I don't believe there will ever be any artificial intelligence, not with Markov chains (next token prediction), not otherwise. Especially not now when the current ML hype is already winding down. And yes this is a matter of belief since I don't think any science precludes agi from existing nor is there any reason to be sure it could someday materialize. I honestly would rather believe societal collapse hits us before agi can even be theorized.
AI was here in the 1970's too for that matter, in the form of expert systems. "AI" is the label that perennially gets applied to whatever current technology does something that was previously considered similar to human intelligence, then later on gets removed and applied to something new.
You'll know were making progress towards AGI when LLMs start being called LLMs again, and something new starts being called AI.
Kind of, but I wouldn't exactly put it like that since AI has never meant anything more than automated intelligence/decision making of some sort. The bar isn't moving, just this almost meaningless label is just forever getting slapped on the latest shiny new thing.
You could legitimately call a thermostat "AI". Expert systems were previously called AI. Today it's Large Language Models. Tomorrow it'll be something else.
I think the issue is that technologists use the term AI to mean one thing, which is something we've had since the 60s with lisp, forth, and prolog. And the average person uses AI to mean something more like Data from Star Trek, which we don't have and may or may not have this century. We're all talking past each other.
There is no rigorous definition of intelligence, let alone artificial intelligence. What you're referring to is people simply not knowing what they're talking about.
More to the point, there has always been a cottage industry in predicting an amazing future, just around the corner. 'AGI' is just the latest incarnation.
"quit my FAANG job" as in they simultaneously worked in Facebook, Amazon, Apple, Nvidia and Google? Or did the op work at Netflix and is too ashamed to admit that :P
Source…? I see this claim thrown around so much without any backing and my experience with mac os (and other darwin variants) is just terrible. My previous portable was a 2015 mac book pro with 16 (!) gigs of ram (mind-blowing for 2026 standards, I know) and the os just became terribly and unbearably sluggish with all the useless updates (that nb mostly removed features, e.g. ClearType was wiped off the face of the Earth in back in mac os Mojave) caking on, at some point I just gave up on trying to fix it with continuously reinstalling and balancing it. My current portable has quarter the amount of workmem and is just incredibly snappy esp. compared to that mac book. And I don't even use portables for anything too heavy, I have an actual PC for that.
The stress tests with multiple Youtube windows open simultaneously, editing 4K video, and doing heavy duty image editing tasks, in some cases doing stuff like that all at the same time have been very impressive. It does have to fall back on swap, but even then seems to soldier on really well considering. Are you better off with more memory? Absolutely, but it still seems perfectly capable of managing even many low to medium duty pro workloads if you don't mind a performance hit.
MacOS has had memory compression since Mavericks in 2013, but the M series chips also introduced a wider memory bus that makes for faster swap, and hardware accelerated memory compression/decompression.
A lot of this tech is inherited from the work done on iOS and the A-series SOCs to maximise performance and minimise resource utilisation for the phones. And of course the Neo uses an A-series phone SOC.
My macbook was UMA and memory compression has been used everywhere for dozens of years now. Symbian had memory compression. Is it just apple users catching up to what a snappy computer actually feels like…? (Doubt, since as I said I used Apple before) The article doesn't address the world outside apple either, and Darwin is objectively slow by its obsolete architectural design, down to the kernel. And not a single objective measure was brought up in replies, so it's my experience vs theirs. Not helpful.
> Two accounts, both lowercase four random letter names respond to me within two minutes time apart, what do I make of it? :P
That a lot of accounts have obfuscated or meaningless names? That some people value anonymity?
Either way, I agree with them, FWIW.
> I find it hard to believe memory management would vary so much between those two CPU architectures on a single-codebase OS.
Linux is a shitshow when it gets OOM, it takes at least half an hour to get out of it, if it ever does. Windows is not much better.
In contrast, the other day the Force Quit window showed up on my Mac Studio because the OS was running low on memory thanks to a misbehaving app that was taking 70 GB out of 64 GB physical RAM. Overall, almost 120 GB were used, most of it was compressed and a lot of it was swapped. It had absolutely zero effect on how useable the computer was, there was no unusual lag. Either Windows or Linux would have shat the bed long before that point.
> Linux is a shitshow when it gets OOM, it takes at least half an hour to get out of it, if it ever does. Windows is not much better.
That's why you usually want a userspace early oom service. Most preconfigured distros ship one by default. Linux is mostly focused on embedded targets, not servers or workstations. There is not a notion of mobile-style app lifecycle either, not in freedesktop environments that is, but XDG portals are working on addressing that sometime in the near future.
> In contrast, the other day the Force Quit window showed up on my Mac Studio because the OS was running low on memory
Windows does that at since like XP and likely earlier. BeOS did that before Darwin based macOS was a thing. On Linux, I don't know which distros do that, but you're definitely much more likely to see an app die rather than be asked whether to kill it. Freezing, once again, is a result of not having a [working] early oom service.
Linux is not that bad, but traditional freedesktop model kind of is.
It's still much better than mac OS.
Also those replies just look like bots, they were really fast and not providing any value, that's what I meant.
And here's yet another four-lowercase-letter-name for you, then. Dunno about the other two, but I've been using this handle for over twenty years, it was originally the auto-generated username I got assigned on one of my university's servers (generated from my initials).
