Everyday AI Podcast – An AI and ChatGPT Podcast - EP 242: AI Tools to Supercharge Research

Episode Date: April 3, 2024

An underutilized superpower of generative AI? The way it can completely change how research is done. So how is AI impacting academic and scientific research and what tools are out there to help? Avi Staiman, Founder of sciwriter.ai, joins us to discuss AI tools to supercharge your research. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageLearn more in today's newsletterJoin the discussion: Ask Jordan and Avi questions on AI and researchRelated Episodes:Ep 89: AI’s Role in Responsible ResearchEp 121: Faster and More Accurate Results From ChatGPT with ScholarAIUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTimestamps:01:20 Daily AI news05:16 Simplify writing process for academic researchers with AI.07:36 Democratization of science and skepticism in society.10:33 ChatGPT has diverse data sources, impacts reliability.18:36 AI researchers responsible for high standard, education.19:32 Using AI in research22:37 AI tools improving accessibility and accuracy.27:30 Explore pain points, offer AI research tools.31:17 Questioning scientific studies, AI's importance emphasized.35:05 Encourage critical thinking when using AI tools.37:18 AI can enhance research and understanding information.Topics Covered in This Episode:1. Importance of Research Accessibility and Transparency2. Utilizing AI Tools in Research and Writing3.  Integrity and Reliability of AI-Generated Content4. Responsible Use of AI and Source Attribution5. Problems of AI-generated Content in ResearchKeywords:skepticism around science, accessibility in research, transparency in research, AI tools, democratization of research, research integrity, research reliability, research fraud, misuse of AI tools, scientific literature, attribution in research, Jordan Wilson, ChatGPT, Sykespace, Illicit, Sciwriter, academic writing, AI Tool Up Tuesdays,Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live and Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. How is AI being used in academic and scientific research?

And with AI, can we all be researchers? Should we be? And what are the AI tools that we should all be using that can really help supercharge that journey? All right, we're going to be talking about that today and more on everyday AI. Welcome. Thanks for joining.

My name's Jordan Wilson. host of Everyday AI. We're a daily live stream podcast and free daily newsletters that are helping everyday people like you and me, not just learn generative AI, but how we can all actually leverage it to grow our companies and to grow our careers. So if you're joining us on the podcast, thank you. As always, make sure to check your show notes and go to Your EverydayAI.com. We'll have a recap of today's show in our newsletter that goes out a couple hours after our live stream here. And if you are joining us on the live stream like Tara joining us from Nashville or Brian joining us from Minnesota, let me know what questions do you have about

academic research. All right. But before we get into that, let's first do as we do every single day, go over what's going on in the world of AI news. All right. So Replit has introduced code repair in AI coding assistant powered by real-time coding data. So Replit has just unveiled code repair, the world's first low-latency program repair AI agent. So this AI agent is informed by Replit's unique data on developer intuition and is designed to automatically fix code in the background. So the program uses real-world use cases to enhance its ability to repair code efficiently. And with code repair, developers can save time and improve productivity by having their code

automatically fixed without manual intervention. Pretty big news from Replit. If they haven't heard of Replit, even for non-developers like my, I always use Replit to just go try things and to deploy them. So it should be essentially, I put a lot of bad code into Replit. So I'm excited to see how this new coding assistant can fix it. All right, our next piece of AI news. Stability AI has released a new audio tool called Stable Audio 2.0.

