|
|
| Page 34 of 34 |
vurt
Posts: 14161
Location: Sweden
|
|
| Back to top |
|
 |
LeoNatan
☢ NFOHump Despot ☢
Posts: 74719
Location: Israel
|
 Posted: Thu, 16th Apr 2026 02:52 Post subject: |
|
|
| vurt wrote: | My "piracy" recipe for stealing hardware  |
The company I work for apparently has a datacenter with 12 H100 racks for on premises Claude. Who's up for a brisk midnight adventure?  |
|
| Back to top |
|
|
vurt
Posts: 14161
Location: Sweden
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 74719
Location: Israel
|
| Posted: Thu, 16th Apr 2026 03:39 Post subject: |
|
|
OK, I'll PM you the address. I'll be ... right over there, on the lookout
|
|
| Back to top |
|
|
vurt
Posts: 14161
Location: Sweden
|
| Posted: Thu, 16th Apr 2026 04:11 Post subject: |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Fri, 17th Apr 2026 09:42 Post subject: |
|
|
https://x.com/elonmusk/status/2044990537145753894
Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI.
AI/robotics will produce goods & services far in excess of the increase in the money supply, so there will not be inflation.
Ai stealing jobs huh.
|
|
| Back to top |
|
|
couleur
[Moderator] Janitor
Posts: 14863
|
| Posted: Fri, 17th Apr 2026 10:15 Post subject: |
|
|
AI is also "stealing" cognitive capacities as it makes us, well at least the kids, dumber (why use my brain when an AI can do it better) and then resells it to us. Its the classic scheme.
"Enlightenment is man's emergence from his self-imposed nonage. Nonage is the inability to use one's own understanding without another's guidance. This nonage is self-imposed if its cause lies not in lack of understanding but in indecision and lack of courage to use one's own mind without another's guidance. Dare to know! (Sapere aude.) "Have the courage to use your own understanding," is therefore the motto of the enlightenment."
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 74719
Location: Israel
|
| Posted: Fri, 17th Apr 2026 10:24 Post subject: |
|
|
| couleur wrote: | | AI is also "stealing" cognitive capacities as it makes us, well at least the kids, dumber (why use my brain when an AI can do it better) and then resells it to us. Its the classic scheme. |
I am dealing with this shit at work. We have UI/UX designers that are fully captured by this way of "thinking". We need a string for a UI element, so they "ask" the AI slop to come up with some shit, and the type of sentences that come up is just mind boggling, but there is nobody to complain to, as these UX "designers" don't even understand the criticism. I'd probably get better results out of the slop pachinko machine myself, if I skipped the middleperson, since I can come up with a more technical prompt.
These are the same kids that fear AI is taking their job. It is—because they are not much more than prompt operators at this point, and it has infested their ability to do independent work. We have "engineers" whose instinct these days with crash reports is to "run it through claude" and if it returns nothing or something really stupid, they are stumped—without even looking at the crash report details. We have people that insist on "writing" PR descriptions using generated slop, then when people respond to their slop, they don't understand the feedback since they didn't even read what the slop returned before copypasta to Bitbucket. |
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 11560
Location: Florida, USA
|
| Posted: Fri, 17th Apr 2026 10:43 Post subject: |
|
|
@LeoNatan
I agree, vibe coding needs to die. Vibe coding is just a 'cool sounding buzz term' for: Fuck it let AI do it, it's good enough, I guess...it..works.
Companies that outsource call center support should stops saying they outsource it, and start saying they offer "We are now Vibe Supporting you". Same to me: offload you job to something else worse at it.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 74719
Location: Israel
|
| Posted: Fri, 17th Apr 2026 10:54 Post subject: |
|
|
It's not even vibe coding at this point, it's just laziness. "Look at my git diff and write a PR description". "Come up with a string for a title for a welcome screen". I'm not kidding, it's at this level. The younger generation doesn't even know or want to do the most mundane things their low grade job title asks them to do.
|
|
| Back to top |
|
|
vurt
Posts: 14161
Location: Sweden
|
| Posted: Fri, 17th Apr 2026 10:59 Post subject: |
|
|
every programmer i know of laughed at AI coding a little more than a year ago, today almost everyone of them uses it because its just way faster to use and as good.
so no, "vibe coding" (God do i hate that word!) / AI-coding wont die and doesnt need to die at all, we will eventually end up with far, far better code than humans can produce.
