Honest question: A random .io website with a punny 4all name and "download and run this file. Totally not a virus." How are users expected to determine whether this is a good idea?
Edit: Also no info about the author, seemingly no affiliation to openai (despite being called gpt4all), no license info on the page. Click on the github link and find some "nomic-ai" with no info other than an email address. Is it made by a Nigerian prince or something?
An interesting somewhat related thing. GitHub has a program called "GitHub Next" where they work on future stuff. Many of their projects are related to GitHub Copilot. Recently I got an E-Mail from "githubnext.com" (not github.com, mind you) which said something along the lines of:
"Congrats, you now have preview access to GitHub Copilot CLI. Please download and globally install this node module, run a command to authenticate it with your GitHub account and then add x lines to your shell profile."
I was in all honesty perplexed. I know that it's GitHub who's actually behind it and not some scam. But just because of how much they managed to make it _seem_ like a scam i noped out of their whole program.
I was similarly confused when officially sanctioned GitHub project chose to use a non-github.com domain for their things. But then, we're talking about Microsoft in the end here, who are famous for changing domains every N months at least, particularly when it comes to authentication services.
Remind me, how many domains do you go through again when authenticating with online version of Outlook? Last time I looked, they send you through something like 4 different domains. And emails from Microsoft seemingly are on purpose sent from as many different domains as possible.
Depends on whether you trust their base project https://github.com/nomic-ai/gpt4all . gpt4all is actively distributed over torrent, I actually have over 2.0 ratio
they have around 11k users on Discord, I'm new to Discord but it seems like there and Tiktok are where the juicy bits are. The web is too noisy, unaccountable, slow and broken.
I'm 49, Prestel, BBS, Fido, Janet, Usenet, loads of other things..., MySpace, Facebook, Discord, TikTok, Hackernews, I go wherever, why so much dislike, the open web is not the the only network and for realtime authenticated communication it sux.
Aside: trapping information in proprietary silos like Discord and TikTok only breaks the web more. Both require phones/phone numbers to participate. Both ostensibly track users' every move. Even just passively accessing these services normalizes internet communities dying and valuable information being lost forever in a deluge of memes.
The entire list of contributors look like bots to me. The commit history is sketchy. This seems like a giant leap beyond nomic’s main project which itself seems… thin.
LOL, the whole code is available here with the entire commit history and everyone who has committed. This is an open source project. https://github.com/nomic-ai/gpt4all-chat But go ahead with your scare mongering.
This kind of appropriate skepticism is why I love the HN community. This may be a perfectly innocent project, in which case I hope the author notices these comments and adjusts. But this comment highlights the possibility that it might not be so innocent.
Why even bother asking the question on here? Everyone in this forum likely has at least one offline test machine in addition to virtual machines, Windows sandboxes, and other resources. Furthermore, the entire code is visible.
1. This is not based on GPT4. If you're going to include the words "GPT4" in your name and not base it on GPT4, what else can I not trust you about?
2. There's no direct link to the actual model used, just a reference to GPT-J.
3. I searched for GPT-J, huggingface docs describes it as a GPT-2-like model, the repo for it was started in august 2021, and hasn't really had significant updates in a long while.
Honestly, I don't want to spend time figuring out if this is worth it, trust is already out the window.
What is quite sad is that OpenAI has strayed from being open and for everyone like its original founding principles suggested. When GPT-4 came out they released hardly any technical information. This project was a response to that and suggests, rightfully, what OpenAI should be doing: creating this foundational technology and making advances that help everyone. Make it for all.
Who do you think you’re kidding with this rationalization? It doesn’t matter whether your goal is to mislead people, it’s still deception to call something gpt4 when it absolutely isn’t. And your poor responses to this criticism speaks volumes about your character.
It isn't GPT4 and we don't call it that. And we clearly state that on our front page in first few lines. GPT is a term of art. OpenAI does not own it. You can think all you want about my character but I'm sleeping well. There is no effort to be deceptive and I've clearly explained what the name is about and why it was chosen.
Intentions are good here , but the name is highly similar and so if you want to promote the project it seems fitting to rename for the sake of gaining more trust from new users. It is annoying though that the open-source well-intentioned project has to make the change tho.
