Everyday AI Podcast – An AI and ChatGPT Podcast - EP 347: Google's Gemini Gems - What they are and how they compare to custom GPTs

Episode Date: August 29, 2024

Win a free year of ChatGPT or other prizes! Find out how.Google silently released its answer to custom GPTs with its Gemini Gems. In short, Gems allows users to create a smaller or more customized version of their big model, Google Gemini.↳ But how does it work?↳ Is it any good?↳ And how does it compare to ChatGPT's custom GPTs?We bring you the answers.Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan questions on Google GeminiRelated Episode: Ep 253: Custom GPTs in ChatGPT – A Beginner’s GuideUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. Comparison between GPTs and Gems2.  Features of Google Gems3. Demo of Google GemsTimestamps:02:15 Daily AI news03:47 High-profile organizations opt out of Apple's AI data collection.7:00 Google Gemini Gems explained12:21 Enable Google Gemini for all users and platforms.14:30 GPT-4 chatbot leaderboard shows significant improvements.18:55 Mixed feelings about inconsistent unique feature. Limited functionality.21:50 Improve productivity with free PPP course. Visit podpp.com.24:31 Use Google Workspace gems to search efficiently.28:53 Google Gems failed to follow instructions effectively.30:04 Evaluating text search for Google Docs efficiency.34:11 Google Gems' ability to guide accurately needs improvement.37:11 Google Gems may pull unwanted information, problematic.41:57 Break down process into 5 steps.43:57 Referencing Google Docs and sharing prompt.46:05 Google gems are not as good as GPTs.Keywords:Google Gems, GPTs, OpenAI, Google Workspace, Google Gemini, Gmail, AI newsletter, ChatGPT, Google IO developer conference, custom gems, Google gem manager, Google account, Gemini 1.5 Pro, tech issues, Customer testimonial, Screen sharing, Live documents, Custom versions, Jordan Wilson, Nvidia earnings, Apple Bot Extended, AI training data collection, everyday AI, AI for daily practical use, AI experts, Instruction following, Admin console settings, Document access, Dynamic docs, GPT 4 Model.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 in 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. Google has finally announced Google Gems.

So it's essentially a smaller, more customizable personal version of its large language model, Gemini. And it was honestly made in direct response to chat GPT's custom GPTs. So today on everyday AI, we're going to be telling you. what's new inside of Google Gemini gems, how they work, and ultimately how they compare to chat GPTs custom GPTs. All right, I'm excited for this one, y'all. I hope you are too. If you're new here, thank you for tuning in. My name is Jordan Wilson and this is Everyday AI. We're a daily live stream podcast and free daily newsletter helping everyday people learn and leverage generative AI to grow their companies and to grow their careers. So if that sounds like you, maybe you're

listening on the podcast for the first time. Thank you. you, make sure to check out your show notes. You're going to want to click that link for the free daily newsletter because every single day, not only do we bring you, everything that you need to know to stay ahead and be the lead person of AI at your company, but we also break down each and every day's podcast and live stream with the most important takeaways. So you're going to want to make sure to check that out. All right, let's jump into it. And before actually, I got a shout out. Go check out our thanks a million giveaway to celebrate a million downloads at Everyday AI. We're doing a giveaway. So we're talking about Google Gemini today. Maybe you want a year of Google

Gemini on us. That's fine. Or a year of chat GPT. That's fine. It's super simple to sign up for the contest on the website. Once you do, you'll get a referral link and you can post that on social media, email, text message, whatever. And then the winner of the competition is going to get a year of their favorite generative AI tool on us. All right, let's talk about what's going on in the world of AI news before we talk about Google Gems. So first, Invidia has reported strong earnings, but stock drops in after-hours trading.

So, Nvidia's latest earnings showed that its significant impact on the AI market, yet investors reacted cautiously despite those strong financial results. So, Nvidia reported a net income of $16.6 billion with adjusted net income at about $17 billion, and revenue up a startling, 122% year over year to $30 billion. Despite these impressive numbers announced during their earnings, Nvidia's stock fell nearly 4% after hours, indicating that investor expectations may have been too high. So analysts pointed out that while Nvidia frequently beats,

earnings estimates, this quarter's results exceeded expectations by a smaller margin than in the past, which may have disappointed some investors. So that one's important to talk about because NVIDIA has at least for the last year or two kind of been the bar on investor sentiment, which in theory impacts tech stocks here in the U.S. and the U.S. economy as a whole. So even though NVIDIA beats, beat its earnings forecast, its stock dropped slightly 4%. It's 4% is actually a big, big drop overnight. All right. Next, AI news.