So with Stable Audio 2.0, it's a new update from Stability AI. It allows users to generate music tracks up to three minutes long at a higher quality. as well, just based on AI prompts. So put in text, get a up to three-minute song. Pretty cool, right? So it also has a feature that allows users to manipulate any audio sample using text-based prompts. So the tool has a content recognition filter to ensure compliance with copyright laws as

well. The company, if you listen to the show, it's faced some controversy recently with its previous AI models, training on copyrighted material, leading to the resignation of the company's BP of audio, and I believe their CEO just left as well last week. Speaking of that, our last piece of AI news for the day, musicians are banning together to fight that very thing, to fight the potential impact of AI on the music industry. So a group of over 200 musicians have signed an open letter calling for protections against the use of artificial intelligence to mimic human artists voices and their likenesses

in the music industry. So notable names to sign this kind of open letter include Nikki Minneman. Steve Wonder, Billy Eilish, and a lot others. So the letter demands that the technology companies pledge to not develop AI tools that undermine or replace songwriters and artists. Seems like that's already well ahead, right? It seems like that's already happening. But they're wanting more responsible use of AI technology that could benefit the industry, but concerns over copyright infringement and labor rights remain.

And obviously, with all these new tools, there's been an increased debate over the use of artist likeness after their death as well with AI tools, raising ethical debates. Well, those debates aren't going to go away, but we'll always be talking about them here on everyday AI. But you didn't tune in today to talk about music. You tuned in today to learn about how AI is really changing just the game of research, right? Because it's ever evolving. And you know, you don't worry. You don't have to listen to me blab on about it. We actually have an expert today to come on and help us understand this a little better. So please help me welcome to the show, Avi Stamen, the founder of SciWider AI.

Avi, thank you for joining the Everyday AI show. Thanks so much, Jordan. Great to be back second time. So hopefully that means the first one wasn't too bad. And looking at those pictures at the introduction, I'm looking for an AI tool, which can make us look like we did 10 years ago when we took those photos. If you come across anything, let me know. Yeah, gosh.

That's a good point, right? Like, should I just be, should I have like an AI filter that just makes me look 10 years younger on live video at all times? Maybe. But, Avi, real quick, give us, give us an overview of kind of what you do at SciWiter AI. Yeah, that's a great question. Thanks so much. So, SciRider AI is an attempt to make the writing process for academic researchers much more simple, straightforward, and streamlined. So imagine your typical researcher, you may know some that likes doing science, like spending time in the lab,

but may not love actually writing up their results and having to communicate and document them. That's super critical for the scientific literature and for the academic record that there's actually a recording and documentation of the results of their study. But no one is, you know, especially when it comes to the sciences, no one's looking at it, as a literature prize for writing. So when I'm asking the question of it, is there a way to semi-automate through AI that process from the lab to the paper without stealing really valuable time and resources from the researcher? And if we keep in mind that the average researcher, the average paper, okay, Jordan, let me ask you

this. How much do you think the average paper costs taxpayers? Just one paper, one study. Well, the fact that you're asking me means it's either very cheap or very expensive. So I'm going to go with the latter, Avi, and I'm going to say, I don't know, $5,000, $5,000. Who knows? All right. Add on a number of zeros.

It's a half million dollars for everything. Okay, yeah, it's a pretty wild number. Now, when I say paper, that includes obviously the infrastructure and the lab and the staff that you need. But science is expensive, right? It requires very specific tooling. So we want to maximize, I mean, every single one of us, our taxpayer dollars is going to fund science, and I think that's a good thing. But we want to make sure that we're, you know, maximizing that investment.

and that we're actually letting the scientists stay in the lab. And they spend a lot of time now through a very inefficient process, writing and revising articles. So that's great. And this is really what we're going to be diving into today. And as a reminder to our live stream audience here, what do you want to know about this intersection of AI in research? Now is your time to get your questions in. But I want to start high level here, Avi, before we get into these tools that can, you know, these AI tools that can help supercharge the research process.

