You know, it's funny. Your comment made me realize something about LLMs:
There's a famous line in Hesiod's Theogony. It appears early in the poem during Hesiod's encounter with the Muses on the slopes of Mt. Helicon, when they apparently gave him the gift of song. At this point in his narrative of the encounter, the Muses have just ridiculed shepherds like him ("mere bellies"), and then, while bragging about their great Zeus-given powers -- "we see things that were, things that are, and things that will be" -- they say "we know how to tell lies like the truth; we also know how to say things that are true, when we want to."
This is the ancient equivalent of my present-day encounters with the linguistic output of LLMs: what LLMs produce, when they produce language, isn't true or false; it just gives the appearance or truth or falsity -- and sometimes that appearance happens to overlap with statements that would be true or false if they'd been uttered by something with an internal life and a capacity for reasoning.
LLMs' linguistic output can have a weird, disorienting, uncanny-valley effect though. It gives us all the cues, signals, and evidence that normally our brains can reliably, correctly identify as markers of reasoning and thought -- but all the signals and cues are false and all the evidence is faked, and recognizibg the illusion can be a really challenging battle against oneself, because the illusion is just too convincing.
LLMs basically hijack automatic heuristics and cognitive processes that we can't turn off. As a result, it can be incredibly challenging even to recognize that an LLM-generated sentence that has all the cues of sense has no actual sense at all. The output may have the irresistibly convincing appearance of sense, as it would if it were uttered by a human being, but on closer inspection it turns out to be completely incoherent. And that inspection isn't automatic or always easy. It can be really challenging, requiring us to fight an uphill battle against our own brains.
Hesiod's expression "lies like the truth" captures this for me perfectly.
I agree with the person you're answering. LLM-assisted coding is like reading a foreign language with a facing translation: most students who do this will make the mistake of thinking they've translated and understood the original text. They haven't. People are abysmal at maintaining an accurate mental accounting of attribution, authorship, and ownership.
I disagree. When you add an abstraction layer, the user of that layer continues to write code. That's not the case when people rely heavily on LLMs. They're at best reading and tweaking the model's output.
That's not the only way to use an LLM. One can instead write a piece of code and then ask the tool for analysis, but that's not the scenario that people like me are criticizing or concerned about -- and it's not how most people imagine LLMs will be used in the future, if models and tools continue to improve. People are predicting that the models will write the software. That's what people like me and the person I agreed with are criticizing and concerned about.
I'm uncomfortable with the idea not because it's outside of my area of comfort but because people don't understand code they read the way they understand code they write. Writing the code familiarizes the writer with the problem space (the pitfalls, for instance). When you haven't written it, and you've instead just read it, then you haven't worked through the problems. You don't know the problem space or the reasons for the choices that the author made.
To put this another way: you can learn to read a language or understand it by ear without learning to speak it. The skills are related, but they're separate. In turn, people acquire and develop the skills they practice: you don't learn to speak by reading. Junior engineers and young people who learn to code with AI, and don't write code themselves, will learn, in essence, how to read but not how to write or 'speak;' they'll learn how to talk to the AI models, and maybe how to read code, but not how to write software.
I can't see the original post because my browser settings break Twitter (I also haven't liked much of Karpathy's output), but I agree. I call this style of software development 'meeting-based programming,' because that seems to be the mental model that the designers of the tools are pursuing. This probably explains, in part, why c-suite/MBA types are so excited about the tools: meetings are how they think and work.
In a way LLMs/chatbots and 'agents' are just the latest phase of a trend that the internet has been encouraging for decades: the elimination of mental privacy. I don't mean 'privacy' in an everyday sense -- i.e. things I keep to myself and don't share. I mean 'privacy' in a more basic sense: private experience -- sitting by oneself; having a mental space that doesn't include anybody else; simply spending time with one's own thoughts.
The internet encourages us to direct our thoughts and questions outward: look things up; find out what others have said; go to wikipedia; etc. This is, I think, horribly corrosive to the very essence of being a thinking, sentient being. It's also unsurprising, I guess. Humans are social animals. We're going to find ourselves easily seduced by anything that lets us replace private experience with social experience. I suppose it was only a matter of time until someone did this with programming tools, too.
(FYI: you can easily bypass the awful logged out view by replacing x.com with xcancel.com, I use a URL Autoredirector rule to do it automatically in Chromium browsers)
Also 6.8 is at least 2 years out, so there's still time to work the remaining issues out. As far as I know only speech input remains a major problem, so hopefully they'll figure that out.
I'm sure the good people at KDE wouldn't mind delaying 6.8 if more work is needed for something as important as accessibility. And let's be real: if you don't need voice input then KDE on Wayland is ready.
Wayland very first release was 17 years ago in 2008 and QT didn't support it until 2015. xdg-desktop-portal first stable release was in 2018 and PipeWire in 2023.
I thought we had peaked with systemd when it came to FUD here but Wayland might give it a run for its money.
No one is "forcing" anything on you, and I'm finding this argument increasingly disingenuous. If you prefer the current versions then you're entirely free to keep them.
But you do not get to demand that future versions are only ever implemented your way. If that's what you want, fork the project or pay someone to do it for you. Acting entitled about the work of volunteers who are sharing it with you for free is not a good look.
Mine still works as well as expected after 17 years, 5-6 of which it spent with heavy daily use, another 2-3 with light use, only occasionally afterwards, and overall a lot of travel and airports. I could disassemble and reassemble it to the last screw easily, no special tools besides a screwdriver, no glue, upgradeable RAM and storage. Actually my one major complaint is Lenovo's use of whitelisting for wireless cards.
But I wouldn't pay $1300+ to bring it up to speed. The batteries are done, the screen is small and the backlight is yellowed and dimming. That laptop would need a lot more love to make it fully usable as a daily driver so I'd rather keep it as it is, as a memory.
It's small, sturdy, maintainable, and aesthetically pleasing. And one can still get (original) parts. Throw in enthusiast projects like this and you can have your own "Laptop of Theseus".
And that’s about it, I’d say! I find that everything else is really, really bad. It creaks, it wobbles, it warps, and it did so from day 1. The fan is loud and kicks in quite early. Well maybe the X200 isn’t as bad, but the X220 certainly is. And even after 14 years, it still smells when it gets hot.
Sorry for the rant. I really want to love it, but I just can’t.
Quality also went down while with later models - back in 2014 I was laptop shopping, based on the X2xx series reputation I tested an X240 and it was crap, even the keyboard was super bad, I ended up getting a Dell xps13 whose keyboard was miles better and it still works today.
Well "use it" is a bit of a stretch. I’m a bit of a device hoarder. It's one of my experimentation platforms for Linux stuff, currently running Fedora Kinoite (with Universal Blue).
My daily driver (of sorts, don’t really need a laptop anymore) is a MacBook Pro Late 2013, with NixOS. It’s so much better in every regard, it’s not even funny. It also still has its original battery.
Not sure, but I bought used x201 in 2014 and it died few months ago (faulty charging port, weak monitor joints). Replaced by P14s gen2 with AMD. Of course it is better in every aspect, except one disc port and overall durability.
Likewise. I don't always do this, but for problems that cost me much time or effort, I like to try to make sure that, if I wanted to reproduce a bug or problem, I'd know exactly how to write it.
Writing and understanding working correct software is, it turns out, a rather different skill from that of writing and understanding broken (or confusing) software. I'd also wager good money that the latter skill directly builds the former.
reply