Some prominent publishers have opted out of Apple's AI training data collection, signaling a shift in digital rights management. So notable organizations such as Facebook, Instagram, The New York Times, and Condonass, among others, have opted out of Apple's AI training data collection. So the tool enabling this action is called Applebot Extended, which allows, website owners and content producers to prevent their data from being used by Apple's AI for training while still permitting the original Applebot to index their content. So according to originality AI, approximately 7% of high traffic websites are currently blocking this Applebot extended.

So what this means in short is they're still saying, hey, Apple, we still want to be discoverable, but we do not want you to scrape our content for use of your future AI models. This is a pretty significant shift that Apple kind of splits this in half when most companies do not. So, you know, at least as of now, you know, if you want to block Google's from using your content for AI, you also can, you know, inadvertently block it from indexing your site in general. So it's a risky game to play. All right. Last but not least, Open AI is in talks for a funding round that values the company at more than $100 billion. So Open AI is reportedly negotiating a new funding round that could elevate its valuation to more than $100 billion, reflecting the company's rapid growth in the

artificial intelligence sector. This development is significant as it underscores the increasing interest and investment in AI technologies. So reportedly Thrive Capital is leading the funding round and plans to invest a billion dollars, according to confidential sources. Open AI's valuation has surged from $29 billion last year to a reported $80 billion earlier this year with annualized revenue exceeding $2 billion. Microsoft OpenAI's largest backer, which reportedly invested $13 billion a couple of years ago, is also reportedly participating in this funding round, although they have declined to comment on the matter.

So recently, there's been a lot of hype around Open AI, new models. We've heard about Strawberry, Orion, GPT-5. We've been covering it all here, as well as Open AI has started to roll out early access to tools such as search GPT, a Google competitor, as well as its advanced voice mode. So make sure to tune in. We'll be keeping you up to date on that. All right, that's enough AI news for the day, y'all. Let's get straight into it and talk about Google gems. So if you didn't know Google at their I.O. Developer Conference in May.

announced Google gems. And I was kind of taken aback by this. I wasn't expecting this. But because it was kind of, I think, a direct play against OpenAI. I think the industry saw how powerful and popular that OpenAI's GPTs were when they released those back in November 2023. So essentially what this concept is that Google gems and custom GPs inside of chat GPT do is, it essentially allows you to create a more personalized or customized version of the model based on your data and some custom instructions.

So let's say that there's a task that you repeatedly do inside of a large language model like Google Gemini or ChatGBT. These smaller, more personalized versions of the model allow anyone without coding experience to usually within a couple of seconds or a couple of minutes to create a pretty robust version of the big, model and then you can go back and use that for those purposes whenever you need it. So with that, Google Gems is pretty much the same. All right. So let's do a quick introduction to Google Gems. And if you are joining us on the podcast, I'm going to do my very best to describe. We have some visuals.

We're going to be doing a little bit of screen sharing today. But hopefully you should be able to keep up and keep along. I think this one's going to be pretty tight and succinct. So first, let's talk about an introduction to Google Gems. So to build them, you have to have a paid account. Okay. So you can use the five built in Google gems if you have a free account, okay? But you do need a paid account to Google Gemini in order to build these Google gems.

All right. And essentially, like I said, they are a customizable version of the larger Google Gemini model. and they are built in real time. Conversation, simple. You just say, hey, build me a version of Google Gemini that does this. You can try to connect your data, which we're going to show you here in a little bit. And then you can go use that gem whenever you need it. You can create two gems, 20 gems, 200 gems.

All right. So like I said, you do need, if you want to build them, though, you are going to need to pay $20 a month for access to Gemini Advanced. All right. And here is some snippets here from Google themselves, kind of what they're saying this is. So Google Gemini advanced in Google Gems. Here's what they're saying. So they're saying introducing gems, custom AI experts for any topics.

