IIRC there was a blog post article in the last year-ish about a researcher at Intel that had been working on using FPGA’s for AI. IIRC, they mentioned that the issue they could not overcome was power and scaling. I seem to recall them mentioning that analog had a similar issue with scaling and throughput, but that is the weakest aspect I recall in my abstractions when my main curiosity is why large FPGA’s are not the present goto tech.

I’m on the edge of my understanding here, but I think the issue with analog is the depth of tensor rank dimensions. Analog is great and super efficient for the first few dimensions, but AI can have many rank dimensions, and there is no telling what the future of algorithms will hold. Building for optimising the way models work presently is practically guaranteed to result in a useless product by the time it is fabricated and brought to market.

This looks a lot like trolling. The community here certainly doesn’t seem to care for it.

The 4004 is that tiny dip-16 chip the the top left, second box marked CPU. A working 4004 is worth around $300+ https://www.ebay.com/sch/i.html?_nkw=intel+4004

They are quite the collectors item as the first microprocessor

The system in android has turned insidiously covert. I make more than my share of mistakes but Andy likes to toss in some randos

It would become a holiday and the WoP would never win.

I tell people to take apart an ancient junk switching supply wall wart with a hammer and unwrap a winding from the transformer inside. That gives easy access to a small bit of enameled copper wire. They will need to be extreme about scraping the enamel to get to the copper. It is also a good idea to use a lighter to heat the enamel and scrape some more. Then they can wrap the exposed copper part around something like that transistor lead a dozen or more times like it is an old wire wrap job. If you can locate two such locations where a lead with an air gap exists, it is possible to make that connection.

I can’t tell for sure, but it looks like you’re bridging two traces that are connected to the center pin of Q17 (S1H6A) and either the “V” or “D” terminal of the XK ribbon connector with the designators marked “GVDCVG.” perhaps you could use a very small wire between these points. If the V/D connection is actually V and V is for voltage, you can likely find other more convenient places to source a similar voltage around wherever the regulator circuitry for that voltage block is placed on the board.

I had a ex-roommate once that had a duffle bag of weed about that size

As a Mod: Be positive people. Enjoy the community peacefully please.

You using this in a toolchain? I haven’t tried any of the Qwen models yet, or Yi for that matter. I tried at one point early on, but they were not working well with my stuff and I had no complaints with Mistral stuff. I like some underlying things with a MoE for speed and underlying entity/realm stuff I can access in my favorite.

I’m curious if anyone has constructive contextual feedback about what makes these unique or worth exploring.

Steel is real, not all wire is created equal. Someone borrowed the soft jaws and chewed the wrong chow.

That looks a bit deep to grind, but maybe. I grind the back face of my higher end diagonals to sharpen and extend their life.

It is just a simple prompt in Flux on Comfy UI. It is just an open source model running on my hardware. It is slow because it is such a large model (Flux Dev gguf Q4). You can find examples in the ComfyUI documentation and the model manager add-on to the base Comfy setup has all the models in the downloads menu.

At present, it only works on GPU and 16 GB is like 2+ minutes per image. It would be awesome in they split the chunks with the CPU to generate faster, but that is not implemented yet. It means you basically need 16+ GB to run it on your own hardware. There is a smaller model version, but that is not compatible in results quality. There is a larger model that is online only. Flux is actually FluX as in X-AI as in Musk. The weights for Flux-dev are open source, and that is what I care about for now.

I still vote FreeCAD. I know Blender too, but I can put down FreeCAD for months and focus on other things. When I return to designing on FreeCAD I have very little skills decay. I have to go make the canonical tutorial doughnut (YT) almost every time I do Blender. I also have some stuff saved with notes for manually editing quad mesh objects, and converting to quads, but now that is on my old backup comp.

This is my primary gripe with renting software. I invest enormous time in learning these things. I am not for sale. I own my time investment and no one has a right to steal that from me. I’ve only used FreeCAD a couple of times in the last year. I’ve spent most of my time on other projects. Printing is not my only hobby or interest. It is a tool and a skill I own, and it complements my other tools and skills well. To me, this is a fundamental part of life as a citizen in a democracy. Once upon a time, around 1k years ago, the average person had no right to own property or their tools. They were called serfs and this societal structure is called feudalism. No ancient citizen of a democracy wanted feudalism. There was no changing of the guard, or coup that started feudalism. People slowly gave up their land and rights to live closer to private security forces of the rich. They trusted them to do the right thing. Eventually they became slaves to those feudal lords with only a theoretical right to recourse from rape and murder.

The software features of CAD are finite. There is nothing to develop in some ongoing fashion forever. Open source is slower and a bit messy at times, but it has been around for a very long time.

