without people willing to let them own it. As soon as money is out of the equation many powerful lose their power.

The owners in any society almost always have the guns on their side, and they can outgun anyone else. That's what makes them the owners. They can outgun any farmer, because the farmer spends all their time farming but the owner's goons spend all their time training in violence.