That’s a silly way to look at it. Surely people can realize that something made through natural processes has more of an appeal than something made in a factory?
That said, most of the gems I’ve purchased in my life have been lab grown.
I'd argue that most people appear to prefer manufactured things these days.
No one's wearing clothes their mothers' spun because mothers' see that their children prefer the "higher quality" that is manufactured in factories.
Few choose to spend their time engaged in walking about in nature and choose instead to gaze endlessly at their factory built device that provides them content ground out in other sorts of factories. (content farms etc).
Something made in nature can be more appealing, but it seems to me that the modern preference is not at all for natural things. Hell, even in the diamonds we're talking about. No one's proposing with a natural diamond. People propose with carefully curated, carefully manipulated, and carefully presented diamonds. There's nothing natural about it, really.
The absolute favorite activity of my dog is to chase groups of turkeys that are in our yard. She’s only 15 pounds (and 15 years old now) and has never caught one, but I think being in the middle of a group of 20 turkeys all desperately trying to fly up to trees quickly is quite the experience for her.
She’s never caught one, even the younger ones. It seems like they can actually fly easier than the adults.
I walked out one morning and had a whole flock of them on the pergola on our second story deck. I was pretty surprised that they managed to get up there, three stories off the ground.
At the same time, they dont seem to be able to fly well when they panic. I let my dogs out one morning not knowing there was a turkey in their fenced area. The turkey freaked out and flew straight into the fence. Never seen a dog move as fast as my 70lb chow/cattle dog mix moved that morning.
I am guessing you live in a place where turkeys were introduced as game birds (such as California). Because in their native range turkeys are wary of humans and other preditors. That’s why turkey hunting involves calls and camouflage and patience…
To your point, if we had truly unlimited context to the point where at least that instance of a model could “learn” and have what seems like a continuous “consciousness” I think many of us would think that we’ve attained AGI.
Right now we have an incredibly smart thing with severe short term memory loss, and it’s hard for us to reconcile that as it’s so different from us.
Quite a few people were already led to believe that these models are conscious when we had a fraction of current context lengths. Right now the biggest problem is that the "session" info in form of the current conversation gets lost too quickly, but that has become largely an implementation detail. You could fit an entire life's story into modern context windows. With some clever context management, you could probably build something that feels like what you describe. If we truly had this sort of short-term to long-term memory (i.e. from prompt context to weights) system on a technical foundation, we'd probably be closer to runaway superintelligence than mere AGI that could beat most humans on most tasks.
Not a bot (although I have been accused of it, due to my activity here, and on GitHub, but I’ve been this way for longer than LLMs have been a thing. I’m retired, “on the spectrum,” and don’t participate in any other social media).
I’m currently working on a rewrite of an app that originally took two years. It’s been about three months, and I’m probably about 70% done. It’s a total “from scratch” rewrite; both client and server (two versions of each, as I also have administrative code). It’s a pretty big system, for one guy. I couldn’t do it, without the LLM.
It’s not been a cakewalk. I’ve needed to toss out large swaths of LLM-generated code, and rewrite by hand, but, for the most part, it’s been a huge help.
But I’m also not doing it in a manner that eats tokens. I just use the standard $20/month subscription as a chat. I suspect my workflow is not one that Anthropic or OpenAI really wants out there.
But I also bet that many HN accounts are bots; although I think many may be ones run by enthusiasts, not some AI cabal.
For 5 million comments like yours I haven't seen a single one with the old code vs. the new code. I understand that not all code is public that way, of course, and I don't mean to put you on the spot personally. But where are all the open source projects that now do the same with better error handling using less resources? Where are 100+ MB Electron apps reduced to more correct sizes like a few MB, or even a few dozen kB? Why aren't startup times getting slashed across the board? Why isn't RAM usage falling faster than RAM prices are increasing?
Feel free to check out my GH profile. I'm working on a closed-source app, now, but several of its component dependencies have had significant LLM work, and they are open.
Other than that, I am not boosting AI, and have absolutely zero interest in doing a bunch of work to satisfy some random Internet Guy, who can't be bothered to examine my pretty damn extensive open portfolio.
And how did any of that relate to "Showing actual improved products and features. Showing actual code. etc." ? It's the opposite, someone says "I'm sick of milk and orange juice all the time, I want some water", and you reply with nothing but offering them a cup of milk.
> random Internet Guy, who can't be bothered to examine my pretty damn extensive open portfolio.
You cannot even be bothered to examine the comment you reply to, maybe get off your high horse.
And the main part of my comment was about something in the common realm, open source software, and hard performance/quality improvements. Not wishy-washy products and features, not yet another tone deaf cool story.
Eh, whatevs. When someone interacts with me, here, even if being unpleasant, I generally check out their profile, first thing. Sometimes, it has changed my opinion of them, and of myself.
For instance, I checked out yours, and there's not much, except a whole bunch of challenging people here. I am wondering if you came here to "set us straight." I know that a lot of folks have low opinions of HN, and not all of them are wrong, but I find this place a fairly good place to hang out. Being challenged, is one of the draws, for me.
By the way, have you tried the new unhomogenized heavy cream? Good stuff!