but why does research or why should research still matter to everyone and in day and age when it seems like information is at our fingertips? Right. I mean, you know, it's, some would say that science is under attack, right? In terms of, you know, do we believe in science? But for those who do, and I count myself as among them, I think it's really critical and important to realize that part of the skepticism around science was that for, traditionally, it was a very closed box. Okay. So everyone's heard of the ivory tower where the, you know, the scientists go and do

their magic and then we're just supposed to trust them. And I think we live in a different world the same way we don't just trust our doctors. We actually double check their work. And, you know, so long as we're doing that responsibly, that makes a lot of sense. And I think what is really fascinating about what AI has done in a very short period of time to research is it's democratized it. So let's take an example of, you know, God forbid you have a family member who has, you know, an illness who's been diagnosed with an illness right now. So back in the day, you would go to one doctor, maybe you would go for a second opinion if you could afford it. And then you kind of just have to take their word for it and do that. Nowadays, I think the first people think, the first thing people are doing is they're

starting to Google. Now, the problem when you Google is most likely you're going to come across a long slew of really dense, heavy, boring academic articles that for someone who's been in the industry for 15 years, I don't know what to tell you what it means. What AI has been a game changer for, And then how do I know, not only that, how do I know which ones are important, which ones are really peer-reviewed, which ones are leading in the field? That's kind of a, again, a mystery. Maybe it's not a mystery for the researchers in that specific area. But for me, as a patient or as a family member, I really want to know that. So there are some really great tools now that do a number of things.

First of all, they distill a lot of different papers and a lot of different scientific information into lay summaries, into ways that we can easily digest and absorb that content. and me and you can say, oh, I get what they have. I understand what the potential treatments plans are. I understand what the potential drawbacks of those plans are. So that's just one example of how research and making an AI accessibility to research is really critical and changes the way that we perceive our interactions and are the way that we engage with the world around us, really.

Yeah, and you bring up so many great points there, Avi. And I love the concept of just democratizing research, right? Because I even thought to myself, you know, a week or two ago, I'm reading, quote unquote, reading research papers now, right? Because I can talk with these papers, you know, using, you know, a chat GPT or a perplexity or a consensus or something like that. But can you talk a little bit about there and, you know, about how now all of these AI tools can really turn anyone, maybe not into a researcher, but someone that's,

can benefit from being able to actually have a conversation with the paper. How does that work and how does that change how we all consume information? Yeah, I mean, I think we're all familiar with chat GPT, it's pluses and it's minuses. And I think that, you know, probably your audience knows that the training data that it ingested for GPT was a big popery of things, right? Everything from the New York Times and some peer-review journals all the way to, you know, Reddit and other. you know, maybe less authoritative information. And I think that the question we need to ask ourselves is when we're building, when we want reputable, reliable information, or when we're building business cases that really need to be based on fact. So I think we have a really

wonderful resource in the scholarly literature to build out those models. Even if they're in finance or they're in business or they're in law, we want to go based on what is the the content that's most verified or that's highest value or highest quality. And I think that the peer review process that research goes through makes it as such, similar to like the way the New York Times would be for journalism. It goes through a lot of reviews and edits and facts checks. So we want to be basing, I think, our ideas, assumptions, and use cases and businesses based on reviewed content, not based on polluted or diluted content.

So I am a big fan of chat GPT. I use it all the time. But I actually think if we're thinking about actual business use cases in, you know, areas where reliability and trust in the sources is critical, then we have to sort of turn to the consensus or to perplexity. And like you said, it allows, you know, me and you to really get answers that are based on, you know, materials that were inaccessible to us prior. So, you know, I think that there's multiple ways that we can.

can look at how AI is impacting just the academic scientific research industry. But we've kind of been talking about, you know, on the back end, right, after the research has already been published and we're trying to make sense of it all. How is it being used on the front end, right? Because presumably, you know, someone like me would hope that, you know, all research papers, yeah, these research papers that cost $500,000 of taxpayer money, you know, you would hope that there's some original content in there, right? But at what point should we be worried about, hey, is AI being used too much for original research papers?

Is there a downside of that? Yeah, it's a big mess in our industry right now, right, around like research, because we've got this weird love-hate relationship in the scholarly world with AI. On the one end, I mean, AI companies have already come up with incredibly complex chemical compounds, for example, that would never have been able to be discovered by humans because it requires the synthesization of so much data or even running data analyses with synthetic data data that just would not have been possible doing it in the traditional way. And all of a sudden, that's been a game changer.

