ChatGPT'S New 'BROWSING UPDATE' Surprises Everyone! (Access Internet, Plugins + Examples)

So ChatGPT just released a massive update here. You can see from OpenAI, they said we're rolling out web browsing and plugins to all ChatGPT plus users over the next week. Moving from alpha to beta. They allow ChatGPT to access the Internet and use the 70+ third party plugins. So this is a major, major announcement in the world of AI because of course, as you know, this has been absolutely an incredible forefront. I mean, one of the major problems with ChatGPT is that it can't access data closer than 2021. But in today's video, I'll show you some examples of how ChatGPT's new browsing feature works and how great it actually is. Because you're going to be surprised at just how it compares to some of the other stuff online, just like Bard and like the one that already uses GPT for Microsoft's own Bing. So let's get started. So the update to ChatGPT has been pretty interesting. Many of you may not have realized that now there were two different icons. So one says GPT 3.5, which is of course available to free users, the one that everybody uses. Then of course we have GPT4. Now when you click GPT4, essentially what you'll see is this drop down menu right here and this is where the browsing tab is. Now this is default without, without browsing. And of course we have browsing here. Now I've got to be honest, when you first try to use browsing, it might be somewhat underwhelming because of what you see, but you have to understand that if you give it time, it does work very effectively. So I'm going to show you a real time Word example of an example that I did recently and the examples from Bard and from Microsoft Sping, so that you can then make a justified decision on what's going to be best for your browsing capabilities. So let me just show you how browsing actually works. So what I did was I actually asked them to recap Google's IO event in 2023. If you don't know what the IO event was, essentially this was an event where Google pretty much showcased off a ton of new generative AI tools and, and a bunch of stuff that they're launching at Google, such as new phones, new software, yada yada. So I knew that this was a recent event and I wanted to see how GPT4 or ChatGPT with browsing actually works this out. So you always want to include the date and then I'm going to click Enter to show you what this looks like. So once I click Enter, you're going to see the model is web browsing and then you're going to see ChatGPT at work. So one thing that's very interesting is that you do see what ChatGPT actually does. You see its entire thought process. This isn't something that we see with Bing or Bard. Whereas they just give the response immediately. You can see that it is actually reading the information, it's managing to click on certain link and it's managing to then extract that information. So right here, this is a great example, because the last time I did this, what I did actually see with ChatGPT, was that it was actually clicking on many links and wasn't able to find a lot of information. And this was very well after the event had taken place. So right here you're seeing that it's giving a very decent summary. And what I'm going to do as well is I'm going to show you exactly the differences in these examples, because you might not know whether or not this is good, because you can't compare it to Bard or to Bing, which I'm going to do in just a moment. So once this finishes up, I'm going to show you what Bard did, and then I'm going to show you also what Bing did as well, because there are some very interesting differences there that I think you ought to know. You can see right here that it's managed to click on a link, click on the developer's Google blog, read the content, and then finish browsing. Now, what is actually interesting is the differences in responses that I got. So this is the same prompt that I used, and I'm going to show you a different response that I got that was actually better. So when this went live, I decided to actually use this. And you can see the prompt right here. If this does load in. So this is the same exact text prompt that I just used. And it honestly did some very interesting stuff. You see right here, what actually happened was it actually decided to search through more pieces of content. And this was very interesting for me because at first I thought that this wasn't going to work. You can see it said, click failed, click failed, click failed. And this was actually quite shocking. But then it read some information, clicked on the Verge.com, read around, clicked on some other Google search results, then had some failed clicks, and finally finished browsing. Now, what I liked about this from ChatGPT from the first generation that I did get, because you can See, I decided to use the same exact prompt is that with this we actually got some very nice citations. So imagine you're doing research for a research paper. Maybe you're writing a blog, or maybe you're just generally trying to give someone a certain piece of information towards an argument, and you're trying to reference that with a reliable source. You can see right here that there are different sources of articles that you can literally include as hyperlinks, which is very, very effective in the way that it is done. It's not all posted as a link at the bottom, but but is actually posted as a hyperlink with a small number text. Now, the only bad thing I can say about GPT4 is that it is quite remarkably slow. As you know, GPT4 only allows 25 messages every couple of hours, which is of course quite the disappointment if you're going to be needing to pull together a research paper in just a couple of hours. Now, you might be