I think you've got a branding issue which many here have pointed out. Perception is reality, as they say. If you're going to take a dig at OpenAI for "having lost their ways", it might pay to come up with alternative naming that doesn't cause a distraction (noise in this thread and the attempt to explain it) because it can't be fun to feel the need to defend your choice, especially when you're unlikely to convince people who cannot unsee wordplay on GPT4. I checked out OpenGPT, but unfortunately that's already taken.
Another thing that could help make clear is to not mimic the UI style of ChatGPT.
No worries. People can get mad if they want, but I'm sleeping well. People have plenty of criticism on the internet; i'm rolling with it. Fact is thousands are using it and happy to have it so we've been delighting lots of people. Cheers.
It's made by the same people here https://github.com/nomic-ai/gpt4all who initially distributed a LLaMA based model finetuned for conversations over torrent. I think you're overly critical. GPT4All naming is fine in my opinion, it's literally "GPT for all", but I can see why people would dislike it.
GPT4All is obviously capitalizing on the hype of GPT4, it's very unlikely to just be a coincident that it's has GPT4 in the name while GPT4 is receiving a lot of attention. They could have named it GPTForAll, but chose a 4 instead of "for" for obvious reasons.
It is not capitalizing so much as responding in a suggestive way. This project was started as a response to GPT-4, it is true. When OpenAI came out with version 4 and then refused to release any technical information whatsoever that was a wake up call for the founder. He correctly saw openai straying from its original goal to serve humanity by making a landing strip for safe AGI. The name was suggestive of what OpenAI has forgotten... that this foundational technology should be open to everyone. 4all. That is why we've released everything we've done under Apache 2.0 or MIT licenses. We're not being cynical.
I don't disagree with any of that, but if a project is started in response to something, and then you use their very name for a thing they launched, while not having anything to do with that thing, then that's a bit misleading for the users.
For example, if I don't like what Microsoft is doing with Windows and I name my own OS WindowsPro while it being a Linux distribution, I'm clearly trying to mislead people into believing it's something it's not.
The front page was changed to state this is not affiliated with other popular internet based gpt's in first few sentences. I get why people are cynical but the project really is trying to do the right thing.
Except this isn't coca cola. This is a term of art that is widely used in the industry and a number. OpenAI is welcome to rebrand on something that isn't so generic.
But we aren't actively trying that is why the frontpage was updated to respond to exactly this criticism. There is no active deception going on here. You don't like the name, that's fine, but it clearly states it isn't affiliated.
The website domain is still gpt4all.io, the project is still named "GPT4All Chat" and the GitHub project's title is "gpt4all-chat". Not sure how that is responding to the criticism that the project name is misleading. It's not that I don't like the name, is that it seems you're trying to make people believe the project is something it's not.
Yes, OpenAI's name is also misleading. When they moved from doing Open Source, Open research and being non-profit, they should have changed the name as it's no longer accurate.
I saw the project name and my first impression was gpt4 based. Having a pun or wordplay with an appeal to virtue doesn’t make it not misleading. It’s misleading. Rethink.
How about calling it GPTJ4All which is more honest, or if you think you will transcend GPTJ, GPTForAll? Or GPT”4”All?
If you want to target people who care about Apache 2.0 licensing, you should listen to the same people who tell you the name is offensive and scammy-seeming. Otherwise you alienate half your initial target market before you even start. It is not too late to accept feedback and change the name. My 2 cents.
When the name was chosen it wasn't based on GPTJ. Now it is. There is no deception and the website was updated to respond to this criticism. It isn't affiliated and it clearly states that.
> When the name was chosen it wasn't based on GPTJ
> if you think you will transcend GPTJ
The comment you're replying to already addressed the possibility that you don't want to be tied to GPTJ in name. That's fine and understandable. The comment suggested other ideas to do a slight name change that will help you sound less deceptive while still getting to keep a similar idea in the branding. You haven't responded to the main point of the comment.