On the other hand, it makes issues of research fraud all that more prevalent and all that more problematic. So let me explain. You know, some of you may have seen recently a few cases of very senior researchers, even presidents, the president of Stanford University, the president of Harvard University, actually had to resign because of issues with research integrity, right, with the reliability and trustworthiness of their research. And I won't get into the specific details now about each one is its own case, but bad actors or researchers who are trying to cut corners and even good researchers who have,

maybe are not naive to how these AI tools work may just take outputs from LLMs and say, okay, you know, I trust this. It did a data analysis for me. Well, here's the results. Well, can we trust that data analysis? How do we, is there a way to reproduce it, right? Today it might give me one answer and tomorrow might give me a totally different answer. So there's this really, I think, you know, there's excitement because like we said before, it can get me back to the lab in less time and make my work more efficient, take off some of those frustrating tasks that I don't want to be doing. On the other hand, if we use it without looking, without being careful, then I think we run into a problem. So there was a, there was a funny yet sad case that a few weeks

ago of a rat. I don't know if you saw this Jordan, but there was a, there was a peer-reviewed article, a journal, research article that had a rat with a giant testicles in it. And it was supposed to be an atomic display and description of the rat. And it was just an AI, you know, baloney image that made it into a journal. And that made it to the world news. I just saw it was on Stephen Colbert was making some funny jokes about it the other night. Like, it's funny, but it's also quite sad and scary

because, okay, that's a very obvious case where science failed or the scientific process didn't work properly. But how many cases are there that are more subtle that we don't necessarily notice? And can we trust when we're reading these summaries, can we say with a full heart that, yeah, I can rely on what's being?

output here? Or are we worried that, you know, this push towards what's known as publish or perish, right? Do we always need to be publishing articles? It's going to push researchers to cut corners or to falsify data or to generate text, which isn't really based on legitimate studies and research. So I think it's, you know, we have to, on the one hand, push forward with our experimentation, but on the other hand, to always be asking ourselves, does this pass the barometer for research integrity or does it need to be checked at the door and maybe we need to wait until it's been tried in other use cases first. Adobe just introduced an entirely new way to create, bringing the power and precision of its

creative suite into one conversational experience. Meet Firefly AI Assistant, now live in the Adobe Firefly app, the all-in-one creative AI studio. Powered by Adobe's Creative Agent, Firefly AI Assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the assistant. The assistant orchestrates multi-step workflows, drawing on 60 plus pro-grade tools across Adobe Creative Cloud apps, including Photoshop, Illustrator, Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for

common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director. Adobe Firefly AI assistant now in public beta. See it today at firefly.adopi.com. Yeah. And, you know, it's a good point.

I'm going to bring it up here. I don't know. Like, do we have to put like a rated R thing on today's live stream? You know, so here's what I was talking about the rat with the giant testicle that actually somehow made it into actual academic research. So that's interesting. But, you know, Abbe, you bring up a good point here. How does the role of humans change with AI being, you know, relied heavily on every step of the process? How should, you know, humans, specifically those working in and around academic research, how should they be changing their outlook, changing their

role to make sure whatever does get published isn't an obnoxiously large rat, testicle. Yeah, I mean, there's a lot of actors kind of along the way, right? So there's the researchers that are doing the research. There are the universities that they're kind of like, you know, the boss in a certain aspect. The researchers have a lot of independence. So the research universities need to kind of take responsibility and say, we need to make sure that, you know, at Purdue, at Illinois, at Harvard, we are stamping, giving our stamp of approval that the researchers are doing legitimate research. And then there's the research publishers, right? They're the people who are actually putting out the content. They also have a responsibility.

