41–50 of 54 posts
If you try to talk about automating SE and you aren’t clearly explaining how people who know nothing about engineering will directly interact with this automation to get what they need, then you aren’t saying anything.
Even with vibe coding there is a certain skillset involved.
Earlier quoted context omitted.
> And there's still someone who says to you: "best I can do is only 1% - 2%" 1% of the global economy seems fine to me. If you actually believe in their vision, money is going to be worthless anyway.
Isn't it the other way around? The productivity gains would increase overall wealth so you would get more for your money, meaning money would actually be worth more. Or do you mean it in the sense that everyone would already have everything they could ever want so the net utility of additional money would be 0?
It is possible to increase productivity while also centralizing that profit. That's what communists would call "exploitation of labor"
Regardless, software dev I consider way more dimensional (eg nuanced) than Dota 2, even if a lot of patterns recur on a smaller scale in the code itself. If they weren’t able to crack Dota 2, why should I believe that software eng is just around the corner?
What if I told you that's not the key question, and the "more data" approach has obviously and publicly hit a wall that requires causal reasoning to move past?
Earlier quoted context omitted.
If your car can't be owned and operated by someone with zero knowledge of how cars work you've failed. And yet to this day every car owner is aware that their lack of any such knowledge would/does cost them dearly at the mechanic and at the dealership too. And of course mechanics and car salesmen love to see such clients coming their way.
Yea, I mostly agree with this. Consumers tend to have some sort of understanding of the car market. If you can't cater to them, you've failed. Correspondingly, it's very easy to imagine someone disgruntled for having to deal with code (i imagine hotwiring a car in the context you provided). People know their lane; you can't force them to change it.
> The key question now is: what data do we need, exactly? What if I told you that's not the key question, and the "more data" approach has obviously and publicly hit a wall that requires causal reasoning to move past?
OpenAI spent several years prior to ChatGPT getting AI to play Dota 2 well. They got some good results out of that, but it was a subset of the game: only a handful or two of characters. I’m not sure why they stopped; maybe that was when they pivoted to something more general? Regardless, software dev I consider way more dimensional (eg nuanced) than Dota 2, even if a lot of patterns recur on a smaller scale in the co…
Earlier quoted context omitted.
Pretty sure that however humans crossed it, it wasn't just "a data problem".
what makes you sure about that?
What experienced software engineers have is a sense of taste - this looks like good code and/or design, that doesn't. But they don't have data; they have, at best, a couple of anecdotes. It's more a sense of "that was harder to work with than it should have been; that approach seems to have drawbacks". But you only get a few examples of that in a career.
And there are very few outfits compiling usable data that could shape the approaches that software engineers use.
So I don't think how humans got there was primarily data.