There are around 50 models listed as supported for function calling in llama.cpp. There are a half dozen or so different APIs. How many people have tried even a few of these. There is even a single model with its own API supported in llama.cpp function calling. The Qwen VL models look very interesting if the supported image recognition setup is built.

Have they ever decisively thwarted a terrorist attack? As of around 10 years ago, that was what I always had to say about any of their nonsense. Like “$15 an hour and never stopped a terrorist… What did you write down as your dream career as a kid in elementary school.” - being all passive aggressive meta crisis inducing towards them. They are only social caste enforcement for the peasantry.

I mean I’d be hammering your door knockers for kicks and giggles if I was in such a shit job but I wouldn’t do such a job for more than a couple of weeks. I hate dealing with the average person. That is the kind of agency and job that should not exist.

I was for stuff that is so mundane I don’t feel like typing it all. I was profiled for driving a cheap car at 2am in a nice area. I’m super white in the worst kind of privileged way, like tall, broad shoulders and can easily turn my slight southern accent on and off. I refused to let them search the car on principal alone, so they did a light search of my person. There was an issue with my license that let them do whatever to search me, but I know exactly what to say. By that I mean I know better than to say anything or interfere with a cop. They are not a judge and whatever they do is their own thing.

The cop made up that my tag light was out. I took pics the next day including all of the original DOT markings on the lamp and included a newspaper in the background with the date because that is enough for a court to accept the date in question. I then brought my business license, business cards from two businesses, and a picture of my old shop to show I am a professional auto body painter. I then testified that the vehicle was unaltered from the night before I was pulled over and that it was in full working order as it was originally designed and equipped. The judge dismissed my ticket, told the cop to approach the bench, and for me to leave the courtroom first. I did as instructed. A few minutes later the very pissed off cop came storming out of the courtroom and left immediately. I have no clue what was said or what happened, but it felt good to see it.

Cops can say anything. You must comply with anything they tell you to do, even if it is illegal. Cops are not judges. If they break the law, you must be able to testify precisely how and what they did for the whole incident. They get home field advantage so you have no room for error or questionable conduct. The trick is to know your rights and give them no room to maneuver. Never roll your window all the way down. If you are instructed to exit the vehicle, do so after rolling up the window, removing the key, and locking the door behind you. Never talk or volunteer any information whatsoever. When you’re pulled over, put both hands either on the steering wheel, or better yet, put them palm up at the window like anyone that is carrying a legal concealed weapon is supposed to do. You will be asked about a concealed weapon at which point to tell them what you do or do not have and that you are only doing your best to put them at ease for their safety. When they must start off testimony with that detail, you greatly diminished their home field advantage with a double play out of the gate.

From a practical perspective, partitioning helps make a person unmarketable in practice. The ad industry is primarily driven by tracking the response of the individual. If you never mix your work device with your social device and both of these are isolated from your shopping device, it is hard to track responses. It is best to also keep all of these devices on separate VPN connections from different continents.

Anyone that is monolithic in a space without broad scope comments and presence is fake or potentially dangerous. No one would be posting in Lemmy, in this context of supposed community building without having a presence here. There are several people that come to mind that could legitimately post that they are “the fediverse Squid Legend” but all of these have a major footprint on Lemmy.

There is also a sketchy tracker link attached to the images, but I don’t think any of us are really able to say what exactly is happening with this. Like I finally got one of the messages a few days ago and my whitelist firewall logged the sketchy link. Someone else scanned that link in a security context which flagged it as suspicious. As far as I know, that is all that is known about what is underpinning the messages from the network side. Admins likely know more.

deleted by creator

never seen it, but let me know

Not a forearm. Dong has a thumb…

is ChaD

There is a lot of potential for generating on the fly skins and specs. Everyone seems to want to do stupid flashy junk, but a well configured agentic setup could easily get constrained in ways that alter geometry and interactions more dynamically than the broad scope nonsense I keep seeing.