But as often is the case in these situations, because there are multiple bodies that are responsible, sometimes things fall through the cracks or there's expectations that the other one is going to kind of take care of this problem. So it starts with researchers educating themselves about the possibilities of AI, but also the potential downsides of AI. And then it's up for us as a society and as research funders, research universities to come along and say, no, we're going to make sure that the research that's being published or the research that's coming out of our institutions is of the highest standard. And in different fields, that means different things, right? So if you're a historian, making sure that you're going to primary sources is going to be key and not just relying on, let's say, the LN. On the other hand, if you're a scientist, making sure that you're, you know, that you're double and triple checking your protocols, that's great. That's going to be really important.

But for example, a protocol. So that's a, let's say, a formula that I go through. in order to design a research study. Well, who's to say that us as humans, we have a better ability to design a research study or to say which chemicals I need to put in, or which testing environments than an LLN? At the very least, I think that it's every researcher's responsibility

to use it as a springboard, to use it as an idea factory to bounce off of. We may decide that, you know what, we're actually gonna reject the LLM's suggestions or we're gonna continue prompting it. in different directions because of things that we know. But I think that whenever it comes to large data sets, it's really going to be important for us to have the humility to say,

you know, we're going to include AI here. The bigger question, I think, becomes what is our error rate that we're okay with? And this is the same question that we ask when it comes to self-driving cars, right? Are we okay with one out of every 100,000 people, you know, getting into an accident as a result of a self-driving car? Well, how does that relate to one at every 10,000? when we're not using that technology. And that's more bigger philosophical, societal questions that I think, you know,

we need to become more comfortable with the fact that it's not going to be perfect, but then again, neither are we humans. So let's just be a little bit, I have the humility to recognize that as well. Yeah. And speaking of, you know, accuracy when it comes to using large language models, a great question here from Tara. So thanks for this one, Tara. So she's asking, how can we assure accurate attribution to original sources?

in the AI-driven realm of writing and research, thereby giving proper credit where it's due. How does this work? How should this work, Avi? Yeah, that's a really good question. And that's why I think that scientists have reacted quite strongly and sometimes negatively to GPT because of the lack of sources or because of falsified sources.

And that's why I'm a big proponent of some of the tools you mentioned before, Consensus. Scholar AI is a great add-on to GPT. So you can actually use it within JetGPT. And all you need to do is then when you're querying or when you're prompting and you're doing your prompt engineering, you actually get results that are sourced, which is really great. And perplexity does the same thing. And I think that's really important because, you know, we need to make sure that we can always trace it back to the source.

And most of the time we need to actually go back and read that source. It's not really any short cuts to, you know, in the meantime, at least that I know of, to fact-checking, you know, LM. So I think that, you know, the whole scientific record is based on attribution in a proper way, right? I'm here is what, basically a scientific paper is saying, here are the studies that have been done previously. Here's a gap in the literature. Here's a question that we haven't been able to answer yet. I'm going to try and come and fill up that gap, but I need to be able to connect it to what came before me in order to move forward. I can't just do research kind of at a thin air. So that's a really, yeah, that's an important sticking point. So, you know, we, you kind of just mentioned there,

I think a lot of, you know, great AI tools that the everyday person can use. So, you know, chat TPT luckily did, you know, announce earlier this week about improving its citations. I think it's, you know, obviously a work in process. But, you know, we talked about even if you are using, you know, a large language model like chatbt, its ability to, to work with consensus, scholar AI. You know, we mentioned perplexity. Is there anything else, Avi, whether it's the everyday person or even people who are more in academic and scientific research, is there any other tool that is going to kind of better ensure, because you can't get it 100% right? Is there any other, you know, tool that can better ensure accuracy, whether it's

for reading, writing, summarizing, et cetera? Yeah, there's two tools that I personally like using a lot when it comes to, if I have a question and I just want to, you know, get a reliable answer and a sourced answer. And one of those is a size space. And size space, it really does a great job because what I can do is I can get, first of all, I can get what I call a meta answer. So an answer that summarizes all the literature and tells me kind of what the bottom line is, but I can also look at individual studies and it'll summarize those as well. And within those studies, I can summarize different parts of the article. So let's say I want to understand what methods they used in this article. Well, I can summarize that. Or if in another article, I want to understand what their

conclusions and results were. Well, I can analyze that. And, and And that's a really, so that kind of becomes more granular and detailed, and it's a dynamic way for me to summarize. So I'm a big fan of a size base. The other one that I like to mention is illicit. Similar functionality, elicit.org. And elicit is, again, a tool where I can ask a question. So maybe Jordan, you just want to like click on one of those questions that's coming up there so people can get a feel for it. It might be cool.