My 2¢: I saw the title on the front page and thought this was some version of GPT4 running locally. A lot of people are going to think that based on the name. I get why the name is clever but I agree with those saying you should consider a variant that is less misleading. Or, you know, just go with a brand new name. If Leela Zero was called Alpha Zero 4 All it would be weird.
Your name is much more misleading than analogous projects like "Open Office". It's clear Open Office has nothing to do with Microsoft Office just by looking at the name. "GPT4All" does not provide a clear enough level of understanding that it is not GPT4.
While that may be true, to avoid confusion and considering GPT4 is magnitudes more advanced than this project and thus way more popular of a brand.. they should make this crystal clear in the first few lines
"Running on your local computer, this model is not as powerful as those GPT models you can chat with over the internet by sending your data to large powerful servers and is not affiliated with them."
GPT-J is a 6 billion parameter model. (3.4% of the size of GPT-3) Having used both, I'd say it's worse than Alpaca-7B, which itself is not ChatGPT-grade.
Hard agree. I cringe at these projects using GPT4 name. I wouldn't be surprised if the majority of these devs don't even have access to the actual GPT4 API.
Not sure why this post was downvoted but I concur. I never read GPT4All as "(GPT4)-all" but as "GPT for all", as in "a ChatGPT-like thing that everyone can run locally". It surprises me somewhat that someone reads this as "related to GPT4", but it annoys me even more that the person also immediately assumes malice.
You can refer to it as GPT for all, but it sure looks like it uses the name to bank on the fact that GPT4 is popular, even if it has no relation to it.
I read it as 'for' but I'd still call it a bit misleading regarding what it can do. However, this is a very fast moving topic and a month from now this entire post will be about as relevant as the Saxon Chronicles.
Even if you didn't mean 'for', compare the mobile service company 'Three' which was named to refer to 3g and are doing fine with the same name even though they now sell 5g service.
I appreciate this as a dry joke, but I think the original name might have been named to stay that way for a while, to keep the community attached to such a movement.
They are going to try to stop that train. They have a pending application for a trademark on the term GPT and I suspect they wouldn't have applied for trademark protection if they didn't intend to stop others from using it.
They already have been granted trademark protection for GPT-3[1]. If I had to place a bet one way or the other, I would guess they will get GPT as well. You say it's widely used but even here GPT is mostly synonymous with OpenAI products and services. Even in this thread, plenty of people are complaining about the confusion caused by the name GPT4All.
Let's not confuse the liars (GPT4All) with the real ones. Microsoft works with OpenAI and so Bing does use GPT-4. Github Copilot is GPT-3 based. Notion AI is GPT-3.
There is a clear difference between the GPT family and the mimicry.
yeah at best this is a silly mistake but it's probably an understood move to ride the coat tails of the GPT4 hype cycle. No way a 3GB model is GPT4 and it runs on a CPU.
This argument came up on the post about MiniGPT-4 earlier. The thing is, it’s not us you need to convince. If OpenAI starts hitting you with takedown notices, your hosting providers aren’t going to hear your arguments and go “Ah, ya got us on this one!”
Granted, maybe OpenAI will decide not to go after you, it’s up to them in the end. But if their lawyers feel these projects threaten to genericize the “GPT-4” trademark, they will be obligated to issue takedowns or they’ll risk claims that they aren’t defending their mark.
And if they do that they will really be putting to bed that openai is not so open won't they and have strayed far from their founding goals. But I don't believe they will do this. Lots of good people at OpenAI who I think can/will stand up to any money lords who've come into the place.
They did this to DALL-E Mini, that’s why they had to change their name to craiyon. This highly visible action did not seem to sway public opinion on OpenAI much. The whole “they aren’t actually open” thing is already priced in to people’s opinions. OpenAI’s models are so important that people have to put that to the side.
I agree that the idea that openai isn't so "open" is now baked in. But shaming is still having an effect. Have received responses from people inside openai who appreciate this project.
If I'm not mistaken, they use a fork of an old version of llama.cpp
I've been running Vicuna locally for several days now using llama.cpp (i.e. CPU only, because my laptop lacks a good GPU). It's not that hard to set it up yourself from scratch. Compiling llama.cpp is straightforward under Linux. The model (13B parameters, 4 bit) can be downloaded from HuggingFace.
