There was a post asking people their opinions about Edge and many people seemed to liked the idea of Edge and seemed to be ok having it on Linux (Blasphemy)
Also, can we all agree how fast Edge went from joke to a threat? I mean, it’s good now alright! It was good back then, but it’s better now. Money man!!! Money!
Personally I hate MS, but I can’t help but see the fact that there is no alternative to Bing GPT and many features Bing offers on Linux.
If there is an open source ChatGPT how would it look? Who would bear the costs? How would we solve the server problem? i.e., it would take a ton of server space and bandwidth. Just wondering.
I am pretty sure MS products will improve greatly due to their integration with GPT what do us poor folks on Linux do?
Just want to know the answers, I don’t want to discuss (aka can’t comment, I need to study), but just curious!
what do us poor folks on Linux do?
Run llama.cpp and any of the models listed here, that stuff has been around for months.
TheBloke has a lot of models converted to GGUF format which you need for
llama.cpp.Quick Start Guide (requires Nix, otherwise compile llama.cpp manually):
$ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/TheBloke/guanaco-7B-GGUF $ cd guanaco-7B-GGUF $ git lfs pull --include=Guanaco-7B.Q4_0.gguf $ nix run github:ggerganov/llama.cpp -- -m Guanaco-7B.Q4_0.gguf --instruct > Write haiku about a penguin A penguin walks on ice, Takes a plunge in the sea, Hides his feet from me!a package manager that can pull, build, and run from git with one command is pretty neat
I ran it on my pc with a gtx 1070 with cuda enabled and compiled with the cuda compile hint but it ran really slowly how do you get it to run fast?
To make use of GPU acceleration you have to compile it with the proper support (CUDA, OpenCL, ROCM) and add
--gpu-layers 16(or a larger number, however much your VRAM can handle). If that’s not enough, than the GPU/CPU is probably to slow.You can try a smaller model, those run faster, but give worse results.
Thanks I might try that out later.
KoboldAI has an “KoboldGPT” mode, which is for answering questions. You can try it at https://lite.koboldai.net/ or run it on your own.
Here’s an example conversation I just had with it (emoji icons added):
🤖 Hello, I am KoboldGPT, your personal AI assistant. What would you like to know?
🧑 What is the name of the original version of the game Monopoly?
🤖 The original version of the game Monopoly is called “The Landlord’s Game.” It was created by Elizabeth Magie in the early 20th century. However, the most well-known version of Monopoly today is based on a variant called “Monopoly” that was designed by Charles Darrow in 1935.
🧑 Did “The Landlord’s Game” have any purpose besides being a board game?
🤖 Yes, “The Landlord’s Game” had a specific purpose beyond being a board game. Elizabeth Magie created it as a way to illustrate the negative aspects of concentrating land in private monopolies. She believed that it could be used as a tool for teaching economic concepts such as rent, taxes, and the distribution of wealth.
🧑 What economic ideology was associated with this game?
🤖 Elizabeth Magie was influenced by Henry George’s philosophy of Georgism. Georgists believe that land should belong to all people equally, and that taxation should be based on the value of land rather than on income or production.
Host it locally https://github.com/nomic-ai/gpt4all
As others said you can host Koboldai on your own, and if you don’t have a powerful GPU, anyone can use powerful llm models via the AI horde
i don’t know a single thing about whatever these things your talking about are, but i just want to say, you’re the first person i’ve ever seen suggest that edge isn’t a joke.
There are quite a few of them out there. Just a few from my notes (in reverse chronological order)
- https://github.com/persimmon-ai-labs/adept-inference
- https://github.com/shellc/laozy
- https://github.com/imartinez/privateGPT
- https://github.com/Josh-XT/AGiXT
- https://github.com/tloen/alpaca-lora
The self-hosted AI/ML system has been here for a while. Granted, the vast majority of them require downloading pre-trained models before they can be used due to how much it costs to build a system of weights from scratch.
I sometimes wonder if it would be possible to build and train a truly open source model with BOINC or something. The last 30 years of history show that it’s entirely feasible to build a massively distributed computing cluster, why not leverage this to build a model? I know how naive that sounds immediately after writing it, mostly because of the difficulty of getting a large enough training data set, which unfortunately has risk written all over it (read: people poisoning the model, ala Microsoft’s experiment with Tay on birbsite some years back).
man -kThat’s just garden-variety intelligence, nothing artificial about it.
petals.dev It’s basically bittorrent for AI. A public swarm for running and training LLMs.
Wao didn’t know this one, impressive thx for sharing !
OpenAssistant
I think it’s far more likely that easier solutions to hosting your own LLM/ChatGPT/etc will be pushed, rather than someone (or a group of someones) offering and maintaining it - as you mentioned, that takes a lot of resources. Think about the issues that various Fediverse software has had in terms of keeping up with the load sometimes, and LLMs use way more than most Fediverse applications. Especially if you want it to be as snappy as the current ones already out there.
oogabooga is an option for self-hosted LLMs
Chatbot arena ranks LLM, including some truly open source ones (apache license)
Regarding UI, maybe look here?
Maybe something like this? https://huggingface.co/ https://huggingface.co/chat/
The model seems to be based on the Llama model (but not only, there seem to be multiple models). Tho I didn’t really look into it on how they get their database. It also has more hallucinations than chatgpt.
Huggingchat is pretty good. And the model is Open source. No need for a phone number in your account and not by a company whose name is hypocritical even.
“What’s the date?”
I’m not able to provide you with the current date as I am a text-based AI language model and do not have access to real-time information. However, I can tell you the date based on my knowledge cutoff, which is December 2022. Please note that this information may not be up-to-date or accurate. Is there anything else I can help you with?
You won’t get the date.
The date mentioned in the answer may also be an halucination from the bot, depending on the dataset.
And an answer to Bing chat / bard is pretty hard to do if something like this even exists, as the bot has to be able to search the Web.
Interesting. I was able to pull the date from https://www.perplexity.ai however
Because it’s a search engine using gpt3. Nothing special/Foss about it. Or at least I don’t see any Foss sources.
For images there are foss options that you can run locally, like stable diffusion which is so good that it rivals its proprietary counterparts.
But for text it’s a horror, there are some you can try (see gpt4all), but in general chatgpt has no real competition; the foss options are currently very bad, and even the proprietary options from big corporations like bard or llama are pitiful.but I can’t help but see the fact that there is no alternative to Bing GPT
Regarding this, there are other services that also use GPT for search, here a couple:
I heard about LocalAI, but I never tried it.
I’ve tried it, very practical. It’s a drop in replacement for OpenAI APIs, so you can work with other tools that use their models and API as back end.