It's like debating if we should have calculators or using humans and abacuses instead. While the later is absolutely possible humans will never be better at these things, this is tasks made for computers to do.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 74719
Location: Israel
|
| Posted: Fri, 17th Apr 2026 11:24 Post subject: |
|
|
“Just as good” as what? The slop Claude spews is not even as good as a junior engineer, while it spews the crap it spews with utmost confidence of a know it all junior that knows nothing but thinks they are on the top of the world. It hallucinates stdlib classes and functions, hallucinates patterns, hallucinates compiler directives. When told it hallucinated, it sometimes argues, but eventually apologizes … and hallucinates more nonsense.
|
|
| Back to top |
|
|
Il_Padrino
Posts: 7900
Location: Greece by the North Sea
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 11560
Location: Florida, USA
|
| Posted: Fri, 17th Apr 2026 11:52 Post subject: |
|
|
I don't think AI coding needs to die per say. I think people with jobs as it, or getting jobs as one need to learn how to code, before passing off tasks to AI.
Not same-same. But I expect a Pilot to know how to fly a plane, before relying on auto-pilot to fly it for him. Same feeling to me.
Want to code for yourself, or hobby stuff and dont want to learn it all for tiny self projects? Go for it with AI. If your new career is a programmer/software engineer, You should know what you (and it) are doing. You and whoever you work for relies on you (buck stops with you if its bad/wrong/broken/needs fixing).
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
vurt
Posts: 14161
Location: Sweden
|
| Posted: Fri, 17th Apr 2026 12:23 Post subject: |
|
|
| LeoNatan wrote: | | “Just as good” as what? |
As good as most programmers. You can check swe-bench results. Yes, it' still not better than the best programmers in the world, so yes, if you can afford to hire those guys, which obviously there aren't too many of, you will have the edge.
I program with it daily (like stated earlier, currently plugins for audio), i let real programmers look at the code too.
Hallucinated code etc, no, that would just not produce the stuff i am doing, a total no-go where latency and performance is key. The hardware (MetaModule, a tiny hardware module) i run the code on is Dual-core 800MHz Cortex-A7 plus Cortex-M4 co-processor, 512MB DDR3 533MHz RAM, so very restrained in comparison to a modern PC. I compare with commercial plugins all the time, my performance is top-notch, not buggy (would directly show) or has any more latency than other plugins...
Setting up the environment properly is key too, the instructions for the AI, if you don't do that part properly the models will not produce good enough results. They still do need custom instructions for whatever they are going to do. That's mostly what people don't seem to understand or are good at setting up, you must research this first, even just asking something pretty simple might still require some tweaks to the default settings of the model.
|
|
| Back to top |
|
|
vurt
Posts: 14161
Location: Sweden
|
| Posted: Fri, 17th Apr 2026 12:27 Post subject: |
|
|
| Il_Padrino wrote: | | Calculators are accurate because based on mathematics. AI are statistical approximations, and by definition can never be 100% accurate unless your input matches its training data 100%. |
Not true at all, that would make them pretty much useless.
LLMs can produce correct outputs on inputs never seen in training. Generalization is the entire point of learning.
Believing partially made-up and buggy code would produce an entirely new game or a new plugin or DSP-code no one has ever made etc? Wut? how would that even work, it wouldn't. You would not be able to run it or at best start it and then it would produce errors or freeze.
That's kind of the cool thing with doing audio plugins, i can test every part in the code by triggering my modules at audio rate (which is insanely fast), the module would just crash instantly, in milliseconds. So pretty good playground for code stability, performance, latency and especailly so running on very low-end hardware.
Another good example btw;
With Claude Mythos they were able to find 500+ vulnerabilities, some which has been undiscovered for 23 years. The dude found more bugs and vulnerabilities in a week than in his entire life. You can do this type of work in seconds, vs likely years or never been able to find it. This is both good and bad of course, because it could be used by hackers instead of people doing the fixes..
So no, if AI could only do things from its training data, matching it 100% it could not do it. All you would have a is a database + a simple chatbot, it could probably run on a 286, we would not have the issue with either expensive RAM or VRAM, power issues we are facing etc. This is as hi-tech as it gets, beyond human comprehension of how the models actually works and why it works. The researchers have a general idea only.
|
|
| Back to top |
|
|
Il_Padrino
Posts: 7900
Location: Greece by the North Sea
|
|
| Back to top |
|
|
| Page 34 of 34 |
All times are GMT + 1 Hour |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB 2.0.8 © 2001, 2002 phpBB Group
|
|
 |
|