My answer to "show your work" was "No." I am not going to go through my code, and show a bunch of supporting evidence for a casual comment, in which I have exactly zero investment. I really don't care that much what people think of me. I was just sharing my personal experience. If you guys want to write me off, then knock yourselves out.
"No" is a complete sentence. What part of "No" didn't he understand?
> My answer to "show your work" was "No." I am not going to go through my code
An interesting answer to literally "Just what kind of evidence do you suppose they could have? - Showing actual improved products and features. Showing actual code. etc."
> "No" is a complete sentence. What part of "No" didn't he understand?
See above. After pointing this out you immediately started down the path of "why didn't you looko at my profile and followed the link to my github".
Not even close. The opposite even. Read more carefully.
> For 5 million comments like yours I haven't seen a single one with the old code vs. the new code. I understand that not all code is public that way, of course, and I don't mean to put you on the spot personally. But where are all the open source projects that now do the same with better error handling using less resources?
It’s not been a cakewalk. I’ve needed to toss out large swaths of LLM-generated code, and rewrite by hand, but, for the most part, it’s been a huge help.
But your anecdote is much more balanced and more in line with my personal findings. Not like all the AI astroturfing that happens here. I like to use LLMs as well, but in a very targeted way in places where LLMs shine.
Yes, LLMs can make you much more productive. But so could assembly -> C -> Python or Rust, or switching an IDE with code completion and support for refactoring. Each step makes you more productive.
Sure, LLMs can spit out greenfields projects. But on large projects with complex requirements, you still need senior engineers to guide them, carefully review the output, and as you say throw out code and write it from scratch in a better way.
I had some friends who ended up in bits of AI psychosis. They exclaim that a swarm of agents was writing all their code, but every time I ask them to show the end-result, all they have is a pile of code that they don't understand, nor doesn't really work either. At the same time, they stopped getting any actual work done.
At any rate, somebody had a great analogy on HN recently: think of it is a vector, LLMs can significantly increase the magnitude of the vector, but you still have to make sure that the orientation of the vector is correct.
> At any rate, somebody had a great analogy on HN recently: think of it is a vector, LLMs can significantly increase the magnitude of the vector, but you still have to make sure that the orientation of the vector is correct.
> It’s not been a cakewalk. I’ve needed to toss out large swaths of LLM-generated code, and rewrite by hand, but, for the most part, it’s been a huge help.
Same here :)
> not some AI cabal.
There are enough enthusiasts to make it feel like one. Also an unhealthy doze of marketers, people buying into hype, AI psychosis etc.
> There are enough enthusiasts to make it feel like one. Also an unhealthy doze of marketers, people buying into hype, AI psychosis etc.
There's absolutely no question that AI is a real thing, and that there's going to be a lot of money made, so there's a bunch of folks with commercial interest in pushing it.
It's just different from crypto. This has actual real-world utility for just about everyone. I am increasingly hearing people say "Ask ChatGPT," where they used to say "Google It" (where they used to say "Look it Up at the Library").
I had to convert a build pipeline from just one linux distro to multiple and then get arm64 going. Not the most difficult thing in the world but quite annoying when there's 100 binaries and a complex dep tree with lots of moving pieces. Anyway AI for sure increased project cadence by at least 2x. Not sure why there's so much denial in these threads.
I can also claim a bunch of things. If you manage to read the comment I was originally replying to, and my reply:
--- start quote ---
- Just what kind of evidence do you suppose they could have?
- Showing actual improved products and features. Showing actual code. etc.
--- end quote ---
Note how you provided neither. It's just claims.
> Anyway AI for sure increased project cadence by at least 2x.
As in: you claim this. Also, no one denies that you can ship a lot of code much faster with AI. However, somehow, very little actual evidence of grandiose claims (see farther up in the context) besides anecdotal "I'm so faster and features are being shipped left and right".
Oh, I certainly believe this. LLMs tend to be quite good at the: I'll give you a well-designed example, now extrapolate to other cases-cases.
I think it's a great example of using LLMs effectively. In the end becoming more productive is understanding where LLMs work great and where they fail miserably.
But it is a step similar to, say going from assembly to a higher-level programming language, not the silver bullet that AI astroturfers like you to believe (fire all the programmers to buy more tokens!)
To be fair, my assumption on that line was always that they were saying they're balancing out the overall news world. Still dirty, of course. Regardless, everyone knows how they lean.
What grinds my gears is NPR. Every member drive they explicitly talk about how their coverage is fair and unbiased, which is way more egregious than some tagline. As far as having a bias and not owning up to it, I think they're the worst offenders right now.
Without looking it up I’m guessing I’ve had my P1S for 3 or more years and I’ve never had a single issue with it, to the point where I wonder what issues your friends have?
You can use their slicer (which works well!). If you don’t want to, you can use one that sends through their Bambu Connect software, which Orca Slicer doesn’t want to support for…reasons. Or you can use it in LAN mode. Or you can just transfer the gcode via an SD card or flash drive like ye olde days.
Despite the tone of the other reply to your question, they are absolutely the easiest printers to work with. I don’t love their new multicolor solution for how slow it is compared to other options, but that would be the only real fault with their newest line.
That said, most of the gems I’ve purchased in my life have been lab grown.
reply