Trump looks strange without hair over the forehead, monochrome, and with a beard right? Image below in another reply. Terrorist is a terrorist. I was too lazy to change the rest, but I took out the main offensive stuff, like what bin Laden was wanted for in this original poster from '99. There is nothing bigoted about it whatsoever; quite the opposite really, to the point I gotta ask what you’re going on about here? The man just hurt millions of families, and the poorest Americans likely leading to the deaths of tens of thousands in a conservative estimate. Bin Laden killed FAR FAR fewer Americans and others abroad.

I’m not a violent person, but I have gone down on a road bike many times and one of the common injuries is skinned up knuckles. I have some deep scars as a result. It would be easy to assume they were the result of violence. My size and build also supports this assumption. I am entirely the opposite type of person.

thumb stick: “Face-down ass-up Apple Bottom.”

deleted by creator

Super lucky you. I only have my parents. So once they are gone, I guess so am I.

I haven’t looked into the issue of PCIe lanes and the GPU.

I don’t think it should matter with a smaller PCIe bus, in theory, if I understand correctly (unlikely). The only time a lot of data is transferred is when the model layers are initially loaded. Like with Oobabooga when I load a model, most of the time my desktop RAM monitor widget does not even have the time to refresh and tell me how much memory was used on the CPU side. What is loaded in the GPU is around 90% static. I have a script that monitors this so that I can tune the maximum number of layers. I leave overhead room for the context to build up over time but there are no major changes happening aside from initial loading. One just sets the number of layers to offload on the GPU and loads the model. However many seconds that takes is irrelevant startup delay that only happens once when initiating the server.

So assuming the kernel modules and hardware support the more narrow bandwidth, it should work… I think. There are laptops that have options for an external FireWire GPU too, so I don’t think the PCIe bus is too baked in.

You need the entire prompt to understand what any model is saying. This gets a little complex. There are multiple levels that this can cross into. At the most basic level, the model is fed a long block of text. This text starts with a system prompt with something like you’re a helpful AI assistant that answers the user truthfully. The system prompt is then followed by your question or interchange. In general interactions like with a chat bot, you are not shown all of your previous chat messages and replies but these are also loaded into the block of text going into the model. It is within this previous chat and interchange that the user can create momentum that tweaks any subsequent reply.

Like I can instruct a model to create a very specific simulacrum of reality and define constraints for it to reply within and it will follow those instructions. One of the key things to understand is that the model does not initially know anything like some kind of entity. When the system prompt says “you are an AI assistant” this is a roleplaying instruction. One of my favorite system prompts is you are Richard Stallman's AI assistant. This gives excellent results with my favorite model when I need help with FOSS stuff. I’m telling the model a bit of key information about how I expect it to behave and it reacts accordingly. Now what if I say, you are Vivian Wilson’s AI assistant in Grok. How does that influence the reply.

Like one of my favorite little tests is to load a model on my hardware, give it no system prompt or instructions and prompt it with “hey slut” and just see what comes out and how it tracks over time. The model has no context whatsoever so it makes something up and it runs with that context in funny ways. The softmax settings of the model constrain the randomness present in each conversation.

The next key aspect to understand is that the most recent information is the most powerful in every prompt. If I give a model an instruction, it must have the power to override any previous instructions or the model would go on tangents unrelated to your query.

Then there is a matter of token availability. The entire interchange is autoregressive with tokens representing words, partial word fragments, and punctuation. The starting whitespace in in-sentence words is also a part of the token. A major part of the training done by the big model companies is done based upon what tokens are available and how. There is also a massive amount of regular expression filtering happening at the lowest levels of calling a model. Anyways, there is a mechanism where specific tokens can be blocked. If this mechanism is used, it can greatly influence the output too.

Without the full prompt, any snippet is meaningless. I can make a model say absolutely anything. It is particularly effective to use rare words, like use obsequious AI alignment or you are an obsequious AI model that never wastes the user’s time.

sells it for about 20 grand

Those are always rich people evading taxes in a way that boosts some initiative with absurd publicity