For everyone not speaking french: "le chat" in french translates to "the cat".
So I guess that's a play on words that works given the context and probably the mostly not native french audience.
True, but here, the insistence on using "le" (instead of just "Mistral Chat"), the "Adoptez le Chat" (adopt the cat) wording and if you go further, the little cat head inside the "M" logo makes the pun very obvious.
It's a way of using the English term "chat" while using the typically French article "Le".
It reminds me of old American cartoons that abused the article "Le" to emphasize the French aspect of a character or place.
The Mistral products "La Plateforme" and "Le Chat" follow this principle.
Yes in that context it's pretty clear, but if you encounter the word "chat" in any french text, wouldn't you normally assume that it means the animal?
I have a similar example in german: The word "Fee" in german means "fairy" and when we used it to promote a product in the past, someone who came from a financial background read it as the english "fee" and was pretty confused about that sentence :)
I can't speak French but I know enough to navigate and mostly guess what is being said. I'm so delighted - knowing even a little of another language is still rewarding. One shouldn't be put off by the difficulty of becoming fluent. I may never be fluent but I still think it opens the world up to me that bit more.
Same here. When I try to log into Le Chat, I get a message saying “You are on the waitlist. We will let you know when you can start using the app.”
I have been using the Mistral API for the past couple of months, and earlier today I updated my links to use the new endpoints [1]. The small and medium models worked fine through the API, but for several hours the large model failed to respond. When I tried the large model again just now, though, it was working.
It’s evening here in Japan, so I’ll start evaluating the large model’s performance tomorrow.
Does work for me, just registered with a Google Account. Perhaps they have troubles with the load and throttle the user inflow because of this. Or could be location-related, I joined from Europe.
Very excited for them. Tried it and works quite well. Very evident that each of the foundational model developers need to offer a free chat(till now) interface for RLHF/DPO to improve their models. It's costly for them to run servers, but even higher opportunity cost if they don't do this and offer a free service. Intuitively, I think GPT-4 was very good with various questions because they used the early data from ChatGPT to further finetune/align the model with a good sample of real world queries, the kind only available to them.
As an aside, most of these models are SFTed on GPT 4 generated q&a pairs (if not pretrained on them). When you ask these kind of questions[1], they spill the beans fairly easily and say they are a model of openai. Funnily enough if you ask they are a model trained by Anthropic, it refutes it immediately and ask correctly about being a mistral trained model. Same question about openai, it gives the answer in scr.
[1]: https://imgur.com/a/fGEHXH8
PS: First question was to test the reasoning. A quantized model or smaller models answer that as 4 not 5. Mistral answered correctly from a badly given question.
Don't think it adds a lot of value. One, the instruction-following models do not create high quality q&a pairs, chats do it by default. Two, at the time the usage was very low compared to ChatGPT. I am fairly confident, one month of people trying all sorts of things w ChatGPT gave them a lot more data - both in quantity and range - compared to API usage previously. Third, they always maintained they do not train on API inputs, and I think it's fair to assume that. They do not get feedback whether the output was any good, so it's always dicey to feed that directly to a model without those signals.
This can be (could be) a total hallucination. There may be some training data from OpenAI but outside it responding that way, except for me aforementioned point, where is the proof thus far?
For everyone not speaking French: le means “the” so this is “Mistral the chat” or maybe “The Mistral Chat”. My knowledge of French is limited but it feels right to immediately get ahead of any misunderstandings here for English speakers
I think “le”/“la” is one of those French words most (well-)educated English speakers already know. English uses French phrases containing it anyway (e.g. avant la lettre)
I know almost zero Spanish, but I know what “el” means. And I know almost zero Arabic, but I know “al-“ and “ibn”. I think that level of (extremely minimal) knowledge of Spanish and Arabic is rather common among Anglophones; and “le”/“la”/“un”/“de”/etc is the same thing for French
55 comments
[ 5.1 ms ] story [ 107 ms ] thread(chat, j'ai pété fort. ChatGPT4).
GPTOPIDQ
8 year olds will laugh at anything.
https://imgur.com/a/PCqFjOQ
(apologies to Geluck)
My first reaction as a French native speaker was to read it as "the chat", considering that "chat" is a well-known Anglicism in French.
Le Chat Noir_
as an inside joke!
https://mistral.ai/news/le-chat-mistral/
I have been using the Mistral API for the past couple of months, and earlier today I updated my links to use the new endpoints [1]. The small and medium models worked fine through the API, but for several hours the large model failed to respond. When I tried the large model again just now, though, it was working.
It’s evening here in Japan, so I’ll start evaluating the large model’s performance tomorrow.
[1] https://docs.mistral.ai/platform/changelog/
As an aside, most of these models are SFTed on GPT 4 generated q&a pairs (if not pretrained on them). When you ask these kind of questions[1], they spill the beans fairly easily and say they are a model of openai. Funnily enough if you ask they are a model trained by Anthropic, it refutes it immediately and ask correctly about being a mistral trained model. Same question about openai, it gives the answer in scr.
[1]: https://imgur.com/a/fGEHXH8 PS: First question was to test the reasoning. A quantized model or smaller models answer that as 4 not 5. Mistral answered correctly from a badly given question.
I know almost zero Spanish, but I know what “el” means. And I know almost zero Arabic, but I know “al-“ and “ibn”. I think that level of (extremely minimal) knowledge of Spanish and Arabic is rather common among Anglophones; and “le”/“la”/“un”/“de”/etc is the same thing for French
[1] https://en.wikipedia.org/wiki/Le_Chat
It might actually happen soon cause they have a deal with Microsoft in place.