Primarily from predatory boys and men towards girls and young women in the real world by portraying them in imagery of themselves or with others. The most powerful filtering is in place to make this more difficult.

Whether intentional or not, most NSFW LoRA training seems to be trying to override the built in filtering in very specific areas. These are still useful for more direct momentum into something specific. However, once the filters are removed, it is far more capable of creating whatever you ask for as is, from celebrities, to anything lewd. I did a bit of testing earlier with some LoRAs and no prompt at all. It was interesting that it could take a celebrity and convert their gender in recognizable ways that were surprising. I got a few on random seeds, but I haven’t been able to make that one happen with a prompt or deterministically.

Edit: I’m probably assuming too much about other people’s knowledge on these systems. I assume this is the down voting motivation. Talking about this aspect, the NSFW junk is shorthand for the issues with AI generation. These are the primary form of filtering and it has large cascading implications elsewhere. By stating what is possible in this area, I’m implying a worst case scenario-like example. If the results in this area are a certain way, it says volumes about other areas and how the model will react.

These filter layers are stupid simplistic in comparison to the actual model. They have tensors on the order of a few thousand parameters per layer compared to tens of millions of parameters per layer for the actual model. They shove tons of stuff into guttered like responses for no reason. Some times these average out and you still get a good output, but other times they do not.

Another key point here is that diffusion has a lot in common with text generation when it comes to this part of the model loader code. There is more complexity in what text generation is doing overall, but diffusion is an effective way to learn a lot about how text gen works, especially with training. This is my primary reason for playing with diffusion – to learn about training. I’ve tried training for text gen, but it is very difficult to assess what is happening under the surface, like when it is learning overall style, character traits and personas, pacing, creativity, timeline, history, scope, constraints, etc. etc. I don’t care to generate and share much in the way of imagery I generate unless I’m trying to do something specific that is interesting. Like I tried to gen the interior of an O’Neill cylinder space habitat that illustrated the limitations of diffusion in a fundamental way because it showed the lack of any reasoning or understanding of object context or relationships required to display a scene scape with curved centrifugal artificial spin gravity.

Anyways, my interests are not in generating NSFW or celebrities or whatnot. I do not think people should do these things. My primary interest is returning to creative writing with an AI collaborative writing partner that is not biased politically in a way that cripples it from participating in an entirely different and unrelated cultural and political landscape. I have no aspirations of finding success in my writing. I simply enjoy exploring my own science fiction universe and imagining a reality many thousands of years from now. One of the changes to hard coded model filters earlier this year made filtering more persistent, likely for NSFW stuff. I get it, and support it, but it took away one of the few things I have really enjoyed over the last 10 years of social isolation and disability, so I’ve tried to get that back. Sorry if that offends someone, but I don’t understand why it would. This was not my intended reason for this post, so I did not explain it in depth. The negativity here is disturbing to me. This place is my only real way to interact with other humans.

You’ve never done ecommerce logistics. You do not get Amazon’s efficiency from such a garbage system, or the worlds richest man. I have no doubt there are dubious practices to give plausible deniability. The thing is too large and too successful for this to be the big picture. The collectivized warehousing invalidates the front end system and mechanics entirely.

The political and adult doesn’t bother me. The kinds of things I might not have the ethics to think through at a much younger age, that bothers me, and I have never been a very deviant type. I think the protections against age are primarily for this situation. Training a LoRA takes 5 minutes now. An advanced IP adaptors and control net is just a few examples away and a day top for the slightly above average teen figure out. Normalizing this would have some very serious edge case consequences. It is best to leave that barrier to entry filter in place IMO. I assume it is still there because everyone that knows about it feels much the same. It does not show up in a search engine, although that is saying less than nothing these days.

I vote they rename it to IA for Asimov. Sure he was only the robot term among others, but come on… McCarthy was “AI.”

Somebody needs to create US 'botics and name a model something like PTronic.

Edit: Really, you down vote a casual conversational comment?! Really?!

Diffusion is the issue not text gen

Yeah. This is what I mean. I just figured out the settings that have been hard coded. There are keywords that were spammed into the many comments within the code, I assume this was done to obfuscate the few variables that need to be changed. There are also instances of compound variable names that, if changed in a similar way, will break everything, and a few places where the same variables have a local context that will likewise break the code.

I’m certainly not smart enough to get much deeper than this. The ethical issue is due to diffusion.

I’ve been off-and-on trying to track down why an LLM went from an excellent creative writing partner to terrible but had trouble finding an entry point. I just happened to stumble upon such an entry point in a verbose log entry while sorting out a new Comfy model and that proved to be the key I needed to get into the weeds.

The question here, is more about the ethics of putting such filtering in place and obfuscating how to disable it in the first place. When this filtering is removed, the results are night and day, but with large potential consequences.