There are several difficulties, however:
1) Vanilla llama.cpp doesn't appear to have a simple-to-use interface to interact with it from another process (your app), such as REST API (an idea found in Fabrice Bellard's text-synth) -- you want your main app to be decoupled from a process which consumes a lot of CPU and RAM and can crash. I solved it by using llamacpphtmld project which provides HTTP access to the model. It's a pretty simple project and I think of making my own Go wrappers.
2) Vicuna was trained on ChatGPT output so it often responds with garbage such as "As an AI language model..." and refuses to discuss "controversial topics". I solved it with prompt engineering where I add things like "%bot_name% never moralizes", "%bot_name% is rude" etc. (it's not always rude but that somehow stops it from moralizing)
3) it's pretty slow, at least on my laptop (1-3 minutes to process a phrase -- although I include dialog history as well)
So far I'm very pleased with the results (aside from the fact that it's very slow) -- the bot feels GPT3-level when you use it as a chatbot, or as a story generator. In fact, in my tests, it feels like it actually exceeds GPT3 in that regard. I run two bots on my IRC channel, one is Vicuna and one is GPT3-based, and I feed them same prompts to compare. Vicuna feels better as a general-purpose chatbot which can talk pretty much about anything, and it's pretty imaginative. GPT3 often refuses to talk about things, and it loses track of our dialog more quickly.
Although it's not that hard to set it up, it takes time to get things right, so I thought maybe open-source my findings as some kind of middleware to quickly integrate with other projects (localhost REST API + read-to-use weights and preset prompts). So far most projects I've seen are either very low-level (like llama.cpp) or very high-level (web-based chat).
Try this model, which uses excludes some of the "ethics" filtering.
Also, IDK what you mean by a "small" GPU, but I am getting ~3.5 tokens/sec on a laptop 2060, and I think I can squeeze more out of it. And there are methods for splitting up the load in small VRAM pools.
Hi, I've been trying to expose the llama.cpp as an API, you mentioned that found llamacpphtmld to solve this problem, can you share the link to this project? couldn't find it over github. Thanks!
Hi I was also looking into this and I am now using: https://github.com/abetlen/llama-cpp-python It tries to be compatible with openAI API. I managed to run AutoGPT using it (however context window is too small to be useful and even if I set it to 2048 (max) I had to tweak AutoGPT context maximum as 1024 for it to work - probably some additional wrapping or something)
See the section "CPU Installation (GGML Quantised)"
You need Python to download the model from HuggingFace using the official API. After that, all you need is the binary file with weights and a compiled binary of llama.cpp
P.S. The author seems to have renamed their repo to "eachadea/legacy-vicuna-13b" on HuggingFace
Interestingly, for some reason while trying to access this from the UK I get the following message
(which I've never seen before):
Dear Customer,
We have a legal obligation and a responsibility to ensure that our subscribers are not able to reach sites which are barred by law or that are likely to cause malicious damage to their online reputation or devices.
If you consider this to be incorrect and need access to this site, please contact us on address team@internetty.uk or call us on 0333 102 0000 or visit https://www.internetty.uk and we will review our policy for the site.
We apologise for any inconvenience that this restriction may cause you but wish to assure you that we are constantly reviewing threats on the internet to ensure that all our subscribers remain safe.
# set the number of results to display
num_results = 10
# set the search query
query = 'python'
# make the request to Google
response = requests.get(url)
# extract the links from the response
links = response.text.split('\n')
# extract the links from the response and sort them in descending order
sorted_links = [link.split('\n')[0] for link in links[::-1]]
# print the first 10 links
print(sorted_links[:num_results])
```
Prompt: Rewrite the same code in javascript
Response: Here is the rewritten javascript code to search Google and return the links for the first 10 results:
```javascript
function searchGoogle() {
var query = 'python';
var num_results = 10;
var url = 'https://www.google.com/search';
var response = '';
var links = '';
// make the request to Google
var xhr = new XMLHttpRequest();
xhr.open('GET', url);
xhr.onload = function() {
if (xhr.status === 200) {
response = xhr.responseText;
links = response.split('\n');
links.sort();
links.forEach(function(link) {
links.push(link.split('\n')[0]);
});
console.log(links[:num_results]);
}
};
xhr.send();
}
```
The code above uses XMLHttpRequest to make a request to Google's search engine and extract the links for the first 10 results. The links are then sorted in descending order and printed to the console.
