Exploring ChatGPT's New Plugins and Bing Integration for Research: A Frustrating Journey
Discover the potential and pitfalls of using ChatGPT's latest plugins and Bing integration for academic research. Is it a game-changer or just hype?
File
ChatGPT Plugins Are They Really Helping Researchers or Just Hype
Added on 09/03/2024
Speakers
add Add new speaker

Speaker 1: ChatGPT just got some awesome upgrades, but let's see what they're like for research. I am here in my ChatGPT Plus account and if we head up here to the GPT-4 version we can see that now we have Bing, browsing and plugins. Interesting. What are they? The first thing I want to talk about are plugins because I think they're by far the most interesting thing. So if we go to plugins, you can see here I've got no plugins enabled. So I go to the plugin store. All makes sense so far. Or go to all, you can see we've got 20 pages of these and there are plugins for nearly everything you want to do. But what about research? I'm going to supercut the ones that I think are valuable for research and we're going to give them a go a little bit later. I think a scraper would be good for getting all of the stuff from websites. HTML papers. Web pilot looks interesting. I'm having that. Access link. Access links on the web and get the information you need. Hmm, want that. Scholar AI. Unlock the power of scientific research with peer-reviewed papers from all of those places. Yeah, I want that. Chat with PDF. Yes, please. Chat with website. Yes, please. Maybe not useful right away, but open lecture. Discover and access the right moments in open courses lectures for targeted learning. Nice. I want some of that. X papers. Effortlessly find real academic papers on archive. Yes. Next paper. AI. Fetch the latest research papers on a specific topic from PubMed. Yes, please. So once we're in here, we can see that we've got all of our plugins here. Let's just enable all of them. Oh, three of three enabled. So I can only have three enabled at a time. That's interesting. So let's have a look. Let's go with Next paper. Let's look for things. So let's have a look at next paper. Let's have a look at Scholar.AI and let's look at X papers. Because one of the things that ChatGPT wasn't very good at is going and getting references. It used to make things up. It got better in GPT-4, but let's have a look and see what happens. So I've got down here, acting as a researcher, find the latest papers on transparent electrode materials. And if we click send, what you'll see is that because it's in the plugin mode, it will start to think about which plugin to use. If we open up this bit, we can see that it sends a request to next paper and it gets some stuff back here. Client error not found. All right. Well, that doesn't work. Well done next paper. Scholar.AI. So here are the latest papers on transparent electrode materials. So 2012. Not really latest, but maybe we need to be a little bit more specific. So I'm going to stop this and say, I'm going to edit this one and say acting as a researcher, find the latest papers from 2022. Let's try that. And let's try it again. You can see what it's doing in real time. Error response. Never mind. All right. Next paper.AI. The thing about these plugins is that they're third party. So some of them may or may not work and OpenAI really has nothing to do with them. But we can see that Scholar.AI is doing quite well. So let's have a look. It's got query publication date and 2022. Let's see if it's actually pulling up the appropriate stuff. It's got a DOI. We'll be checking those. 2012. Not 2022. But is it a real thing? So let's go and check out the actual DOI to see if it's a real paper. Transparent electronic materials for simultaneous blah blah blah. Okay. Well, that's good. But you can see that Oh, we've got kind of stuck. It didn't get it from 2022. So that's a bit of a shame. It really is at the very early stage of being useful. Here our network error. There are some papers and a network error. So a little bit of a frustrated experience if I'm going to be completely honest. Let's regenerate the response and see if it can do a little bit better. It really has missed out on the date thing. It's got real papers, but it's not really done a great job. I'm going to stop that because that's now really annoying me. So let's go to a new chat and let's use plugins and let's have a look to see what it's like with chat with PDF. I don't know. How do you chat with a PDF? I guess you send it a link. How do I chat with a PDF? All right, so it does need to be accessible online. So I'm going to upload it to Google Drive to see if I can make it accessible. So what I've done is I've uploaded one of my papers about planar silver nanowire, carbon nanotube and PEDOT PSS transparent electrodes. And I've just said that anyone with the link can access it. But I'm going to copy that link and I'm going to put it into chat GPT and see what it's like. So here I've got summarize this paper. Let's see what happens. So it should be able to access it. Chat with PDF, it's using chat with PDF. Great. And let's see what it's doing. It's got the PDF URL. Good. So it's actually identified that now it's asking. Oh, now it's got all of the information. That's interesting. So there's the entire paper, I think. All