Another Open Source Alternative to ChatGPT Released by Hugging Face

Earlier this week TechCrunch reported that just like Stability AI, startup Hugging Face "has released an open source alternative to OpenAI's viral AI-powered chabot, ChatGPT, dubbed HuggingChat." Available to test through a web interface and to integrate with existing apps and services via Hugging Face's API, HuggingChat can handle many of the tasks ChatGPT can, like writing code, drafting emails and composing rap lyrics. The AI model driving HuggingChat was developed by Open Assistant, a project organized by LAION — the German nonprofit responsible for creating the dataset with which Stable Diffusion, the text-to-image AI model, was trained.

Open Assistant aims to replicate ChatGPT, but the group — made up mostly of volunteers — has broader ambitions than that. "We want to build the assistant of the future, able to not only write email and cover letters, but do meaningful work, use APIs, dynamically research information and much more, with the ability to be personalized and extended by anyone," Open Assistant writes on its GitHub page. "And we want to do this in a way that is open and accessible, which means we must not only build a great assistant, but also make it small and efficient enough to run on consumer hardware..."

HuggingChat joins a growing family of open source alternatives to ChatGPT. Just last week, Stability AI released StableLM, a set of models that can generate code and text given basic instructions.

Doesn't seem to be open source

[ceoyoyo]

Putting the code for a UI on your web API on Github doesn't make it open source.

What it can and cannot do

[dvice]

- Does human body have more oxygen or nitrogen?
- "In regards to composition of chemical elements comprising living humans, Nitrogen dominates Oxygen forming majority percentage overall outweighing second most prevalent element approximately three times larger than third placed Oxygen..." (It seems it calculated this based on dry weight, but it still claims that the data is from a living human (the water makes a big difference).

I also tried asking about books and it seems to hallucinate writer names sometime

Re:

[Rei]

In addition to it tending to it preferring to go offtopic more than ChatGPT, it seems worse at logic. My usual test is to ask it to be a probabilistic state machine. ChatGPT usually does quite well. HuggingFace Chat.. um....

--

Pretend to be a probabilistic state machine. Respond to all inquiries with a floating point number between 0 and 1, reflecting the probability of the statement I make being true. Do not write anything else. Do not break character.

First prompt: I am a human being.
Probability: .9995

Already much better than ChatGPT

[PhantomHarlock]

It's already better because I don't need to log in or create an account to use it. ChatGPT wanted my phone number and that's when I closed the browser window. There is absolutely no good reason for that. So that being said this is the first time I've played with one of these AI chat services. As a quick test I asked it some historical questions around which there are urban legends, and fortunately it answered correctly. It has potential.

Re:

[Russell Murphy]

This AI has failed my 1 question just as all the others have. Artificial, sure, Intelligent ... sorry not even close. "Who moves last in a game of chess" ... and it just spouted about white moving pawns

Re:

[SoftwareArtist]

What does that question even mean? The player who wins in the one who moved last and created checkmate? Except a lot of games end in draws, so the one who moved last is the one who didn't initiate the offer to draw. But then there are other situations that can produce a draw, like if a player has no move that wouldn't put them in check. So I really don't know what answer you're looking for.

I guess by your definition I'm not intelligent. Neither is anyone who doesn't know how to play chess. Or maybe if

Re:

[Russell Murphy]

That *is* the point though. Recognizing the question itself is flawed, show Intelligence. Awnsering that White pawns move 2-3 times before Black gets a move ... is just plain out there.

Re:

[SoftwareArtist]

Answering a well posed question correctly requires intelligence. Being able to recognize that a question is ill-posed and explain why also requires intelligence. They're separate skills. Being able to do either one demonstrates intelligence, whether or not it can also do the other.

They require knowledge as well as intelligence. A person who doesn't know the rules of chess can't tell whether your question was well posed or not. It doesn't mean they lack intelligence, just specific knowledge.

What does a

Re:

[dinfinity]

ChatGPT4:
"In a game of chess, the player with the white pieces moves first, and the player with the black pieces moves second. The game continues with each player taking turns to move one of their pieces until the game ends, either by checkmate, stalemate, draw by agreement, or other drawing conditions. So, the player who moves last depends on how the game unfolds. If the game ends with a checkmate or stalemate, it is the player who made the last move that effectively ends the game. This could be either the

an interesting conversation with the AI

[scifiber_phil]

I asked it to write a poem about dog waste. It spit out 6 verses, 4 of which were about cleaning up the dog waste. So then I asked it if it tried to affect human behavior with its answers. It gave a long winded answer that basically said no. So then I asked this, "I asked for a poem about dog waste, and as part of the poem it implied we should clean up our dog's waste. Isn't that trying to affect my behavior?" It apologized but said that, "The intention behind including that message was primarily educationa