With gems, you can complete tasks more efficiently. And also, gems are your custom team of AI experts. So this is how Google is marketing gems, right? You can work in Gemini more efficiently. It's essentially a way to build a team of AI experts. Okay, pretty cool, right? So let's dig in a little bit more. But first, we have to talk about accessing Google gems.

This is where it gets a little bit tricky, and I'm going to do my best to explain this to you. So first, if you have to understand, you need a Google account. Okay. So maybe you have a free. email account. That's fine. You still need to sign up for that $20 a month plan in order to build any gems. All right. So it's actually easiest if you have a free at Gmail account. But so many companies, my companies, both of my companies included, right? So we use Google workspace. All right. It used to be called G Suite. It was called something else before then. Google changes the names all the time.

I swear they change the names more than I sometimes change my shorts, which is a lot. But it's actually a little trickier if you have a Google workplace account. So I'm going to go through and walk you through step by step because it's a little easier now, but like two or three months ago, this was a straight up maze. It was hard to find. So like I said, first, you have to have the paid accounts. So you have to do that first. And you do that.

It's actually through a product now called Google One. That's where you sign up through it. So we're skipping that. So let's say you are a Google Work workspace account. Maybe you're a small business. Maybe your company has 500 employees and you're all using Google workspace. All right. So for this, you're going to need to go to the admin council.

So you do that by going to admin. com. And you can already guess, well, you're going to need to be a system administrator. So you might need to ask, right, if you're a larger company, you know, hundreds or thousands of employees, you're probably going to want to talk to someone in IT, someone that's a system administrator because they have to enable this. All right. So even after you pay the $20 a month for Google Gemini Advanced or depending on how your organization is set up, it might actually be called Google Gemini Enterprise. It might be called Google Gemini Business. Yeah, so many names. So, but if you, you're not, but if you're you, you know, you're going to be called Google Gemini Enterprise, it might be called Google Gemini Business. So, so many names. So, but if you you are on a workspace account, you are going to have to have your, uh, admin for your workspace. Go to admin.com. And then there's at least they updated this because it didn't use to be this way. It's a little easier now. There's a new dedicated generative AI tab once you are logged into the admin council. So, uh, under there, there's three, uh, there's three kind of, uh, sub menu items.

Gemini, Gemini for workspace and Gemini reports. So first, again, you have to be the admin, you're going to go to the Gemini tab. And then you need to make sure you turn it on. All right, because even after you pay for Gemini advance, this is not turned on. So you need to either turn it on for everyone or you can turn it on for certain users. Also, if you want Google Gemini's other AI to work around in other places, like we've done some shows, some tutorials on our YouTube channel about, you know, you can use Gemini inside. of Gmail. You can use Gemini inside of Google Docs, inside of Google Sheets, etc. You would also need to go down there to extensions and make sure that is turned on as well. One more quick setting is you need to go into Gemini for workspace. You might as well do this since you're in there. This shouldn't impact your ability to get gems or not, but you need to turn on alpha features as well. So a lot of times, you know, as an example, we did imagine three, you know, Google's AI image generator. right, that was an alpha release.

So a lot of these, you know, products and services come out to a general release after they are an alpha or a beta release, right? So they come out in tiers. So if you do want to try these early on, that's another little tip while you're in there, right? You don't need to do this for Google gems, but while you're in there, you might as well turn on alpha features. All right. We got that out of the way.

So that's a lot. I wish Google made this easier, like Microsoft copilot, like chat GBT, like Claude Anthropic, like just about every other generative AI system out there. I don't know why. Google, you should fix that. Make it easier. It's so complex. Stop changing the names every other day, please.

All right. So now we have to talk about what you're actually getting. All right. And this part is important, y'all. And I'm going to go ahead and tell you and show you why this is important. All right. So you probably heard a lot recently about Google Gemini because in the last month, they've had two pretty sizable releases.

All right. I'm going to go ahead and share my screen here for our live audience just to show you what I'm talking about. So there's something we talk about a lot here on the everyday AI show. This is called the chatbot arena leaderboard. Okay. So essentially, you'll see here, GPT40 from chat. is the best model. And I don't care what anyone says it's the best model by far,

no matter what metric you look at. All right, at least a comprehensive metrics. But you'll see right now, number two and number three, or tied for second, actually, are Gemini, but they are the 0801 and the 0827. All right. So essentially, Google has had two pretty big updates this August, that the model has gotten so much better, right?