And actually get like science-based answers to the question that I'm asking. So all of a sudden I've got, so you see here there's like, like all these papers, right, and the insights, there's a TLDR in case I don't want to read it. But here I have a one paragraph summary that's sourced, right, which is a reliable answer, okay, or at least somewhat reliable answer, right, more reliable than you get in GPT. So already there, I think it addresses some of the major problems that we see in chat GPT already addressed here. Now, what will be interesting is if what you're saying is right, which I assume it is,

that chat GPT is going to start doing this, does that, you know, make these kind of tools already kind of outdated? But I think there's a lot that scientists can do with some of these tools that really make them quite powerful. So that's on the, that's on the kind of scientific search end. And then in terms of what I'm building on, you know, SciWiter, is really has a lot more to do with scientists who are writing their research. But it's a, no one is born knowing how to write a scientific paper. Put it that way. Oh, that's not good.

That it's coming up, coming up with a with an issue. Okay. I'll have to take a look at that. But, you know. But it typed it in wrong there, obviously. Oh, okay. I got you giving a bit of a heart attack there. Maybe I, you know, maybe our website's dad.

Anyway, so, but I think what happens, there you go. All right. That's, that's, that's it. Yeah. So it's really, it's really trying to overcome that blank page and the struggle that comes along with writing that academic paper. And by the way, that starts early in your thesis, right? It can continue on to if you're doctoral student, postdoctoral student.

No one's born knowing how to write in that specific genre. It's a new genre that you need to kind of learn on the fly. universities don't do a great job of teaching academic. So this is almost like a 24-7, you know, assistant pilot, almost like imagine having your own private writing coach available to you whenever, however. And that's kind of what we're trying to build over at Sightwriter. So, so maybe, you know, I'm curious, obviously, you know, you're very well versed in this kind of intersection of AI and research. So let's say someone for the first time is hearing about this and they're excited. Maybe, I don't know, they're in college or they're working to, you know, transition in their career and learn something.

What's some good first steps for people that they can use AI in a responsible way and make sure that they can learn things, maybe a little better, a little faster, and in making sure that, oh, I'm not relying on a bunch of hallucinations. Yeah, that's a great question. And I'm sorry, this is shameless self-promotion, but this is just, I've taken the horns in this field and kind of tried to address these questions. So I ran a course called AI Tool Up Tuesdays. It's an entirely free course.

It's 24 different tools that each one of them is really, you know, very relevant for research and for researchers. And I split them up thematically. So you've got research writing, how to create illustrations for research, you know, research discovery. And what I recommend is I don't recommend to anyone to try all these 24 tools. I think you're going to get overwhelmed. You're going to say, forget it. I can't handle AI. It's not going to work.

But what I do think. can happen is you can pick out two or three tools that you're like, all right, this is interesting to me. I'm going to try this out. I'm going to play around with it. I'm going to see if it works. Think about what your current pain points are. What are your problems? What makes it hard for you to actually do that research? And that's the way that, you know, I think it's just like in bite-sized pieces. You can, you can kind of, you know, handle that. I'm also doing, you know, a boot camp for research institutions and for universities in a more formal way that's, you know, a little bit more structured and actually coming in. But anyone can access the AI tool up Tuesdays and just sign up

and just kind of get a taste of these tools. Because what I think is really exciting about our little kind of corner of this big AI market is that I think people are going to start finding us more and more because I think that in the end of the day, a real business that respects themselves or a real, you know, entrepreneurial venture is going to say, I don't want a polluted, you know, LM that I'm drawing from. I want an LM which is, reliable, verifiable information. So that's why I actually think that it's interesting, because on the one hand, it'll be easier to publish and maybe publishing will be more democratized. On the other hand, the value of this curated content, I think, is only going to increase over time.