Low character count handles are a scarce resource, and are often highly-sought after (people were paying crazy amounts for some names on twitter in its heyday). Almost any 2-, 3-, or 4-character sequence is going to be either a word or an abbreviation of something that's meaningful to someone out there.
One is a four letter set of characters without a vowel, the other spells a word with 5 letters. And so what?
I’ll add on that the change to Apple silicon was an amazing improvement, even in the same OS version. Maybe these anecdotes mean your experience in this regard is dated. (I say this as someone who came reluctantly to Mac, and looks forward to returning to Linux)
Have you used any of the Apple silicon machines? In discussions about modern Apple devices and RAM, I don't know that pre-Apple-silicon experiences are all that relevant?
My first was underspec'ed and I used Resolve, Lightroom and Photoshop on top of the usual other stuff and it was quite impressive. The relationship of performance to RAM for earlier machines felt incomparable.
Luddites didn't protest against automation, but rather against inhumane working conditions in factories and smashing their machines just happened to be efficient at destroying profit. So in a way, you could say luddites acknowledge technology for what it is, it's just techbros who fail to comprehend basic facts about machine learning and pushing text compression models as a replacement for everything they (don't) know anything about.
You scrape sites, okay, but what's "ai" got to do with that (I assume ai means chat bots in this context?)
I'm genuinely curious, whatever you're doing sounds cool, but more details beyond the buzzword pitch you'd tell your manager would be welcome on a hard technical site like hn?
(ftr, I'm skeptical of all applications of machine learning, but I keep experimenting with all the various kinds of it, generally with no good result; last real-world useful [to any extent] ml model I tried was BASnet, but whatever you tinkered out sounds cool and if it actually scrapes and filters clothes the way you describe, that'd be quite cool [perhaps even product worthy…?], cuz there are way too many clothes online to look at all of them manually and then esp. on fast fashion sites, there are oftentimes reviews you want to be wary for that indicate low quality products… anyway, that just sounds impossible to automate in my experience, but feel free to prove otherwise)
What "ai" got to do with that would be that he didn't write a scraper and a clothing style ("vibe") categorizer to build a database to process entries in to pick a shop. They just prompted the "ai" (I really don't know why you're putting that in quotes), and it in turn did that for them.
Was it a technically impressive effort from the prompter? No. Are the tools created in the session somehow a massive technical achievement? No. But was it a very useful result? Yes. It took the kind of task that would likely never get done otherwise, and turned it into the kind of thing that got done on a whim.
Doesn't mean that your laptop needs "AI buttons" though.
Ah so what they meant was like a 'vibe coded' a scraper? I thought they meant something like turning descriptions/reviews/photos of clothes into embeddings, as in like sentiment classification but way beyond that? Because the latter would be somewhat cool if it's actually achievable (I doubt it is tho…)
(I mean honestly the project idea[?] they posted sounds like daydreaming some science fiction scenarios, otherwise with all the hype and investment around chat bots, this way of shopping would definitely be mainstream already. If it weren't daydreams, that is. But if my grandma had wheels, she would've been a bicycle, no…?)
You could turn clothing descriptions into embeddings and have a fashion vector database, but doing that would mostly just net you the ability to find adjacent clothings, rather than the ability to navigate available clothing or clothing fitting certain requirements.
What was done is more like using the LLM as a personal assistant that doing long manual labor to find what you might be looking for.
This way of shopping is already a thing. "Hype and investment" goes into how the companies can monetize AI harder (ads! integrated LLM shopping! business development! premium pro max enterprise data policies!), it doesn't really focus much on how the individual can save time and money through non-flashy tasks.
> You could turn clothing descriptions into embeddings and have a fashion vector database
Well, that assumes descriptions are extremely accurate down to the last seam, which is not true. You'd be better off considering reviews and photos, esp. user provided photos, you also need to take into account the model/s in the photos are not necessarily shaped the same as you, so you need to somehow counter that bias in training. This is simply not a task achievable with current ml techniques, however again, feel free to prove otherwise.
(and ftr, I'm of course making a basic assumption that we're past the topic of markov chain/'llm' based chat bots at this point? Those are completely irrelevant to the goal of categorizing clothes based on some characteristics [i.e. the so-called 'vibes'])
Okay but then 'AI' is just a noise generator when it comes to very specific information… I mean just try asking any chat bot to search for something like specific photography gear for some specific scenario and in my experience it's just as good as simply picking some random stuff, except the chat bot will also gaslight you into thinking you made the right choices, so you don't question them… :/
For me, some projects I start by writing a readme.txt by hand. That saves me time in cases I realize I'd be making something pointless. (I don't use chatbots when coding though)
I shall now drive my fart car back to my cozy meatcave from the public meatspace so that I can do some good old ape coding with my smelly carbon-based friends in peace.
Who isn't collecting biometric data using shady sites nowadays? Even Hetzner wanted to match my id with a selfie using some really shady site that almost certainly violates gdpr… What's next, will I have to send a stool sample to view posts on Twitter?? Funny how that's not even unlikely nowadays, with slippery slope becoming less of a fallacy and more of a universal law…
ejabberd is so much easier to set up than prosody, especially containerized. I would highly recommend checking multiple servers out before settling on one tbh.
reply