"Other people were wrong about something else so that invalidates your argument"
Why are half the replies like this?
Why are half the replies like this?
I’m open to all possibilities. There might be a near term blocker to improvement. There might be an impending architectural change that achieves AGI. Strong opinions for one or the other with no extremely robust proof are a mistake.
Call me if you find a good reason. I still have not.
I always forget the name of that law but... it's hard to make somebody understand something if their salary depends on them not understanding it.
Why do you think your observable reality is the only one, and the correct one at that? Looking at your mindset, as well as the objections to the contrary (and their belief that they're correct), the truth is likely somewhere in-between the two extremes.
The funny thing about the rest of your comment is that I'm in full agreement with you but somehow you decided that I'm an extremist. I'm not. I'm simply tired of people who make zero effort to prove their hypothesis and just call me arrogant or old / stuck in my ways, again with zero demonstration how LLMs "revolutionize" programming _exactly_.
And you have made more effort in that direction than most people I discussed this with, by the way. Thanks for that.
Still, I don't think it's a hypothesis that most are operating under, but a lived experience that either it works for them or it does not. Just the other day I used ChatGPT to write me a program to split a file into chunks along a delimiter. Something I could absolutely do, in at least a half-dozen languages, but writing that program myself would have distracted me from the actual task at hand, so I had the LLM do it. It's a trivial script, but the point is I didn't have to break my concentration on the other task to get that done. Again, I absolutely could have done it myself, but that would have been a total distraction. https://chatgpt.com/share/67655615-cc44-8009-88c3-5a241d083a...
On a side project I'm working on, I said "now add a button to reset the camera view" (literally, aider can take voice input). we're not quite at the scene from star trek where scottie talks into the mac to try and make transparnet aluminum, but we're not that far off! The LLM went and added the button, wired it into a function call that called into the rendering engine and reset the view. Again, I very much could have done that myself, but it would have taken me longer just to flip through the files involved and type out the same thing. It's not just the time saved, which, I didn't have a stopwatch and a screen recorder, but apart from the time, it's not having to drop my thinking into that frame of reference, so I can think more deeply about the other problems to be solved. Sort of why ceo isn't an IC and why IC's aren't supposed to manage.
Detractors will point out that it must be a toy program, and that it won't work on a million line code base, and maybe it won't, but there's just so much that LLMs can do, as they exist now, that it's not hyperbole to say it's redefined programming, for those specific use cases. But where that use case is "build a web app", I don't know about you, but I use a lot of web apps these days.
> Detractors will point out that it must be a toy program, and that it won't work on a million line code base, and maybe it won't
You might call me a detractor. I think of myself as being informed and feeling the need to point out where do LLMs end and programmers begin because apparently even on an educated forum like HN people shill and exaggerate all the time. That was the reason for most of my comments in this thread.
Honestly, if you don’t think it works for you, that’s fine with me. I just feel the dismissive attitude is weird since it’s so incredibly useful to me.
If you can give examples of incredible usefulness then that can advance the discussion. Otherwise it's just us trying to out-shout each other, and I'm not interested in that.
However, that's absolutely not clear today.
I have spent 24 years coding without LLMs, cannot fathom now spending more than a day without it…
If you have scrutable and interesting examples, I am willing to look them up and try to apply them to my work.
You're essentially saying "the programming to do basic arithmetic and physics took only a year" as if that's remotely impressive compared to the complexity of something like Microsoft Teams. Simultaneous editing of a document by itself is more complicated than anything an Apollo program had to do
I also have no idea what your comment here had to do with LLMs, will you elaborate?
I have built web services used by many Fortune 100 companies, built and maintained and operated them for many years.
But I'm not doing that anymore. Now I'm working on my own, building lots of prototypes and proof-of-concepts. For that I've founding LLMs to be extremely helpful and time-saving. Who the hell cares if it's not maintainable for years? I'll likely be throwing it out anyway. The point is not to build a maintainable system, it's to see if the system is worth maintaining at all.
Are there software engineers who will not find LLMs helpful? Absolutely. Are there software engineers who will find LLMs extremely helpful? Absolutely.
Both can exist at the same time.
What are the likely impacts over the next 1, 5, 10, 20 years. People getting into development now have the most incredible technology to help them skill up, but also more risk than we had in decades past. There's a continuum of impact and it's not 0 or 100%, and it's not immediate.
What I consider inevitable: humans will keep trying to automate anything that looks repeatable. As long as there is a good chance of financial gain from adding automation, we'll try it. Coding is now potentially at risk of increasing automation, with wildcards on "how much" and "what will the impact be". I'm extremely happy to have nuanced discussions, but I balk at both extremes of "LLMs can scale to hard AGI, give up now" and "we're safe forever". We need shorthand for our assumptions and beliefs so we can discuss differences on the finer points without fixating on obviously incorrect straw men. (The latter aimed at the general tone of these discussions, not your comment.)
My issue is with people claiming LLMs are undoubtedly going to remove programming as a profession. LLMs work fine for one-off code -- when they don't make mistakes even there, that is. They don't work for a lot of other areas, like code you have to iterate on multiple times because the outer world and the business requirements keep changing.
Works for you? Good! Use it, get more productive, you'll only get applause for me. But my work does not involve one-off code and for me LLMs are not impressive because I had to rewrite their code (and to eye-ball it for bugs) multiple times.
"Right tool for the job" and all.
But what you'll probably find is that people that are skilled communicators are currently getting a decent productivity boost from LLMs, and I suspect that the difference between many that are bullish vs bearish is quite likely coming down to ability to structure and communicate thoughts effectively.
Personally, I've found AI to be a large productivity boost - especially once I've put certain patterns and structure into code. It's then like hitting the N2O button on the keyboard.
Sure, there are people making toy apps using LLMs that are going to quickly become a unmaintainable mess, but don't be too quick to assume that LLMs aren't already making an impact within production systems. I know from experience that they are.
Strange perspective. I found LLMs lacking in less popular programming languages, for example. It's mostly down to statistics.
I agree that being able to communicate well with an LLM gives you more results. It's a productivity enabler of sorts. It is not a game changer however.
> don't be too quick to assume that LLMs aren't already making an impact within production systems. I know from experience that they are.
OK, I am open to proof. But people are just saying it and leaving the claims hanging.