wondering, how does this compare against some of the other AI softwares? Because of course we do know that this is something that we've wanted for quite some time. And as GPT's limitation was the fact that it only had information up to 2021, but how does this compare to the likes of Bard? So I took the same prompt from here and then I input it into Bard. Now, you can see right here, when I put this into Bard, it actually got this prompt within about, I think, five seconds. But the only thing about this was that it did miss some key things. So you can see right here, it missed a bunch of the AI stuff. Now, of course, all of this information is correct, but it did miss some of the AI stuff. So I think maybe what would be better is if I said recap some of the AI stuff that went on at Google's event in 2023. Now you might be thinking, why are you comparing this to Bard? Well, you need to understand just how good ChatGPT is, because you don't want to be using ChatGPT's browsing feature if it's completely terrible. And you're going to want to make sure that you're using the best AI software up to date. So I'm going to just do a different prompt and I'm going to show you how bard handles it versus how ChatGPT does. So I'm going to say recap Google's IO event 2023 is recap Google's AI announcement at Google's IO in 2023. So I've asked it to recap specifically the AI stuff because of course, as you know, we are an AI focused channel, so I've asked it to do that. And you're about to see the speed of Bard. Now, this isn't a Bard poster video. I'm just doing this so that when we go back to ChatGPT, you can see the differences in something done by Google and then something done by, by OpenAI Microsoft. So you can see right here, this was done in literally about, I guess you could say 10 seconds or so. And this is pretty much everything that there was from AI. I guess they missed a small amount, which was Gemini. But this is definitely something that we do have. Now what's also cool as well is you do get these other drafts you can see right here, new AI powered features in this and this and that. Of course. Then we've got the third draft here, which is also very, very interesting. So I do think that being able to get access to three different drafts is very, very good because it does show a diverse range of information. Now let's use this same prompt and go back into ChatGPT's browsing feature to see how effective this is. So I'm going to put recap Google's AI announcement at Google's IO in 2023 and let's see how ChatGPT does this. Now later on in this video, I'm going to show you some other tips and tricks that you can use for ChatGPT browsing because there's still a lot that you can do even if you aren't going to be researching around the web. So I'm going to show you that stuff. So it's researching on what's going on. And what's interesting is that it actually says it's thinking and then as you can see it, it's actually clicking on many different links. Now this is actually really good because of course, as you know, plagiarism is a huge problem. Now if you do reference something, you might want to of course give reference back to that official article, depending on which kind of project you are working on. So while this is going to take some time, like you guys can see right here, this may generate a better response, but you guys are going to be the judge of that because of course, as you know, this is still in beta mode whilst Bard is of course officially released and updated. So let's just wait to see how long this takes and I'll tell you guys exactly how long this entire process did completely take. Okay, so after a minute or so, chat GBT was actually able to get this right here. Now I'm going to be honest with you, this is actually quite effective because one thing that I've realized about the differences between this and Bard is that whilst this is actually quite slow, Bard doesn't actually give you any sources which you can cite, which is actually, I think, a very big feature. Because of course, as you know, when you are doing research, especially for work related tasks, you may actually want these research sources. Because of course, as you know, one thing that AI has a big problem with is misinformation. So if Bard does hallucinate, you're going to want to know if that hallucination is real or not. So I think in terms of accuracy, ChatGPT does win. Although I didn't find any inaccuracies from Bard in this because I actually watched the entire Google I O thing. But what I will say is that if you're looking for speed and you're looking to get this done asap and you don't want any limitations because of course, as you know, GPT4 does have a limited number of generations per hour, even for the plus users. I would say you'd have to go with Bart in this area. But I do think that once this does become more efficient and more, I guess you could say, comprehensive, I do think that this will win. Now what's also cool is that this does have some other potential applications. Now there was an awesome thread on Twitter which I will link in the description, but it shows many different applications for this new browsing feature. So one of them was that now ChatGPT can now summarize the latest news and cite its sources. So we one of the prompts that they had was list 10 things that happened this week in AI and then put it in a table with links to sources. So I'm going to just put that in as a prompt and let's see how this does. Now I want to see also if Bard can do this. I'm pretty sure it can't, but I'm going to go ahead and put this in Bard too. So I'm going to do this and then we're going to wait. But I do know that Bard actually can use tables, but we're going to see exactly what this is able to do. So after some moments