• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

The anti-AI thread

Page 39 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
AI sucks, but I've seen senior design engineers make similar mistakes (connector is backwards, missing a net connection, etc.). But not all at once like AI does.
 
Was wondering when I'd start seeing this shit, as an EE I'm gonna say this guy can take his AI circuit board and shove it up his ass. How much time did he piss away trying to do something with AI that could be done in like a couple hours or less by a person. If I fucked up a board this simple so badly on my first spin my ass would be fired but for some reason people keep plugging away at it with AI.
 
Was wondering when I'd start seeing this shit, as an EE I'm gonna say this guy can take his AI circuit board and shove it up his ass. How much time did he piss away trying to do something with AI that could be done in like a couple hours or less by a person. If I fucked up a board this simple so badly on my first spin my ass would be fired but for some reason people keep plugging away at it with AI.
I don't know what the purpose of the project was. All I know is what I presented here. It's kind of an amusing exercise that doesn't cost much money. 'See what the computer can do by itself'. However, he talks about V1, so it sounds like he's gonna persist with it. After getting it so horribly wrong on such a simple device, I don't know why you'd try again any time soon. If the next one is correct, is it because the process improved, or did you just get lucky?

That's the biggest problem with ai imo. It's a random results generator. Regardless of how well you structure it, it comes down to chance whether or not it's right. I don't know how you work with that.
 
We should skip the CAD altogether and tie the AI directly to the pcb etcher and pick and place machine! Instead of submitting gerber files you just submit an AI prompt. 😛
 
I don't know what the purpose of the project was. All I know is what I presented here. It's kind of an amusing exercise that doesn't cost much money. 'See what the computer can do by itself'. However, he talks about V1, so it sounds like he's gonna persist with it. After getting it so horribly wrong on such a simple device, I don't know why you'd try again any time soon. If the next one is correct, is it because the process improved, or did you just get lucky?

That's the biggest problem with ai imo. It's a random results generator. Regardless of how well you structure it, it comes down to chance whether or not it's right. I don't know how you work with that.
This is why I'm dubious about the intent, or process. 'ill fated question that results in random/dubious output' was definitely a feature of earlier LLMs but today? AI should be able to smoke such a simple target. I'm guessing something was janky with his process, or he intentionally screwed it up.
 
OpenAI: “we are merely a knowledge base.”



I have the feeling that Schroedinger's Cat tactics are going to feature a lot in 21st century lawyering weaselspeak: "Its true AI!" "Well, when we say intelligent..."
 
I can kinda see the AI company's point of view on this. It's not any different than using Google or even a public library in order to learn information and then use it maliciously. Would be a dangerous precedent if we start holding organizations responsible for simply providing general use info without ill intent. Like holding a steel company responsible for someone buying steel to make a weapon.

I suspect they will lose though, they'll probably be required to monitor prompts more closely and maybe auto detect suspicious activity and then shutdown people that appear to be using it for bad purposes.
 
All rejoice and fear! AI has shown the capability to self-design a RISC-V CPU, named VerCore, from scratch in 12 hours with no human intervention.


"The chip reportedly scored 3,261 on the CoreMark benchmark, a specialized tool used to measure the performance of microcontrollers and CPUs in embedded systems. The modest score suggests that VerCore can barely keep pace with an entry-level Intel Celeron SU2300 from mid-2011, indicating that the new design is unlikely to compete with modern CPUs."

This, despite being designed to be built on a 7nm-class process node.

Methinks Skynet's daddy is intellectually disabled, at least for now. And, out of curiosity, since they don't mention it I wonder how many hardware security design flaws ended up in it?

______

And, in more stupid Mark Suckerburg news, Meta will now record employee screens, clicks, and keystrokes to train AI that may replace them:


I guess Mark will be exempted, since he is already building his own AI replacement.
 
Last edited:
I don't really follow this stuff. How does RISC-V compare to x86 in general? How well would a human designed chip perform?
 
I don't really follow this stuff. How does RISC-V compare to x86 in general? How well would a human designed chip perform?

There really isn't a comparison. RISC-V is mainly used at this point for efficient embedded silicon and IOT type workloads. The companies using it tend to lean towards royalty-free custom designed embedded silicon. It has potential, mainly in that it is a open source ISA that anybody can improve and use for their own needs free of charge. Its main disadvantage is that, unlike open source software licenses, companies that use it and improve it are not required to return those improvements back to the RISC-V community.

It will not be catching up to either x86 (or ARM, for that matter) anytime soon.
 
LibreOffice does what I want it to do. I'm not a huge office program user though, and I don't think I've used MSoffice since ~2005. I switched our company over to libreoffice, and all the document blanks to odt. Seems to work for everyone.
 
I'm probably not a "power user" of LibreOffice, but the spreadsheet program does everything I want, and the text editor does everything I want

Well, I suppose tweaking the line spacing in a desperate effort to try to squeeze all the chords for Drunken Lullabies onto a single page is maybe not an average user thing to do...
 
I actually prefer Libre Office over whatever the hell Microsoft is doing with Office these days. Office 2003 was peak after that it started to go the enshitification route.

Outlook and Excel is mostly what we use at work. Outlook is horrible now. It's so bloody slow and also really finicky sometimes. I don't get how companies think going cloud is better. When we had a local exchange server it was faster.
 
Hey, just for shits and giggles, here's what that ended up looking like... one of those "just because you can, doesn't mean you should" moments. And I was only barely able to do it.

0424261454.jpg
 
Ryzen CPU prices up over 50% in Japan - here comes the rumored CPU shortage segment of the AI madness.


Memory, storage, hard drives - all the AI price spikes were pretty much foreshadowed in Japan before they hit everywhere else. So, I guess we now have to worry about CPUs. Anyone who plans to build a new system might just want to seriously consider going ahead and grabbing a decent CPU pretty soon.
 
Back
Top