right. So if we close all of that, you can see it's actually done a pretty good job at summarizing that paper. So if your papers and the things you want to summarize are accessible online via a sharing link, this could be a really useful tool and better than using something like chat PDF. I really like that. It's actually pulled out the data in a really great way. Like it's pulled out all of the important bits and it's actually sort of like really gone into detail with the sorts of stuff it's extracted. Sometimes you just get bullet points in some of the other services. So a complete win for this plugin. So the last thing I want to try is actually GPT-4 with Bing. All right, then. So I've said acting as a research scientist, go find some recent papers about elephants. I'm a little bit bored with transparent electrodes. Let's see what it will do with elephants. Hmm. There seems to be a server issue. Interestingly, on Bing, it says recent scientific papers on elephants. It's clicked on a nature and then it's thinking and it's going to iterate through this process of clicking links, going to find stuff and then it's going to summarize what it's found at the very end. So let's wait until that and see what happens. You can see it's reading content, thinking, clicking, going back to the last page. It's like sending a little agent out into the Internet to do all of the hard work for you. It's a little bit frustrating because you can see here that it's got like click failed, thinking, reading content failed. Like you can see it failing as it goes along. I'm not sure right now this is mature enough and useful enough to be out in the wild properly. Nonetheless, it's here and let's see what it comes up with at the end. So really, this has found one paper and it says like, here are some recent findings from a paper titled blah, blah, blah. And it's got the actual thing. It's got the reference. I think I can click here and actually go out to the paper, which I can, which is brilliant. So it does do what it says, but it was really frustrating just to sit back and watch it actually try to do it. So I'd much rather just go to a repository where I know there's lots of papers for my field rather than relying on this. Maybe with a little bit more direction, a little bit more prompting, you'll be able to sort of like get it to act a little bit more error in message stream. How annoying. So here are my thoughts on ChatGPT with plugins and Bing at the moment is that it's just so annoying to use. Like we were promised or maybe not promised, but our expectations for these tools were like it was going to change the world. It has changed the world. It's made it a little bit more frustrating to do research. So I'd much rather just go to a place where I know there's papers, search for them. You can actually go to these places and search by date. Tools like Elicit are brilliant for actually finding papers. There are other tools out there. I think trying to use ChatGPT as a research tool at the moment is more frustrating than it is useful because you get these errors. You actually see it failing. It takes a long time. You're waiting around. So at the moment, I think for research, you're better off using the old fashioned ways with specific tools for specific things you want to do. A little bit disappointing is the hype dying of AI in research. Are we past the hype hump? Maybe we are. So there we have it. That was my trial of the new ChatGPT with Bing and ChatGPT with plugins for research. Let me know in the comments if you've tried it. Is it just as frustrating for you as well? I'd love to hear it. But also remember, there are more ways that you can engage with me. The first way is to sign up to my newsletter. Head over to andrewstapleton.com.au forward slash newsletter. The link is in the description. And when you sign up, you'll get five emails over about two weeks. Everything from the tools I've used, the podcast I've been on, how to write the perfect abstract and more is exclusive content available for free. So go sign up now and also head over to academiainsider.com. That's where I've got my ebooks, the ultimate academic writing toolkit, as well as the PhD survival guide. I've also got a resource pack if you're applying for a PhD and grad school. I've got the forum and blogs growing out as well. And remember, it's all there to make sure that your PhD and academia works for you. All right, then I'll see you in the next video. Transcribed by https://otter.ai

ai AI Insights
Summary

Generate a brief summary highlighting the main points of the transcript.

Generate
Title

Generate a concise and relevant title for the transcript based on the main themes and content discussed.

Generate
Keywords

Identify and highlight the key words or phrases most relevant to the content of the transcript.

Generate
Enter your query
Sentiments

Analyze the emotional tone of the transcript to determine whether the sentiment is positive, negative, or neutral.

Generate
Quizzes

Create interactive quizzes based on the content of the transcript to test comprehension or engage users.

Generate
{{ secondsToHumanTime(time) }}
Back
Forward
{{ Math.round(speed * 100) / 100 }}x
{{ secondsToHumanTime(duration) }}
close
New speaker
Add speaker
close
Edit speaker
Save changes
close
Share Transcript