←back to thread

LLM Inevitabilism

(tomrenner.com)
1616 points SwoopsFromAbove | 5 comments | | HN request time: 0s | source
Show context
mg ◴[] No.44568158[source]
In the 90s a friend told me about the internet. And that he knows someone who is in a university and has access to it and can show us. An hour later, we were sitting in front of a computer in that university and watched his friend surfing the web. Clicking on links, receiving pages of text. Faster than one could read. In a nice layout. Even with images. And links to other pages. We were shocked. No printing, no shipping, no waiting. This was the future. It was inevitable.

Yesterday I wanted to rewrite a program to use a large library that would have required me to dive deep down into the documentation or read its code to tackle my use case. As a first try, I just copy+pasted the whole library and my whole program into GPT 4.1 and told it to rewrite it using the library. It succeeded at the first attempt. The rewrite itself was small enough that I could read all code changes in 15 minutes and make a few stylistic changes. Done. Hours of time saved. This is the future. It is inevitable.

PS: Most replies seem to compare my experience to experiences that the responders have with agentic coding, where the developer is iteratively changing the code by chatting with an LLM. I am not doing that. I use a "One prompt one file. No code edits." approach, which I describe here:

https://www.gibney.org/prompt_coding

replies(58): >>44568182 #>>44568188 #>>44568190 #>>44568192 #>>44568320 #>>44568350 #>>44568360 #>>44568380 #>>44568449 #>>44568468 #>>44568473 #>>44568515 #>>44568537 #>>44568578 #>>44568699 #>>44568746 #>>44568760 #>>44568767 #>>44568791 #>>44568805 #>>44568823 #>>44568844 #>>44568871 #>>44568887 #>>44568901 #>>44568927 #>>44569007 #>>44569010 #>>44569128 #>>44569134 #>>44569145 #>>44569203 #>>44569303 #>>44569320 #>>44569347 #>>44569391 #>>44569396 #>>44569574 #>>44569581 #>>44569584 #>>44569621 #>>44569732 #>>44569761 #>>44569803 #>>44569903 #>>44570005 #>>44570024 #>>44570069 #>>44570120 #>>44570129 #>>44570365 #>>44570482 #>>44570537 #>>44570585 #>>44570642 #>>44570674 #>>44572113 #>>44574176 #
1. throwaway328 ◴[] No.44569761[source]
And here's a list of stuff I've seen or that the non-computer-experts tell me they're doing with it, since the last month or two when suddenly even people who were against it are accepting it, along with people who'd never heard of it suddenly using it:

- getting the do-re-mi notes for "twinkle twinkle little star" for the piano, just written out with no rhythm or audio anything

- writing a groom's wedding speech ("the first draft", he said, but I doubt it'll be edited much)

- splitting a list of ten names into two groups, to get two teams for indoor soccer (I know, I know... The tone was one of amazement and being impressed, I shit you not. One fellow used to bring a little bag with the same amount of yellow and red lego bricks and we'd pick one from the bag)

- in a workplace, a superior added a bell that gets triggered when a door opens. The superior left, and one employee went straight to ask chatgpt how to turn off the bell, and went straight to fiddling with the alarm after the very quickest skim of the response (and got nowhere, then gave up)

- and a smattering of sort of "self-help" or "psychology lite" stuff which you'll have to take my word on because it's personal stuff, but as you'd expect: "how to deal with a coworker who doesn't respect me in xyz manner", "how to get a 6-pack", "how to be taller", "how to get in to day-trading"

- and a good dose of "news"-related stuff like matters of actual law, or contentious geopolitical topics with very distinct on-the-ground possiblities and mountains of propaganda and spin everywhere, about say the Ukraine war or Gaza. E.g., one friend asked for specific numbers of deaths "on both sides" in Gaza and then told me (I shit you not!) he'd "ran the numbers" on the conflict during his research

Anyway. All that to say not that these people are silly or bad or wrong or anything, but to say - the internet was new! This isn't. When you were brought to see that computer in the university, you were seeing something genuinely amazingly new.

New forms of communcation would open up, new forms of expression, and a whole new competitive space for the kids of the wealthy to see who could contort these new technologies to their will and come out on top dominating the space.

With LLMs, we're only getting the last one there. There's nothing new, in the same profound sense as what the internet brought us. The internet offered a level playing field, to those brave enough to slog through the difficulties of getting set up.

Put differently - LLMs are similar to the internet, if and only if we accept that humans generally are idiots who can't understand their tools and the best we can hope for is that they get faster slop-generating machines. The internet didn't start like that, but it's where it ended up.

And that's LLM's starting point, it's their cultural and logical heart. I think a large number of technologists have internalised these assumptions about humans and technology, and are simply not aware of it, it's the air they breathe.

Put differently again - if the tech industry has gotten so blind that LLMs are what it considers the next internet-sized-idea, and the only possible future, well, it's an industry that's in a myopic and inhumane rut. We'll go from a world where people click and scroll on their devices for entertainment, fundamentally detached from each other and fundamentally disempowered, to a world where people click and scroll on their devices for entertainment, detached and disempowered. How noble a vision, how revolutionary.

So to sum up, in one sense you're correct - it looks like it's going to "take over", and that that's "inevitable". In another sense, LLMs are absolutely wildly different, as this time we're starting off treating the average user like a complete idiot, in fact assuming that we can never do better, and that considering the possibility is childish nonsense.

replies(2): >>44571488 #>>44576000 #
2. otabdeveloper4 ◴[] No.44571488[source]
The big unspoken deal is local, offline LLMs, especially if (when) easy plug-and-play finetuning becomes the norm.

Suddenly the pendulum swings back the personal computing and you can have your own offline, curated mini-google.

Imagine when a thousand purpose-trained open-weight LLMs appear, for everything imaginable. (E.g., if you want your own private Magic the Gathering robot you now can.)

replies(1): >>44575282 #
3. cruffle_duffle ◴[] No.44575282[source]
Dude, local offline LLM's is what will be transformative. I absolutely hate that these LLM's are designed, built, governed and ran by massive tech oligarchs. There is absolutely no level playing field. It's the same "rich dudes" just getting richer.

Wake me up when you can get offline, open sourced, "peer to peer trained" LLM's that had nothing to do with the "big boys". Wake me up when those things manage to piss off "the establishment".

The current crop are all establishment. They've been neutered and approved by expensive laywers.

replies(1): >>44577060 #
4. namcheapisdumb ◴[] No.44576000[source]
Most of these can be accomplished by 2B models running on top-of-the-line consumer phones. That's the damning stuff! The real pale horse is HBR reporting that most people use chatbots to chat. Reasoning models are pointless for most people. LLM initiatives are seeking its wide adoption, at the same time, the business models collapse if they become too commoditized. The entire industry undermines itself.
5. namcheapisdumb ◴[] No.44577060{3}[source]
Don't worry, the writing is on the wall. Unless Trump cracks down on these novel chinese models, like Kimi, there's just no way for these companies to ever turn green. Either they swallow their losses and move on, or transform silicon valley into some sort of ultra-restricted state monopoly