This is a terrible example. Cars qualitatively opened up something that humans could never do before. Same with the computer. What about LLMs?
LLMs just replace what you were already doing, but don’t qualitatively open up all that much in the frontier. A million lines of LLM slop? Arguably LLMs close the frontier, rely too much on it and your ability to think hard and long ends up getting crushed.
> LLMs just replace what you were already doing
IF you were already doing it. But LLMs also allow people who never were doing something to now do that thing as well. In the same way Excel macros enabled some people to run whole businesses on a spreadsheet without hiring a "real" programmer, LLMs will enable some people to run a whole business on "real" applications without hiring a "real" programmer. And yeah, some of those excel macros were a nightmare, one of the many jobs I've taken over the years was a contract job converting the business logic that used to be in excel macros and later vba script to a "real" language and application. But the business that hired me only had the sort of money they needed to hire me because they had a successful business running off that horrible spaghetti mess. I wonder what businesses will be hiring "real" programmers when their founders hit the wall of what they can accomplish on their own with AI.
> LLMs will enable some people to run a whole business on "real" applications without hiring a "real" programmer
SaaS, and before that a lot of shrink wrap software, has been allowing this for several decades now.
So what? Just because you had one way to do something before, that means having a different way of doing it is useless? We have the C language, and it’s been letting people write applications for decades. Does that mean there’s no value in other programming languages?
> that means having a different way of doing it is useless?
That’s a ridiculous jump to a totally extreme conclusion that’s not supported by any message above. What does one even answer to this without getting sucked into a pointless argument?
You don't understand how a comment chain that went from:
> Cars qualitatively opened up something that humans could never do before. > Same with the computer. What about LLMs? > LLMs just replace what you were already doing
to:
> LLMs also allow people who never were doing something to now do that thing > as well. > ... > LLMs will enable some people to run a whole business on "real" applications > without hiring a "real" programmer.
to:
> SaaS, and before that a lot of shrink wrap software, has been allowing this > for several decades now.
leads one to conclude you seem to feel that if something doesn't let you do something you weren't doing before, that it has no value? Because I'm otherwise unclear about what your comment was then supposed to express. The entirety of your reply was to express that the thing an LLM will enable was something that was also enabled by something else. Given the topic of conversation, it seems reasonable to conclude that expression was in the context of talking about the value of AI tools with respect to how they remove friction. If that wasn't the point, then if you could perhaps use more words to state your point directly instead of leaving it to the reader to infer that your point was tangential to the main thrust of the discussion.
I don't think it's as much as opening up what we _could_ do, it's opening up time so that more can be done.
Slop is too generic of a complaint. Do LLMs write code exactly how I would? No. Is my way right? Probably not. But what LLMs are great at are pre-emptively thinking about edge cases that I forget when writing the code. I'm surprised as how many fewer weird bugs come back to me now after a release. Should I have thought of them? Yes probably, but I don't and they do. So in practice, I do think LLMs write better code than me (if you look at results not the actual code itself).
But more importantly is that it allows me to take on projects that I never would have. Some of them are personal projects, some of them are work projects. I often didn't even consider a project because the time it would take _me_ to do it would not be worth the value. But now that _I_ don't have to do it, I can do more things. At the end of the workday last week, I kicked off a project that I've wanted to do for the last 5 years but could never justify writing it. The LLM crunched for 36 hours, I checked in on it this morning and it's ready for me to deploy next week when I get in the office. Something that would have taken me probably 9 months of focused effort. It would never have been worth 9 months of engineering time, but it sure is worth the 36 hours of background work while I was doing something else productive.
I guess it's not a quantum leap in the frontier of what _can_ be done, but it is a leap in what _will_ be done because the cost/benefit parameters change.
>So in practice, I do think LLMs write better code than me (if you look at results not the actual code itself).
"correct" code is not "good" code. We see this quite a bit with the analogue of AI-generated text. Having all your grammar and spelling correct does not make for a more compelling read.
like, a correct, tested class doesn't necessarily stand up to integrating into a million line legacy code. Executives don't understand this, but this is the value being shedded as they try their 4th or 5th wave of attempts to take the programmer out of programming.
Cars qualitatively opened up something that humans could never do before.
Not really, we explored almost the entire world on foot and horseback. Cars and trains "just" make it easier.
A million lines of LLM slop?
As they say, quantity has a quality all its own. Over the last few years I was working on a side project that needed a lot of tedious logic of the form "if X then Y, unless both A and B, in which case either P or Q depending on the phase of the moon". I gave Claude the problem, it independently came up with a nearly identical architecture to what I had already built, then tore through all the scenarios and produced thousands of unit tests along with validation and fuzzing scripts. The code isn't terribly elegant, and was never going to be. But it works and it's shipping, and that wouldn't have happened anytime soon without AI.