ChatGPT: One Model to Rule them All - Now Even more useful

So chat GPT just got better. Now you can use the same session to create images using Dall E, perform Internet search using Bing, and do data analysis using the Advanced Data analytics tool. And they probably killed a whole bunch of wrappers and plugins for chat with your documents. Now you can do that directly within GPT4, and it's much better than anything that is out there. So in this video, I'm going to show you how to use these different features simultaneously. And in later in the video, I'll show you how to chat with your documents within GPT4. So first and foremost, you want to make sure that you have selected default, not any one of these other options. Okay, so we'll start with a very simple prompt. Brainstorm a few names for my fantasy football team with a rock theme. Okay, so GPT4 came up with some interesting names. I think I like the third one, Touchdown Tarts. Okay, so I said I like the third one. Can you help me design a logo? And it directly started using Dall E3 to create images. Okay, so I actually like the second logo that it came up with. It's pretty awesome. So I'm going to tell it I like the second option. Okay, so it came up with another variation, but I still like the original one. Now let's see if we can use the browsing feature within this chat. Okay, so let's say I want to create a website for my fantasy football team. So I asked GPT4 what would be a good domain name for that and it came up with 10 different options. Now let's assume I like the Second1, Todd Touchdowns.com but now we need to see if this domain name is actually available or not. So for that we're going to be using the Bing chat. Okay, so what I did was I, I said I like the second option. Can you check if the domain is available? And now it's using the browsing feature to look, look up that domain and see if it's available or not. So now as you can see, you are able to use different tools within the same chat. It used to be that you could, let's say, use only dall E3 within a chat. But then in order to use browsing feature, you would have to start another chat session. Right? But now you're able to do everything within a single chat. And I think this makes GPT4 a lot more useful now. Okay, in this case it says I wasn't able to directly check the availability of the domain due to restrictions accessing specific domain registrar tools now in this specific case, it wasn't able to you to access or look for specific websites because you will need the browser capability, but it can do search for you. Now, one more tool that you can directly use within this chat session is the Advanced Data analytics or previously called code interpreter. Now, currently I'm going through this paper, Zephyr direct distillation of LM alignment. I'm currently testing the Zephyr 7B beta model that was just released yesterday. So let's see if we can use this Advanced Data analytics to chat with this paper. Now I uploaded that paper in here, so I'm going to ask is what is this paper about? Now, to be fully transparent, this is not a new feature. It has been part of the Advanced Data analytics, but it seems like they are specifically highlighting the ability to ask questions from your documents. So the model might have improved. And that kind of shows in the answer that I'm getting because for this specific question it came up with a pretty good summary of what the paper is about. So it says the paper focuses on enhancing the intent alignment of chat models through a method that utilizes AI feedback and distilled direct reference optimization. So it actually is able to give us a really nice summary of what the paper is all about. Another thing that I noticed was this. So now it gave me a reference. So you can upload multiple files and it seems like if it generates an answer based on a specific file, it will also give you references on which file it's referring to or using for generating a specific answer. So that is pretty neat. So subsequently I asked what is the performance of Zephyr 7B on the empty bench and Alpaca evolved data set and it's able to extract the information from a table. So we are specifically looking at this table. This table has basically the performance of Zephyr 7B on these two benchmark data set as well as the performance of other models as well. Right. So I think it's looking at that data set and then it says it outperforms this. I think Zin LM7B and is competitive with large models such as llama 2 chat 70 build, scoring better on empty bench and within 2 standard deviation on alpaca. Right. So if you look at here, it outperforms some of the 7 billion model. But then it goes on to say that Z4.7B performs worse than the Visit LM70 build. And there's other 70 build models suggesting that applying distilled direct preference optimization to larger models might be necessary to match their performance. Right. And if you look at here, at least this visit, LM is actually outperforming Zephyr 7B on the empty benchmark for the other one. We don't have scores for empty benchmark, but we have those for the Alpaca Evolved data set. Now then I asked it to provide me the actual scores in a form of a table and it's able to do that. So this is pretty impressive. I think none of the other chat with PDF services that I have used, they are able to do something like this. It seems like OpenAI is taking some inspiration from Amazon. OpenAI is looking at all these different wrappers that are being developed by the community and then look at which one are the best ones. Right. And then they are developing those within GPT4 or chat GPT interface. And it seems to be doing pretty good because you could not only chat with their document, but you can use something like the Advanced Data analytics tool or a code interpreter to do further evaluation on that data. So this is really powerful. Now you can do the same thing with the other file types as well. For example, if you have text or CSV files, you can do similar analysis in here. Now one great feature which I think is going to be very useful for people is that apart from the ability to analyze the documents, you could use the Bing Search to look for the most up to date data as well. Right? So for example, if you're working with a CSV file and has data, let's say till 2022, and you do some analysis and then you want to look at the more up to date data. So you might be able to use bing search within GPT4 interface to retrieve that data. Okay. Another update is related to the training cutoff data. So it seems like the training data has been updated until September 25, 2023. That means by that means that on the day of the recording of this video, we are just one month behind in terms of the latest information that ChatGPT has. Overall, OpenAI is doing really good job at adding these amazing new features. Let me know in the comments section below how you are planning on using all these features combined in a single chat session. If you found this video useful, consider liking the video and subscribe to the channel. Thanks for watching and as always, see you in the next one.