NEW HuggingChat 🤗 - VERY GOOD ChatGPT Competitor (Open Source)

hey welcome back another day another model today we're going to be reviewing hugging chat by hugging face this is based on the open Assistant llama 30b 30 billion parameter model and it's quite comparable to chat GPT 3.5 so let's test it out and then I'm going to compare it to Chachi PT 3.5 let's get into it so this is the hugging face website as a reminder this is where you can find a ton of Open Source llms and a bunch of machine learning tools and libraries and data sets and just a whole bunch of awesome stuff if you're into artificial intelligence or machine learning so I highly recommend checking out huggingface.co but today we're going to be testing out hugging chat now here it is you can see it looks almost identical to chat GPT in terms of its interface but right here at the bottom we actually have the model it's the open Assistant Lama 30b model and it's quite good let's test it out on a few things so first let's start with something easy write me a poem about AI 50 words now it's incredibly fast I'd even venture to say faster than GPT 3.5 certainly faster than gbt4 and the quality is great too check it out artificial intelligence grows a digital entity so unlike us humans it glows with potential and power to lead the way in technology that's never been seen before today a little bit of capitalization issues but this is a good poem let's try something a little bit more difficult so we're going to go up to new chat and one of the examples is code a snake game so writing really good and longer code has been a problem for a lot of Open Source models open assistant is pretty good but let's test it out code a snake game sure here's an example of how you might code a basic snake game in Python and it's nice and formatted just like chat GPT okay so interesting we got the whole thing I think maybe it misspelled this at the end it's supposed to be running but that's okay if it still works and it didn't finish the prompt so pretty good not perfect let's test out the code though so we'll open up vs code we'll open up a new text file and we're going to call this snake dot pi and we're going to come back to hugging hugging chat we're going to copy all the code switch back paste it and let's make sure it all works doesn't have random Okay so one thing that I see it didn't do is import random so let's try that okay now we have it I'm going to click save interesting it doesn't use any code is unreachable okay so I see so the issue that we're seeing here is that it's not indented properly so it wasn't a simple copy paste but there we go and also I see that some of the method names are misspelled so here's draw Snake let's see if we can just fix that and it still does not work is not defined it is right there okay so I don't want to spend time debugging this right now but I'd say it got 80 of the way there there's definitely some bugs some misspellings but it's pretty good I'm a little disappointed that that's one of the examples that it tells you to do uh given that it's not Flawless every time but this is a big step up from a lot of Open Source models let's switch back let's give it some more tests write an email from a bullet list let's see what that does as a restaurant owner writer professional email to the supplier to get these products every week okay so I hope this message finds you well I'm writing on behalf of my establishment restaurant name we are in need of regular supplies yeah I mean this looks like a really good email I'm not seeing any misspellings from a quick glance so yeah this is good let's try the last example assist in a task how do I make Delicious lemon cheesecake this is how you can prepare an excellent and Rich lemon cheesecake ingredients you will need for 10 pieces biscuit milk eggs yolk cream cheese instructions yeah I mean this looks pretty good as well let's give it something a little bit more difficult one task I like to give are reasoning problems so let's give it a simple reasoning problem Sally is faster than Nick Nick is faster than Joe is Sally faster than Joe yes if Sally is faster than Nick who is faster than Joe then it stands that Sally is also faster than Joe really really good okay let's give it a little bit of math now what is 34 times 56 plus and I'm not going to use parentheses either let's see if I can figure it out 109 and then add five it seems there has been a mistake in copying the task can you please repeat the instructions to me correctly so that I know exactly what needs to be done without misunderstanding Thank you okay please calculate and then I'm just going to copy it again and let's see if it can do it training data goes up to 2021 interesting I cannot perform calculations okay that's not great let's try it with parentheses now see if that helps 34 times 56 plus 109 Plus 5. the results of the calculation is 2158 so let's test that out 34 times 56 plus 109 plus 5. not right okay so the answer is not right and I've found that a lot of large language models struggle with math they aren't necessarily word problems they're just pure math we'll test it against GPD 3.5 and we'll see what the difference is so let's do that now I'm going to open up a separate window I'm going to open up Chachi PT we're going to use GPT 3.5 and now let's give it this math problem first therefore the final answer is 1928 and that also is not right let's try it with the parentheses this time 2013. that also is not right so they didn't add that last five at the end it seems so both hugging chat and GPT 3.5 failed Let's test on gpt4 now so you can see it's much much slower it's explaining each step and that got it right so gpt4 gets it right on the first try let's go back to testing against GPT 3.5 now I'm going to say write a snake game in Python let's see if it gets it right the first time okay so I'm going to copy I'm going to switch back to visual studio code I'm going to replace all that code that was in here already from hugging chat and immediately I don't actually see any issues maybe something at the end there but we will see and let's play it's loading up but it doesn't seem to actually be working nope okay let's see indentation error expected indentation block after if let's see if we can do a quick fix for this I'm going to pass the error back into chat GPT 3.5 and let's see if it'll fix it sorry about that it looks like there was missing code block after the if okay right so it didn't actually output the whole thing so we'll replace that we will make sure that the indentation is right I'm going to save and let's play again oh looks like we might have got something nope I think it tried to load and then it just quit so it seems like it got close but not enough for us to actually play the game okay so I'm gonna say they both failed again at that now let's give GPT 3.5 that same word problem and see if it does it right while it's true that Sally is faster than Nick NYX is faster than Joe there may still be a possibility that Joe is faster than Sally that is wrong it gives examples here Sally runs 100 meters in 12 seconds Nick runs 100 meters in 13 seconds Joe runs 100 meters in 11 seconds that can't be right because that would make Joe faster than Nick and Nick is faster than Joe so GPT 3.5 actually failed this reasoning problem where hugging chat succeeded let me know what you think if you liked this video please consider giving a like And subscribe and I'll see you in the next one foreign [Music]

Channel: Matthew Berman

Views: 14,590

Keywords: 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

Id: UM-66lc4oxc

Length: 9min 58sec (598 seconds)

Published: Wed Apr 26 2023