And, you know, getting back to what we, you know, talked about a little bit earlier, Avi is, you know, this kind of, you know, funny example of somehow this, this giant picture, AI generated photo of a rat with a giant testicle made its way into the academic research world. I think for every one of those that gets caught, there's probably, I don't know, dozens or hundreds or thousands of, you know, maybe academic research papers that aren't maybe 100% true because of AI. Is there a problem when we look into the future now that these things are online and, you know, the next version of all these large language models are going to gobble this up?

is there a danger that maybe AI early on was being used in irresponsible ways before a lot of these tools existed that now is going to be gobbled up and presented as fact in the future? Is that a problem? And if so, how do we address that? In one answer, yes, it is definitely a problem. I don't want to put it to this. I don't want to overreact. So what I mean by that is that there actually were some interesting articles.

This is a very hot topic in our industry right now. And there were some articles doing some early studies, and it seems like it's a problem, but it's also not a, it's not, you know, it's not taking over. I want to be really careful to not, so that people don't have, you know, the feeling all of a sudden, well, it's like, well, we can't trust science anymore, so forget about that. No, that's really not it. And I think that we need to be aware that it's an issue, come up with tools to identify these issues, and then, you know, and address it.

But I don't think that we're going to be able to root it out entirely. you know, one good example, I'll give you one example where it's being done pretty well, is in the case of images. So there are a few companies, one's called Prufig, another one called Image Twin. They're specific to the scientific research, probably not super interesting to your audience, but they will be able to go through an image and see if the image has been doctored, or if it's been duplicated, or if it's been copied from somewhere else. And that's the kind of tool where a human being would need to know all the images in the world,

they would never be able to do it. And I think it's somewhere where actually the problem pre-existed AI, and AI is part of the solution, not part of the problem. So, you know, I want to warn of, yes, we need to always think critically. I guess that's the bottom line, right? Anytime you read it, even if you see a research article, don't take it as, you know, God's word to Moses on that, you know, on the Hill. Like, you always need to be thinking about, does this make sense? Are there conferring studies, meaning are there other researchers that agree with this? And also do a search online.

There's an amazing website called PubPier where the entire goal of PubPier is to critique published articles. So what that means is someone will publish an article and say, here's the study I found. And then there's other science sleuths that come along and they're like, eh, this doesn't look right, right? There's something here that's off. The data doesn't make sense. And that's where some of these stories come to light from some of these more higher profile,

you know, heads of universities that were doing things that maybe weren't exactly, you know, as they should be. I think, you know, we should consume, we should respect the academic literature, we should consume it, but we should always have that critical, we should never, like, relieve ourselves of the critical thinking and, you know, the confirmation studies that we're going to want to see in order to, you know, actually make sure that it gets into play. One last story that I wanted to kind of share with with you that I think really drives home how valuable and important AI is in the scientific realm is a story back from the 1970s, a professor to Yu-U. She was a researcher in China in a lab for, you know, for they studied plants and, and, you know, and, you know, and cures for, and disease research. And she came across this, this component of a plant called artemin, which essentially, she realized over time was, could cure malaria, right? And malaria is one of the most deadly diseases killing millions people a year in Africa. And she wrote up her results.

She wrote up her study and she published it in Chinese. Now, I don't know why she published it in Chinese, but no one noticed it. Maybe a few researchers read it. So essentially, we had a cure for malaria for many years. And there you go. And we didn't do anything about it. Like millions of people died unnecessarily. And I can only think that if she had found this in the age of AI,

where we can easily search for research articles and understand them regardless of what language there. And she could have published it in English easily or in 20 different languages simultaneously. We could have saved millions of people. So while I think that we need to put up barriers to AI or safeguards to make sure that it's not overrunning the research literature, I also am concerned of what happens when we don't make those tools available

or we ban them. There have been journals, scientific journals, that said, nope, no, no AI use. that's problematic in its own way. So I just, I think it's, you know, it's kind of running the, the rational middle path where we're careful, but we're also encouraging of people to do experimentation. And as you see here right up on the screen, she got a Nobel Prize, right? As well, she well deserved.