The Gemini 1.5 Pro, and they've also done great things with their flash model, making it cheaper, more affordable for developers. But that is not what you get. All right. So I actually have to go down here all the way to number 10. This is what you are actually getting. You are getting the Gemini 1.5 Pro.

It's technically the quote unquote older model. So if you've been hearing great things about all these new updates recently, that's not what you are getting at least right now. That's not what you are getting. getting inside of Gemini, and that is not what you are getting when you are making gems. So keep that in mind. I personally think what's available on the front end of Google Gemini, y'all, we pay for, I have two paid accounts. I don't use it a lot. Right. I pay for these so we can test them and always bring you the most up-to-date information. It's not very good, if I'm being

honest. I hope that Google updates to these newer, more capable models soon. But keep that in mind, you're not technically getting the latest and greatest when you are going into Google Gemini on the front end or creating these gems. All right, we got that out of the way. Now let's talk about the process of how you can actually create a custom gem. And like I said, we got some screenshots here for our live audience. But I'm going to go ahead and try to do our best to walk everyone through even if you are on the podcast. So once you log in to Google Gemini, so again, you got to sign up for the paid account if you want to create this.

All right. And then so you go to gemini.gov.com. All right. And then on the left hand sidebar, there's going to be something called the gem manager. All right. And then you are going to click new gem. So pretty simple. All right.

Google actually has a great user experience and user interface, probably one of the best, but their product is, I think, a little disjointed. It doesn't always work very well. All right. So then here is how you go through the steps. It's pretty simple. I broke it down here. One, two, three.

All right. So first, you give your gem a name. Like I said, this is a customizable, personalized version of Google Gemini, of the big model, right? Because what ultimately happens a lot, you know, you might be using Gemini or other large language models for similar tasks all the time. And generally what that means is you have to go through a series. of prompts, sharing information, trying to make the output a little bit better. So like I said, a gem is a way that you can kind of shortcut that.

And you kind of start at the end with these instructions. So number one is you give it a name. Number two, you put some instructions in there. And then number three, in this kind of builder, you can preview it on the right hand side. All right. So pretty simple. And if you've used chat GPT, their custom GBTs, it's very similar. And I do know you might be thinking, oh, what about Claude and Claude projects?

They're actually set up much differently. That's why I'm just kind of doing a almost like a head-to-head comparison here between GPs and gems. Because I think that GPTs are very popular. I know a lot of people use projects, but they're not, I would say they're not super popular, even though I think they're really good. So that's why we're doing a comparison here because they are set up and they function very similarly. All right. So here's the other thing to know, or another big difference.

If you are used to creating GPTs, you can do so conversationally in a chat interface. Essentially, you can send 20 short messages to the GPT builder inside of chat GPT and build it that way. Not quite the same inside of the gems builder. Okay, you essentially just put in your instructions once, all right? But there is a feature that I like here that I'm going to get to in a second. Here is probably, I'd say, the feature that I'm most excited about, but also currently most disappointed in, because this is a very unique feature, but it does not consistently work. Okay. The big thing of creating these kind of custom versions of large language models, smaller versions is you want to be able to tie a combination of your data and your instructions.

and then you want to be able to use these smaller versions everywhere throughout the product. All right. And unfortunately, Google Gems missed the mark in many of those regards. So, but one thing, again, lots of promise, but I'm not currently excited about this because it doesn't work very well. But you can add some of your files or add access to some of your different Google products and services. So like I said, this is really only going to be helpful.

you are a Google user, if you're a Google workspace user, if you're a Google Docs user. Because when you are building your gem, you can click the at key, all right. I think in most keyboards at shift two, just like on social media, you can at mention. So you click at, and then you get three options, Gmail, Google Docs, and Google Drive. All right. So we're going to do a little test on that here in a second, but that's important to keep in mind. And that's a big actually, if they can get this to work correctly and consistently, this could be a game changer. But I don't think it works very well. And I'm going to let you

decide for yourself because we're going to try some of this live. All right. One other feature that I actually love. So in that step two, when you are writing your instructions, you can just write very simple instructions and then click the little I little magic wand icon. And it says, and I to rewrite instructions. And this actually works very well. 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.adobie.com. Like surprisingly well. And I wish that other large language models that, you know, projects in Claude or GPT's inside of chat, GPT, had a feature like this. Because what it does, and I did a little bit of testing, is it essentially takes your simple instructions in one click and turns it. it into a much more structured set of instructions, right?

