HuggingChat is the best open source Chat GPT competitor | Hugging chat - Beginners tutorial

hello everybody in this video I'm going to talk about hugging chat which is hugging Face's new chat interface now if you did not know already hugging face is one of the main open source AI communities which has a lot of EI chat models available to everybody and the current version of hugging chat uses open assistance lava model which is accessible on their website when you use openassistant.io alright let's get started on hugging chat now the first thing I wanted to mention here is that any of the results that you see from the testing today is more a reflection of open assistance lava model versus hugging faces functionality itself as you see in the bottom the model being used today is open assistance llama model which is using 30 billion parameters and the generated content may be inaccurate or false but I think keeping up with the spirits of the open source Community this is one of the best models to use versus GPD 3.5 or any of the other models from other companies which are not open source and now if you click on the Privacy button in the bottom here you're going to see this important message that in the v0 version of hugging chat all the messages are only stored on the front end so it's not going to be used in any kind of research or model training usually when you're using something like chat GPT all the queries or prompts that you're making are being used to improve the model itself but every time you use hugging chat all the results are actually just stored on your browser so if you clear your cookies all the results will be gone and actually you don't even need to create an account with hugging face to use this functionality the next thing to mention here is that the code for the entire UI is actually open source so if you click on this link right here you're going to see that you can download this entire front-end code and use it locally for a different model which is not open assistance llama model so there is definitely some space for experimentation if you're not happy with open assistance model that is being used in chat currently and if you do get stuck there is a place here where you can post any queries that you have or any kind of suggestions that you have for hugging chat alright now let's try out the actual chat now I personally prefer this dark theme but if you prefer a more lighter version of the chat you can click this theme button and toggle it according to your needs now like I mentioned you don't need to log into hugging face for this functionality all you do is click this new chat right here and your chat interface is ready to go and remember since this is not from openai the model is not restricted to data up to 2021 so let's start with this question who won the 2022 football World Cup all right it says I don't remember who played but it was France now that sounds incorrect they were playing very well last time I saw them on TV is there something else I can help you now you might think that it just got the answer wrong but if you look at the title it actually says Argentina defeated France and that is correct so that's interesting that the model gets the actual response wrong and gets the title of chat correct so that's something that needs to be worked on with open assistance model and this is just interesting how it came out here now let's try something else related to the World Cup who played the quarter finals in football World Cup quarterfinals and it just stops so that's not good let me try this again which teams played the quarter finals so I don't think the model is doing that well with actual news so continuing with my recent videos about my newsletter I'm gonna say write me a newsletter format for my AI YouTube channel says Dear subscriber name we hope this email finds you well if you wanted to update you on some developments happening at my YouTube Channel we are launching several exclusive online courses well I don't have any courses yet maybe in the future so this sounds pretty relevant but it seems to have cut off the last word here I don't know why let me see what it does if I say please continue all right it seems to be adding a few more paragraphs and this time it comes with a proper ending of the email with my name and my channel name so combined these two messages look pretty good and I can use some of it in my newsletter maybe now of course there's some work to be done in the actual results from this model but maybe there's parts of this email you want to send to a friend or a team member now instead of copying this entire body of response in an email you can click on this button at the bottom called share this conversation and when you click on it you're going to see there's an URL which gets generated so if you copy this URL and paste it in a different browser I'm going to use Edge this time now you're gonna see that the entire conversation shows up here without having any other user log in and of course it shows this message that this application is for demonstration purposes only and when you click start chatting and you can actually continue on this conversation based on this response now I'm going to say change the email to Chinese now I'm not sure how much open assistance model is trained on different languages because usually hugging chat actually comes up with the responses pretty fast but this one seems to be stuck so it seems to have broken the chat a little bit and there is no button here to stop the actual generation I'm going to try something else make the email smaller as you see the last response is kind of still stuck and I'm proceeding to the next response and this time it says I'm planning to start an AI YouTube channel and was wondering about technical requirements and this is kind of not what I was expecting because I wanted to create a newsletter for my existing YouTube channel so if you're not happy with a specific result you can just click on this button called retry and it's going to regenerate this again this time it comes up with a proper subject line and the body came up with the subject line which says saved lives by sponsoring our AI assisted medical equipment and that is nothing to do with my YouTube channel but I'm still curious about how it came up with that response and it also seems to have cut off this email right here now let me test something else so I was in the middle of this conversation up to here when I pasted the thread on this browser and I tried this out which did not work and I tried something else which worked kind of now I'm going to click share this conversation on Microsoft Edge and when I do that now I see there's a bunch of different kinds of sharing options which show up which did not show up in Chrome I'm going to share this conversation and copy this link and go back to my previous browser and see if it adds all the things that I did in Microsoft Edge and now it continues the conversation from where it left off on Microsoft Edge so I think this is a really cool functionality that hugging chat has where you can use a conversation thread across multiple browsers without having anybody to log in so let me try some coding write a simple python script generate random numbers between 1 and 100. all right it's using this random library and it's iterating up to 10 and generating random numbers change that to Java all right it says here's how you can write the same logic in Java all right I think it did a decent job I haven't tested this code out but it looks correct to me now let me try making this into a web page make it into a web page using script and react so yes all right it's using the use effect hook to fetch random numbers only once and pass them through the state all right let me try creating something for myself create a landing page for my AI newsletter using react.js it says here is some sample code that you can use to create a landing page for your AI newsletter it has a place for my logo then it says join our AI newsletters stay up to date on all things AI by subscribing to our Weekly Newsletter it has a place for your name and your email address and a button called subscribe now with my newsletter logo so I think this time it did a pretty good job with what I needed it to do now let me try appending something else on this newsletter page create payment button on the newsletter landing page it says to add a payment button you can use the PayPal package so I can accept payments on PayPal it's requiring me to add my API credentials for the PayPal API then I have a click Handler for my PayPal button then it added this button called place your order with this click Handler which actually executes the payment as it's showing right here using my PayPal API and it also adds a message here which shows an error occurred when creating payment if it does not go through but again I haven't tested this out so there might be issues here but the overall structure I think is at least 80 to 90 percent there and if something does not work out there might be a little bit of debugging to be done but it does come up with the first structure pretty easily and I think all of this here is actually usable now to end this web page I'm going to say add a link to my YouTube channel on the newsletter okay so this did not recreate the entire code again and just told me to add an anchor tag but overall I think open assistance llama model is doing a better job at handling with code than asking for direct queries like these but again this is early stages of this product so I would not be too judgmental and understand that this is still in the works overall I'm pretty happy with what it came up with at its first attempt for my newsletter body but that's all I have for today's video but I hope you understood the potential of hugging chat on how important this is for the open source community and I hope you got some value from today's video if you did enjoy it make sure to click the like button and subscribe to my channel I have a lot more videos coming up till the next one thank you so much

Channel: Testing AI

Views: 5,115

Keywords: Testing AI, AI, Technology, hugging chat, hugginggpt, hugging gpt, huggingchat, llm, artificial intelligence, llm explained, openai, open ai, chatgpt, chat gpt, gpt3.5, gpt4, open source, open source llm, chatgpt competitor, gpt 4, gpt-4, large language models, explained, ai, gpt-3, instruction tuning, llama-i, custom dataset, dolly, cerebras-gpt, gpt4 x alpaca, minigpt 4, minigpt-4, ai vision, minigpt, minigpt-4 model, hugging face, huggingface hub

Id: JjmBPuIsSiA

Length: 10min 32sec (632 seconds)

Published: Sun Apr 30 2023