Live data from Hacker News

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

publicznyprofil.github.io

71–80 of 143 posts

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

#71

Although it's not the most important thing here, I love the IQ badge on the right side of the screen. At IQ 200, the AI is finally qualified to carry a pager for production. I'm in the pagerduty lineup, and shockingly, my IQ isn't even a mere 185.

Perhaps that is the true measure that the author found, and humans have actually been winging it since the inception of programming? ;)

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

#72

You should add a parallel timeline of how many times AI CEOs have claimed their next model is too dangerous to release, AGI is months away, or some white collar job will be obsolete within 6 months. I don’t know anyone in the tech industry who thinks AGI will never happen or that software engineering and white collar jobs can’t be automated. We all read sci fi, you’re not some unique visionary for anticipating AI. Th…

You don’t know me, but I am in the tech industry and believe AGI will never happen. Albeit that is because I did a BSc in psychology where I developed a deep distrust for intelligence research and concluded that intelligence is not a useful term in philosophy nor science (and especially not in engineering).

I strongly agree. In scientific terms, we already know how to achieve 1G constant acceleration space flight. "All we need to do" from an engineering standpoint to achieve it is develop miniaturized fusion reactors and superconducting electronics. Hell, something approximately as good was achievable since the 1950s with a gargantuan ship that pooped out atomic bombs through a giant shock absorbing pusher plate[0]. Trivial, right? Except it isn't. Both of these projects, which are scientifically feasible, are completely impossible to actually build. They are off the table in engineering terms.

We are so much closer to 1G constant acceleration space flight than we are to AGI. We know, in principle, how to achieve 1G travel. We don't know, in principle how to achieve AGI. Our best guess so far is something along the lines of "emergence" which means "maybe if we do enough matrix math in the right way it'll wake up and become a being with agency and intelligence". Another way to say this is "hopes, prayers, and lots of GPUs".

Let's all get a grip. Without a coherent theory of intelligence, you aren't gonna make one in a lab. That's not how science works, it's not how engineering works. Start at the beginning.

[0] https://en.wikipedia.org/wiki/Project_Orion_%28nuclear_propu...

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

#73
post #51

You forgot the first goalpost: the Turing test.

Even that line's very debatable though. Passing always under every circumstance? No.

Passing specific tests to the point that the internet is now full of "Is that content AI generated or not" debates? Yes

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

#74

Earlier quoted context omitted.

I agree. I don't think they're mutually exclusive. You can have something smarter than you which you use as a tool. My point was that there's nothing objectively special about our level of intelligence, so it shouldn't be used as a benchmark.

> You can have something smarter than you which you use as a tool. That sounds like how tech CEO's treat their employees! Kidding aside, I think the notion that you could control something (legitimately) smarter than you is a pretty risky proposition. (Fortunately one that I think is actually far off)

I'm able to control my coding agent even though it writes code better and faster than I can.

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

#75

Earlier quoted context omitted.

I agree. I don't think they're mutually exclusive. You can have something smarter than you which you use as a tool. My point was that there's nothing objectively special about our level of intelligence, so it shouldn't be used as a benchmark.

> You can have something smarter than you which you use as a tool. That sounds like how tech CEO's treat their employees! Kidding aside, I think the notion that you could control something (legitimately) smarter than you is a pretty risky proposition. (Fortunately one that I think is actually far off)

> I think the notion that you could control something (legitimately) smarter than you is a pretty risky proposition.

Well, Trump seems to control a lot of people given how afraid they are.

Trump is also called not the smartest person out there.

So...

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

#76
post #44

Earlier quoted context omitted.

If we don't have access to capital, we will abolish capital altogether. This still backfire to the oligarchs

Are you sure the capital-enabled won't abolish us for trying to do so?

They will certainly try, but they don't have the capability yet

It's probably a really bad idea to keep building them autonomous killing machines though

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

#77

As my company experiments with AI-heavy programming, I don’t feel like my job’s going away, but as a “full stack” programmer who’s also good at and comfortable managing a task board, walking “stakeholders” and SMEs through requirements-gathering, et c., I am increasingly wondering why our “teams” still have a couple non-programmers doing that stuff. Dedicated QA, for that matter—all that test-writing and test-data-ge…

In my experience, those roles never existed to begin with in well run teams and companies. Some companies drive product managers into this role, but that was always a bad use of their time. Even with AI tools, I still don't feel like I have the skills to do the job of a good product manager, which requires the kind of vision and business acumen that I am just not as good at as the best PMs I've worked with.

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

#78

Earlier quoted context omitted.

I’m unaware of any humans that don’t have this error method also.

Look, I don't spend most of my time online criticizing AI progress. But what does your response even mean? People hallucinating work and solutions isn't commonplace at all, right? What industry do you work in where people hallucinate with frequency?

I can't speak to GP's intention, but I've personally witnessed a guy on my team who was trying to position himself as the go-to technical dude. He was jockeying for a management role. When QA or customer support had questions about our products, he'd always have an answer. I would say that at least 50% of the time, his answer was completely fabricated nonsense. He'd wildly misrepresent projects that his teammates were working on. I also saw several incidents of cargo-cult programming from him. Bizarrely, this never bit him in the ass and now he's a middle manager at a FAANG. This experience leaves me without much hope for the future of software development as a career.

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

#79

I like that this goofy fear-based boosterism is on the front page at the same time as an actually well reasoned and well written article about how fear-based boosterism has actually harmed the AI industry instead of making everyone panic-buy like they intended. I assume that’s entirely due to not being able to downvote submissions on HN. As for the article, as another user put it: > Call me when it stops making thing…

I have really enjoyed this period of HN, because there has just been a genuine split here on this topic since it became The Thing, which means I get to see a whole ranges of opinions and arguments on this when I come here. I'll be really bummed if either the doomers or the boosters (or the pragmatists) get run off of the site. A monoculture bubble wouldn't be nearly as useful to me.

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

#80

Earlier quoted context omitted.

Are you sure the capital-enabled won't abolish us for trying to do so?

They will certainly try, but they don't have the capability yet It's probably a really bad idea to keep building them autonomous killing machines though

I wouldn't be surprised if multiple nation-states already had them without calling them such because 'Look, someone pushed a button 3 hours ago before it engaged'. Once the private capital gets their hands on them, we immediately transition to cyberpunk no matter what else happens in the world.
Post reply on HN