Which is great because the key to making these things work is breaking things down, almost like if you've heard of chain of thought, it breaks it down really step by step, right? And it's a good structured set of instructions. So that's one thing I have to give Google gems and the Gemini team credit for. It's a great feature. You know, I think a lot of people might end up just using that and then using the output to create GPTs, if I'm being honest.

I'm going to be using it for that because it does a very good job of taking a simple, natural language, one sentence, a little instruction and turning it into a pretty robust, step by step, well thought out, almost structured output.

All right, so that's a feature I love. So that's it. So once you do that, so like I said, it's one, two, three, title,

instructions, test it out, then it's done. And then on the left-hand sidebar, your new gem, is going to be there. And then to use that gem, all you do is you click on it. And then it essentially starts a new chat. And then you start chatting with your custom version of Google Gemini, with your data and with your custom instructions. Eh, kind of. All right. So let's let's go ahead and

take a look live. All right. Hopefully this works. We've been having some tech issues today. If you're listening on the podcast, we actually had an unsuccessful first round of this. So thanks for everyone on our live stream that tuned in twice. All right. All right. So let's take a look live and I'll show you some of the pros and the cons here. Okay. So let's go ahead and take a look.

So these are now my gems. All right. So I just did, you know, I built a handful of them last night. So let's first do this one, which is called email, responder test. Okay. So I'm going to hopefully everyone can see my screen here for a few live stream audience. Let me see if you can still see in here. All right. So here's essentially what I'm saying. And maybe this will give you an idea of what you could use gems for, specifically if you are a Google workspace user. This is huge. If they can get this figured out.

So I did some custom instructions here. So I essentially said search through my Gmail. And I did the at Gmail. So it's connecting to that extension. That's what those are called inside of Google Gemini. So your Gmail, your Google Docs, and your Google Drive, those are extensions. So in the instructions, you can essentially call to those extensions. So in this case, I'm saying search through my at Gmail and find any newsletters from info at your everyday AI.com. So I'm being very specific here. I'm saying only search through emails

from your everyday AI. That is the sending name and the email info at your everyday AI. Yeah, that's my email. You can email me, right? If you want to get a guest on the show or sponsor or advertise, you can reach out. All right. And then I'm saying specifically, do not search any other emails. These messages should be daily newsletters. The user will ask a question about a certain topic about AI, you will then search my Gmail inbox to find the most relevant answer and present it. All right. So let's see if that actually works. We'll see if this is getting better since the last time I tried it.

So now I just showed you the backend. So I'm going in. I clicked my little gem. I just called it email responder test. And I'm saying this, tell me updates about Open AI from August 2024. Here's why I'm doing this, right? I spend so much time. Our newsletter, yeah, maybe I'm biased.

It's better than everyone else's, right? Everyone else brags. They're like, oh, my AI newsletter is written by AI. We complete the whole thing in 15 minutes because it's all done by AI. I'm a human. We have a team. We have humans build our newsletter, right? Yeah, we give you the normal stuff, the AI news, fresh finds from across the internet, new

AI tools, tutorials, et cetera. But then we also break down our podcast every single day and digestible way for you to better learn. All right. So anyways, I'm saying tell me about Open AI from, tell me updates about Open AI from August 2024 because I always have to read through our old newsletters because y'all, I forget so much because I talk about AI every day. I read about it all day and I forget things. So I actually personally spend a lot of time, whether I'm preparing for a future episode, whether I'm doing consulting for a client that hires us to help them figure out generative AI. whether I'm responding to an email or message, I spend so much time searching through our newsletters, reading old newsletters to try to find things. And even with great, you know, Boolean search inside of my Gmail inbox, it doesn't always work. So let's go ahead and run this prompt. Tell me updates about OpenAI from August 2024. And if it works. So,

so right now you can see, which I like, it's showing me that it's going through and accessing emails. So I kind of see almost like a loading screen here. And here we go. So it says, here are some of the updates about OpenAI from August 2024 found in your Gmail. Fantastic, right? So it's giving me some of the partnerships here. This is amazing. It's already, you know, going in about some of the recent news, products and releases, right, GPT40, very updated recent stuff. Here's, so here's the the good and the bad, right? It shows you sources, and you can toggle these on and then click on them and see exactly what these are.

