One thing jumps out about the person who noticed the AI was wrong on things they were familiar with. It's like when ELon Musk talks about rockets. I don't know about rockets so I take his word for it. When Elon Must talked about software it was obvious he has no idea what he's doing. So when the AI generates something I know nothing about, it looks productive but when it's generating things for which I'm familiar I know its full of shit.
> So when the AI generates something I know nothing about, it looks productive but when it's generating things for which I'm familiar I know its full of shit.
This is why when you hear people talk about how great it is at producing X, our takeaway should be "this person is not an expert at X, and their opinions can be disregarded"
They are telling on themselves that they are not experts at the thing they think the AI is doing a great job at
"This is why when you hear people talk about how terrible it is at producing X, our takeaway should be "this person either hasn't tried to use it in good faith, and their opinions can be disregarded"
I'm playing devil's advocate somewhat here but it often seem like that there's a bunch of people on both sides using hella motivated reasoning because they have very strong feelings that developed early on in their exposure to AI.
AI is both terrible and wonderful. It's useless and some things and impressive at others. It will ruin whole sectors of the economy and upturn lives. It will get better and it is getting better so any limitations you currently observe are probably termporary. The net benefit for humanity may turn out to be positive or negative - it's too early to tell.
> AI is both terrible and wonderful. It's useless and some things and impressive at others
That's kind of my problem. I am saying that it mostly only appears impressive to people who don't know better
When people do know better it comes up short consistently
Most of the pro AI people I see are bullish about it on things they have no idea about, like non-technical CEOs insisting that it can create good code
> When people do know better it comes up short consistently
I disagree with that part and I don't think this opinion can be sustained by anyone using it with any regularity in good faith
People can argue whether it's 70/30 or 30/70 or what domains it's more useful in than others but you are overstating the negative.
Have you considered that it's actually impressive in some areas that are outside of your interest or concern?
Could be, but why would I trust that when it's clearly so bad at the things I am good at?
At this point I just want to restate my hypothesis that you haven't or aren't using it in good faith - or you haven't used it much at all.
Two can play that game
My hypothesis is that you are invested in the success of AI products somehow, financially or emotionally, and that leads you to be blind to their shortcomings
You keep using them whenever possible because you want them to be useful even though in reality their usefulness is really iffy
So - we are at an impasse. Both suspect the other of motivated reasoning and an internal bias that distorts their ability for rational thinking.
It's entirely possible we're both irrational to some degree. But that's irrelevant to answering the question at hand.
Do you claim you are using it regularly and in good faith - enough to honestly form a reliable view on its utility?
I would claim that I am using it in such a way. It would take more effort than I'm prepared to put in to provide evidence of this but please - ask away.
(for the record - I have no financial or professional involvement directly with AI. I simply find the technology fascinating and I use it daily - both playfully and for it's practical utility)
I think I have used it in good faith. A few months ago I was part of a small team at my company tasked to evaluate AI solutions like copilot to see if they are useful to us and could speed development and such
For a couple of week tryout period I tried to use it in my daily workflow pretty heavily. I came away with the impression that it is a neat toy, but not really ready to be a full time tool for me. The other evaluators agreed and our recommendation to our leadership was "This is not really ready for prime time and while it is impressive it probably isn't really worth the cost"
Anyways fast forward and we're getting AI usage OKRs now, being pushed down on us by non-technical leadership, and what I call "formerly technical" leadership. People who did tech 20 years ago but really don't know what working modern tech is like since they've been in management for too long
So yes. I'm definitely negatively biased, and I'm fine to admit that. I absolutely resent having this stuff forced down on me from leaders that are buying the hype despite being told it is probably not ready to be a daily driver
And I'm seeing the hype spreading through the company, being told by junior devs how amazing it is when I am still iffy on their abilities.
And the absolute worst is when I build a cool proof of concept in an afternoon and everyone is like "wow, AI let you do that so fast now!" and I'm like no, this is just what a good developer can build quickly
So yeah, I'm pretty negative on AI right now. I can still admit the tech itself is impressive, amazing even, and there is no doubt in my mind I could probably find some use for it daily
But I think it is going to be a disaster, because people cannot be trusted to use it responsibly
The social impact is going to be absolutely catastrophic. In some ways it already is
Edit: I am also not really sure why I am supposed to be enthusiastic about technology that business leaders are fairly transparently hoping will make my skillset redundant or at best will make me more productive but I will never realistically see a single extra dollar from the increased productivity
This makes a lot of sense. But to be honest it feels more like a story about the pathology of hierarchical organisations than anything about AI.
I mostly work solo. I use AI when it's either a) interesting or b); useful. Our experiences are very different and it's no wonder our emotional responses are also very different.
Fair enough
I'm envious that you work solo. I think that would change my perspective on a lot of things
Thanks for the good faith discussion, anyways
> The net benefit for humanity may turn out to be positive or negative - it's too early to tell.
It's just a tool, but it is unfortunately a tool that is currently dominated by large-sized corporations, to serve Capitalism. So it's definitely going to be a net-negative.
Contrast that to something like 3D printing, which has most visibly benefited small companies and individual users.
Like many things (general purpose computing, the internet) we can carve out our own space once something is released into the public sphere so I don't think Capitalism has the iron grip on this that you're hypothesising. In recent memory I think it's mainly social media where the corporations have mostly succeeded in keeping a firm hold on things and where it remains hard for users to subvert their aims. And that's largely because of the failure of decentralized social media to grow to a mass audience.
I think AI is different. "Good enough" models are already available under generous licenses, fine-tuning and even training is within the reach of groups of volunteers etc etc
Sounds like the Gell-Mann amnesia effect.