I’m as smart as any AI, I can’t code, type or count. 
I think some people just drink the coolaid and believe it’s good for them. 
2 Likes
Well, I think you went too far 
.
I have to inform you (since you haven’t seen my previous statements), that I strongly dislike being told that I am wrong. 
Here is a quick proof of the difference between Artificial and Human Intelligence:
The above answer seems to refer to organic neural something, while quoting Artificial Intelligence is a fake name. A human can see the error, while a machine cannot. 
And here is one of the many examples of saying two contradictory things in one sentence only:
- A machine can originally produce something by itself (negating my statement, which is untrue)
- Simple functions can produce something on their own.
A machine is a piece of hardware. A function is a piece of software. A machine is built by humans. A function is created by humans.
How original is something created by a machine, when the hardware and the software that consist the Intelligent machine are built by humans?
Can you get any answer from a PC, while it is unplugged from electrical power?
Can you get any answer from a PC, while it has no Operating System?
Can you get any answer from a PC, while it has no Firmware/BIOS?
Even terminology suggests that artificial means Made by humans.
Unless you yourself are some planted AI agent, to fish responses from real humans, so you can improve the monster’s capabilities, you should have realized one or both of these two things:
- Machines cannot do anything by themselves
- You should never tell me I am wrong
Cheers! 
Didn’t know then that computers have mothers and boards together with them. If start to play around this linguistics quirks about how same names could represent something opposite.
I define software as a machine/part of machine/machine spirit too.
Currently it’s mostly machines built by machines, while humans mostly watching into the glass windows of a clear rooms.
Functions are not created by humans, no one can create knowledge besides of the God. We can discover them. To say that someone created a function is like to say that Newton created gravity.
Same rule applies to man.
nanos gigantum humeris insidentes
Can you get an answer from a man in diabetical coma?
His cells are literally in the similar to plugged off state, because without insulin there’s no glucose going inside from the bloodstream. No insulin, close to no energy.
Can you get an answer from redditor man without brain or after strong insult? Even though by our faith his souls is still intact if he’s still alive.
Main processing unit and OS shipped with it are missing.
Can you get an answer from man with one of a hundreds mental issues which made him incapable of learning, so basically means he is lacking firmware?
01101000 01110100 01110100 01110000 01110011 00111010 00101111 00101111 01111001 01101111 01110101 01110100 01110101 00101110 01100010 01100101 00101111 00111001 01100111 01001001 01001101 01011010 00110000 01010111 01111001 01011001 00111000 00111000
##### News Flash!!! #####
- Serious communication problems because of the language medium makes this discussion impossible.
- More and more evidense of an AI participating in the conversation flooded the neighbourhood.
Too much Kool aid being consumed. 
He loved Big Brother.
In high school also they forced me to read Fahrenheit 451. Now I don’t even remember anything about it other than the fireman protagonist and the scientists trying to recite entire books to him. 
I’m not a AI expert, and this is a forum for a Linux distro, so it may be better to also go to their Discord (linked at the top of the README.MD) and ask as well, especially since apparently nobody else here will help you (cough @petsam 
). However, I’ll do my best to clarify the steps and answer your questions.
-
Installation
To install, you first download the checkpoint (basically, a partially-trained model) linked, clone the repository (either via agit cloneor downloading and extracting the zip), and then in a terminalcdto the repository dir. Once there, you want to place the downloaded checkpoint file inchat/. Then, you execute./chat/gpt4all-lora-quantized-linux-x86to start the model. You may have tochmod +x ./chat/gpt4all-lora-quantized-linux-x86first, to mark it executable. -
It doesn’t appear it has the ability to fetch data from your system? There’s not much info in the repository, other than its planned to integrate Atlas (a training set database). This would be a question better suited for Nomic.
-
You might be able to add data to the training set, skimming over the code i don’t immediately see how… The JSON is here.
-
There’s no information on the repository, it may only be English. Best to ask in their discord.
The remaining questions are best suited for Nomic. I hope i was able to help a little bit :3
I’ve recommended it before and still think that this project is easier to use, as it’s has GUI and shipping not just LlaMA, but quantized and fine-tuned Alpaca.
Apologies, i mustve missed the link. That is much better than the original repo, might actually use it myself 
AI is created by idiots, for idiots, fed by all idiots in the world simultaneously, then used by idiots in order to train AI on it’s results to complete the circle of idiocy…
I HAVE A PROOF NOW!!!
This thread is the best of everything ever posted!
The singularity is near
One place that I have found helpful to a degree was HuggingFace and looking into the GPT-J model.
I managed to get it up and running on ‘cough cough’ Win 11 and put the guide here. https://github.com/Promootheus/Beginner-AI-Knowledgebase/wiki
I started to make a wiki for myself on the terminology, as there’s a hell of a lot to understand.
I was going to replicate it on EndeavourOS, but I’m starting to feel like the best best is going to be trying to get Llama, Alpaca or Vicuna working instead on GPT-J. Some of the HuggingFace youtube vids give a reasonable overview of how it works in general.
GPT-J required a lot of RAM for me, I think about 30GB
The way I understand it is that the Model is like a brain full of neurons, but no synapses connecting the neurones to each other.
During training with vast amounts of text the ‘synapses’ are created between the neurons, and these synapses have a different connection strength/weight based upon the word patterns in the text it has been fed with.
The Model is useless in itself without the weights that have been learned, in the same way that Einsteins brain would be useless with all the neurones, but no synaptic connections between them.
Training Einstein brain requires more hardware power than is available to the average person. But, what can be done is taking the existing Model and the Pretrained weights and modify it in one of two ways.
- Reduce the number of neurons (make the model smaller)
- Simplify the synapses connecting the neurones (Quantization=32bit/16bit floats converted to integers=less memory needed to store the numbers, so can be run on consumer level hardware)
Llama, the LLM created by Meta is a big brain that took vast computing power to train and is too big for typical consumer hardware to run, so Alpaca was created to simplify the synaptic connections between the neurones of the big brain through quantization. Imagine Einsteins brain recreated in Minecraft.
The models with their corresponding weights when running are are stored in VRAM or DRAM.
Storing in video RAM is faster to read/right to, and if using an NVIDIA gfx card the cards dedicated Tensor cores can be used to do the number crunching. As most of us mere mortals don’t have GFX cards with 22GB of RAM, we would have to run Hamster brain instead of Einstein brain.
One alternative is by using CPU cores to do the number crunching and also store the model and weights in DRAM, which is reasonable cheap for many folks. This is what I had to do with the GPT-J to run it. I have a 2080Ti with 11GB of VRAM, so it was a no-go using the gfx card.
I manageds to get Alpa working using this guide…
This guy has done some really good videos. Would be great to learn how to feed in your own data to fine tune it though.
Ya this is hilarious! It’s like clown laughter.
You can use koboldcpp to run models shared on CPU and GPU.