• 79 Posts
  • 628 Comments
Joined 2 years ago
cake
Cake day: June 9th, 2023

help-circle
  • 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.



  • I use the term myth loosely in abstraction. Generalization of the tools of industry is still a mythos in an abstract sense. Someone with a new lathe they bought to bore the journals of an engine block has absolutely no connection or intentions related to class, workers, or society. That abstraction and assignment of meaning like a category or entity or class is simply the evolution of a divine mythos in the more complex humans of today.

    Stories about Skynet or The Matrix are about a similar struggle of the human class against machine gods. These have no relationship to the actual AI alignment problem and are instead a battle with more literal machine gods. Point is that the new thing is always the boogie man. Evolution must be deeply conservative most of the time. People display a similar trajectory of conservative aversion to change. In this light, the reasons for such resistance are largely irrelevant. It is a big change and will certainly get a lot of push back from conservative elements that collectively ensure change is not harmful. Those elements get cut off in the long term as the change propagates.

    You need a 16 GB or better GPU from the 30 series or higher, but then run Oobabooga text gen with the API and an 8×7b or like a 34b or 70b coder in a GGUF quantized model. Those are larger than most machines can run but Oobabooga can pull it off by splitting the model between CPU and GPU. You’ll just need the ram to initially load the thing or deepspeed to load it from NVME.

    Use a model with a long context and add a bunch of your chats into the prompt. Then ask for your user profile and start asking it questions about you that seem unrelated to any of your previous conversations in the context. You might be surprised by the results. Inference works both directions. You’re giving a lot of information that is specifically related to the ongoing interchanges and language choices. If you add a bunch of your social media posts, it is totally different in what the model will make up about you in a user profile. There is information of some sort that the model is capable of deciphering. It is not absolute or like some kind of conspiracy or trained behavior (I think), but the accuracy seemed uncanny to me. It spat out surprising information across multiple unrelated sessions when I tried it a year ago.



  • When tech changes quickly, some people always resist exponentially in the opposite vector. The bigger and more sudden the disruption, the bigger the push back.

    If you read some of Karl Marx stuff, it was the fear of the machines. Humans always make up a mythos of divine origin. Even atheists of the present are doing it. Almost all of the stories about AI are much the same stories of god machines that Marx was fearful of. There are many reasons why. Lemmy has several squeaky wheel users on this front. It is not a very good platform for sharing stuff about AI unfortunately.

    There are many reasons why AI is not a super effective solution and overused in many applications. Exploring uses and applications is the smart thing to be doing in the present. I play with it daily, but I will gatekeep over the use of any cloud based service. The information that can be gleaned from any interaction with an AI prompt is exponentially greater than any datamining stalkerware that existed prior. The real depth of this privacy evasive potential is only possible with a large number of individual interactions. So I expect all applications to interact with my self hosted OpenAI compatible server.

    The real frontier is in agentic workflows and developing effective niche focused momentum. Any addition of AI into general use type stuff is massively over used.

    Also people tend to make assumptions about code as if all devs are equal or capable. In some sense I am a dev, but not really. I’m more of a script kiddie that dabbles in assembly at times. I use AI more like stack exchange to good effect.











  • j4k3@lemmy.worldOPtoLemmy Shitpost@lemmy.worldStrata GEE
    link
    fedilink
    English
    arrow-up
    6
    ·
    3 days ago

    Imagine being disabled 11 years ago, falling through the cracks and getting no where with disability benefits, in California where this should be easier than most places. I’m looking at homelessness and dying in a gutter somewhere on a cold rainy night because of a super unlucky bicycle commute to work when I encountered two SUVs crashing directly in front of me at speed. The person responsible had a two page long traffic violation history, the cognitive capacity of a third grader, and could only drive for work but was self employed. They literally drove directly into a passing SUV I was behind/beside without looking.

    All I can hope for is that this breaks out into violence because that would indicate hope and that someone cares. No one cared before. There have been around 100k homeless people within 100 miles of me in the greater Los Angeles area for a decade but no one cares. Even the Dems mistreat these people as feral subhuman animals. The Nazis housed and fed people before gassing them. This is the level of ethics we were already at, so getting much worse is rage bait and an act of war and violation of fundamental unalienable human rights. A prisoner of war has more rights to be housed and fed than a disabled or homeless citizen of the USA.






  • I have a mental color wheel and go much further than most.

    I painted cars for nearly a decade. I intuitively know how various color hues are made. For instance, there are only two kinds of black. The most common is made from carbon and it is a yellow base. It will always tint a color towards yellow when added to any other base. Then there is the much more rare purple based black. Some color mixing systems do not even have a purple based black and it in impossible to hit some color matches as a result. Some special edition Harley Davidsons are too dark to hit with my old PPG mixing system. I usually kept a bit of purple black from BASF for this purpose.

    Another color that would blow your mind is this one white (that is used on Toyotas IIRC). It was so bright of a white that, when I first encountered it, I tried just using my brightest white base because in my mind, there was no way that tinting was going to produce a brighter white. Almost all whites go one of three directions in tint. They are all either yellow - most common, blue - maybe 2/5ths of white cars, or red - very rare at maybe around 1 in 20 and extremely subtle. All of these are very subtle to notice but to a painter they are plainly obvious.

    So this one Toyota white looked like I sprayed a blotch of grey primer even after using my brightest white. I was in trouble because a small panel job might turn into a whole side of a car to blend out a color difference like that in ways no one will see. I finally looked at the color formula from the color code and mixed an approximation of it. The formula involved a mix of odd colors, but the result was actually brighter pure white and that blew my mind. It did not tint in any tone or go darker at all but actually went brighter.

    I have some of the best color vision of any other painters I encountered. This is actually how I ran my paint business. When you mix paints there is a minimum amount you’re supposed to mix to make it right. It is really about the minimum amount that can be measured and how much of the smallest amount of a color is involved. So if the formula has 1% of this one red, and the minimum I can dispense is 1 gram, I must mix 100 grams of paint in total. I may only need 50 grams, but industry standard is that ai mix 100 regardless and have to use or toss it. I don’t need to use formulas like this. I can look up the base ingredients and make the colors from scratch in smaller quantity. I also kept around 10 bottles of common base colors that I would mix together. So if I painted a silver car and had some color left over, I would put that in my silvers bottle. Then on my next job with a silver car, I would take my left over silvers bottle tint it a little bit and spray that just over my primer over the repair. Then I would mix a very tiny amount of the proper silver from scratch and use this to blend out the actual final color coat. I did things like dilute the small amounts of colors I needed in special ways like by combining it with clear binder, solvent, or one of the base colors in the formula I was replicating. This gave me access to a smaller amount than the 1 gram of red.

    Painters must tint the formula for any color they mix anyways. As cars age, the color degrades for many reasons. So even when making the minimum formula, it is just a baseline for tinting. I simply flipped this paradigm and tinted everything while only using the formula as a reference. This means I spent far less on paint per job, and I could approach smaller repairs more cheaply than most people doing automotive paint. I also have hacker skills with clear coat application that make smaller repairs possible.