Unfortunately, it took us years to realize that she was deserving of such. So, Harvey, we've covered so much in today's show. You know, we've talked about how AI can democratize research. We went over a lot of those. tools many I haven't even heard of that can, you know, really supercharge your research process. But maybe what's the one takeaway that you want people to leave with today, whether they are working in the researching world or they're trying to take advantage of AI to better learn and understand research papers? What is your one best takeaway for all these people to kind of live

at that intersection of, you know, learning new things with AI in a responsible way? Yeah, I think that the main takeaway is that we all do research, right? Anyone who has a business, as at some point dump some sort of market research, or if you're looking to buy a car, right, you do research. And if you're, you know, as we discussed, if you're looking to, you know, understand a family member's illness, we all do research. Now, when I say do, it doesn't necessarily mean that we're sitting in the lab each and every one of us, but it does mean that we need to understand it, right?

Our whole lives are built on the premise of, I get into the car, I'm trusting the researchers, that have built this, right, and the technicians that it's going to work and it's not going to explode on me. But I think that we realize that we need to all be, educate ourselves more and more as time goes along because there are so many competing narratives, and it's the only way to do it is really for us to understand and to realize. So what I would encourage people to do is next time you have a question or next time you have a decision and you're like, all right, let's try a GPT or let's try an LM. Stop for a minute and ask yourself, is this question important, right?

is the answer going to impact my business decisions, my life decisions, right? And if it has, if it's not just, well, write me a fun poem for, you know, April Fool's Day, but if it's something more meaningful than that or more important or the stakes are higher, take the time not just to throw it into GPT, but also to try some of these other tools. Because I think that many of them are built by scientists who said, GPT is great as a platform, but it's not built for, you know, really reliable, verified, you know, information and content, which should be the basis, at least in my opinion, of all the decision-making or anything that's meaningful in life. And therefore, be aware of where the content is digested,

you know, what access we have to the literature, and educate yourself, right? Before you go to that, doctor, find out what the most recent treatment, you know, plans are. It'll take you 10 minutes. Maybe it used to take you five hours and it's like, no, forget it. I'll just trust my doctor says, now it takes you five time minutes to understand, okay, here's A, B, C, D. Here's how they treat, here's how science treats this illness. And I just think that it will, you know, we should have the humility to know that we're not the, just because we get these, you know, answers, doesn't make us the, you know, the arbiters of truth. On the other hand, we'll have educated ourselves in a way that's meaningful and that impacts our lives in a real way.

Wow. I think that's such great advice. You know, you broke it down there for us, not just, you know, to grow your business, but how we can just, help us in life. I hope everyone that tuned in today got a lot out of today's conversation. Avi, thank you so much for joining the Everyday AI show. We really appreciate it. Thanks so much, Jordan's pleasure. And next time, that'll make a hat trick. So you know, you'll let me know. There we go. And hey, there was a lot. Yeah, we name dropped a lot of different tools, studies, et cetera. Don't worry. Well, first, if this was helpful, please consider giving us a review on Spotify, Apple Podcasts, etc.

Or sharing this with your friends. You know, I think there's a lot of people that need to hear this message on responsible ways that, yeah, AI can kind of supercharge your research and how you even read and understand information. So make sure to go to your EverydayAI.com. Sign it for that free daily newsletter. Will we be recapping this fact-filled show? Thank you for joining us.

Hope to see you back tomorrow and Every Day for more Everyday AI. Thanks, y'all. Meet Firefly. AI Assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com.

day's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayaI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.