you can see right here that I asked it to list some of the things that happened this week in AI and and then put it in the table with links to sources. It wasn't able to do this, but there were some examples on Twitter where people were actually able to get this done. So I would just say perhaps it was my prompt that got this wrong. But you can see right here, the reason this is actually so effective is because it actually reads through a bunch of articles before formulating its final answer. Now, what I did realize when doing this test between this and Bard was that Bard actually needs a very specific prompt in order to get the right information. Because when I said listen, then things that happened in AI this week put it into a table with links to its sources, it denied its ability to be able to do that until I asked it, what did Sam Altman do this week in AI? Then it gave me a very, very comprehensive overview of exactly what happened this week of him testifying in Congress. Now, this was some very interesting information, but of course, as you can see right here, ChatGPT was able to get this information done as well. So this was very interesting because it goes to show that whilst ChatGPT can work from just a very simple prompt, Bard is unable to simply, I guess you could say, browse the Internet. It needs a very specific directed prompt, whereas ChatGPT doesn't. Now, you might be thinking, how does this compared to Bing? Of course, if I was just asked them to list 10 things that happened in AI, put it in a table, you can also see that from the recent prompt that we just talked about, it wasn't as effective as, you know, I guess you could say this Bing AI doesn't really have a main purpose in terms of its complete functionality. And I would say that because it's not something that you can just completely talk to, but. But it's Also now that ChatGPT has browsing, I would argue that ChatGPT is just effectively better. But this is actually surprising as Bing is able to do this in a very effective table format. So I'm not going to lie to you guys, I would say that currently we do have a bit of a balanced one right here, because it seems that a lot of these AIs are able to effectively do a lot of the tasks as long as you manage to tweak the prompt, with ChatGPT being quite slower, but providing more efficient and more, I guess you could say, a general sense of information with users that don't know exactly what has gone on. But Bard is providing speed and depth of information for those who know the exact kind of prompt that they want. And for those who want, I guess you could say, more variations in their answers when you have all of these different drafts. So, so far, let me know what your opinions are. I would say that what is also cool is that we have the ability to see which articles that they click on. Because of course, as you know, like we said before has hallucinations are something that happens quite a lot in AI and if there is something in the statement that isn't true, we'd always like to be able to fact check it. And if there is a certain source that it's taken the data from, we'd also like to have the ability to see where that source has come from. Even if it is still true information, we'd like to see where that article data has been taken from. So I gotta be honest with you guys, overall after taking this entire test, I would say that browsing isn't a complete failure. The the only gripe that I have with browsing is that unfortunately it is quite slow. Which means that maybe if you are on a tight time schedule or if you need a bit more data, I would just go with Bing and Bard for now. Because like we said, GPT4 doesn't have a lot of prompts and it isn't the fastest when it comes to data. But if you are trying to get a vast majority of data out from maybe a specific topic and you don't know where to look, this is not a bad option. So before the video ended I decided why not conduct one more research test to see if just how effective this software was. You can see that right here I said find the top three places to get coffee in Vancouver based on the best reviews. However, after some very, very extensive research which ChatGPT was doing, it managed to click on a bunch of links and all that I saw it say was click failed, click failed, click failed. And then it returned this statement after around five minutes saying I apologize for the delay. I wasn't able to find the top three coffee places in Vancouver. Now let's compare this to the other AI software. You see that Bard was able to get me this within seconds and this was actually pretty effective because I did double check these and these are actually very good places to get coffee in Vancouver. And you can see that it's given me three different tables. Now I then also cross check this with Bing and you can see that right here on Bing on the right hand side, it also gave me three places to get coffee after around 20 seconds. So what is this all to say? I would say that this is currently in beta and I think that the problem that ChatGPT has is, is that currently it's trying to get the most accurate information, which sometimes can lead it to just being not effective at all. So far. I got to be honest with you guys, I do think Bard does take the cake here a little bit better in terms of the speed and efficiency of the software, as long as you know what to prompt. But it's going to be interesting to see if Chat GPT can improve this, because although it is quite effective, it does fall short in its speed, and in some areas it just isn't able to get that information. So maybe it was just a prompt, maybe this test wasn't long enough. But let me know what your personal experiences are, and if you've actually got access to browsing with GPT.