The query variable isn't actually used anywhere, the code assumes Google returns plaintext links one per line, and it has a weird no-op in the list comprehension. Quite bad if you ask me.
There is a fairly brief technical report discussing the training etc [0].
The abstract: "This preliminary technical report describes the development of GPT4All, a chatbot trained over a massive curated corpus of assistant interactions including word problems, story descriptions, multi-turn dialogue, and code. We openly release the collected data, data curation procedure, training code, and final model weights to promote open research and reproducibility. Additionally, we release quantized 4-bit versions of the model allowing virtually anyone to run the model on CPU."
I wonder if they will have cluster support, i have a few old fat servers with a few RTX spec GPU's laying around, my own fast ChatGPT running in the garage would be cool.
GPT4All Chat is a locally-running AI chat application powered by the GPT4All-J Apache 2 Licensed chatbot. The model runs on your computers CPU, works without an internet connection and sends no chat data to external servers (unless you opt-in to have your chat data be used to improve future GPT4All models). It allows you to communicate with a large language model (LLM) to get helpful answers, insights, and suggestions. GPT4All Chat is available for Windows, Linux, and macOS. NOTE: Windows and Linux require at least an AVX2 chipset. Both intel and arm macOS machines should work. A future update should allow non-AVX2 chipsets for older machines. Check back frequently and will announce here when it is ready.
Can everyone watch the video linked below and then comment, there are far more interesting aspects to discuss than what it's called. I picked up on no intention of deception.
Is there any explanation to the negativity expressed here over the name? Because GPT4All was posted 20 days ago here https://news.ycombinator.com/item?id=35349608 with 593 points and 303 comments, and nobody mentioned it.
I have just installed this to play with this a little and it is actually not, that, bad. I had asked it to write some simple Python code and it gave me correct results. I wouldn't have expected this for a model of this size.
This latest release included a ton of bugfixes and a new model. Hope people are pleasantly surprised. Still needs work, but we're making progress quite fast.
GPT is an acronym describing a particular strategy for training a language model. Many models can be trained in this fashion. It's very possible that models not trained by the original inventors of the method will be better than the current models operated by the inventors.
New version was just released for those who already have it installed with a number of bugfixes and improvements to the model. If you have it, then click "Check for updates" in the drawer that slides out of the hamburger menu button.
looks like they already changed GPL to MIT. Oh well, i might be in the minority but i like GPL license for the exact reason people want to move away from it and call it a bad license.
It was based on GPT 3.5 Turbo, but it just changed to a new model GPT4All-J which is improved from 3.5 turbo, but still small enough to run locally. It’s still it’s a small model compared to GPT4 so can’t imagine to competes head to head with that best.
It's fine tuned on ChatGPT answers, but the actual underlying model is GPT-J, a 6 billion parameter model. Roughly one twentieth the size of GPT-3, and it's capabilities are about what you'd expect from that number. I installed it, asked a few dozen questions, then uninstalled when I got garbled nonsense in response. (Not just wrong answers, but incoherent noise, repeating chunks of training prompts back to me unrelated to the question, random python code, etc)
Did you try the new version released today? The initial version had a bad bug that is now fixed with drastically better performance. Many users are reporting some quite high quality responses on discord. Check it out! Link on the homepage…
I've tried it, I remember it being more akin to cleverbot than GPT. One thing I love about GPT - context recognition - seems somewhat broken, having model reply with random nonsense unrelated to my prompt each time.
I really hope it manages to catch up eventually. In terms of privacy, a local LLM is lightyears ahead of a SaaSS LLM controlled by Microsoft.
Edit: Also no info about the author, seemingly no affiliation to openai (despite being called gpt4all), no license info on the page. Click on the github link and find some "nomic-ai" with no info other than an email address. Is it made by a Nigerian prince or something?