So in my case, these are different emails. But if you're looking at the live stream, these are not the emails I asked for, right? So it's bringing in, I subscribe for these newsletters years ago or probably like two years ago and I put a filter on them and I don't really read them anymore. but these are all other newsletters, right? And I don't even see a single email newsletter from Every Day, from Your Everyday AI. I don't see anything from info at Your EverydayAI.com. So even though this gem was technically able to do a nice search and to be able to give me a good result,

it did not do a good job of following the instructions. So that's why it's like, hey, what good? is having in, you know, a live connection to your Gmail, your Google Docs, your Google Drive, if it does not follow instructions, right? Because for the most part, if you have live data that you want to connect to, you probably have certain live data you want and certain data you do not want. So in this instance, in this instance, it did not work well. So in my opinion, Google gems, zero for one there, you know, 0.5 for one.

So I did another one, Google Doc test. All right. So I'm going to show you again the inside here. Very simple. This one, simple as possible. I say look in my at Google Docs to find information that satisfies the user's query. Do not use your knowledge base, the internet, or anything else. Only use my Google Docs account to find this information to answer the user's query.

All right. And this more than anything is a test. How well does it work, right? Can you actually steer it in the direction that you need to? to is it going to accurately comb through documents, right? Our Google Docs, we have probably tens of thousands of documents because since I started my company in, I don't know, 2017, I think it was. So we have seven years of documents.

We have tens of thousands of documents. All right. So this is a good way, you know, right now, the way you search for it is essentially by keyword. So this is a great way to conversationally interact with years of your data, if it works right. All right. So now I'm going to go back in.

I'm going to go into my gem. and now I'm having a conversation. So I'm going to ask, please give me my hot takes for 2024, all right? Pretty open-ended, but you'll see here in a minute that it's actually not as open-ended as you would think. Again, it's taking probably about, it's pretty quick. That was probably about eight, about eight seconds to go through here.

So it says, here are your hot takes for 2024 based on your documents, all right? And these are coming from different documents inside of my Google Drive. All right. So these are pretty good. And these are all accurate. And these are all 100% right. So here is the document that it's actually pulling from. Let's see if I can do this in a new window.

Great. So here is, here's the reality. This document is literally called 2024 hot takes. So I made it as easy as possible, right? However, this is where you get. some trouble, right? Because sometimes you may only want to be talking to certain documents, right? Like I said, we have, our teams have hundreds of folders, tens of thousands of documents. You know, the whole point of having access to live documents is being able to grab the right

information from the right files or the right folders for the right client, for the right customer, et cetera. And you don't want to get that mixed. So I'm trying to make this as easiest So I'm going to have a follow-up here. Right. So I'm going to say, please list off all 24 of my hot takes. So again, this gem already found the file. In this file, it's very simple. Again, the name is 20-24 hot takes in all of my hot takes that I did in this episode,

which by the way, I did this episode in 2023, not trying to brag. A lot of people that I was crazy with these hot takes and almost all of them already came true. Anyways. So let's see how it does. So I'm saying please list all 24 of my hot takes. So again, it's going through. It's letting me know. It's finding my documents, accessing documents, et cetera.

Then it says I couldn't find all 24 of your hot takes, but I found 20 of them across the following documents. So now if I scroll down, now it's bringing in more documents. So it's brought in, it looks like five different documents. And it only listed 20, didn't even number them for me, right? even though I know and this gem knows that there is literally a list called 2024 hot takes that I know that it read through that has these listed out very specifically. So the instruction following and the ability to accurately guide a Google gem to specific documents, at least right now, out of the box, is not great. Sure, we could get there with some basic prompt engineering, some A-B testing, all of that, which I'm sure we'll do eventually.

but out of the box, I am not super impressed with this feature. And this feature, I think, is potentially one of the best features of Google Gems if it can work correctly. All right. That's enough for live. Let's finish this show up. So now, let's just do a quick side-by-side factual comparison between Google Gems and chat GPs, custom GPDs. Because like I said, when Google announced these back in May, this is essentially what they were gunning for. Everyone knew that the GPTs exploded in popularity.

I think this is why a lot of teams and enterprise companies ultimately, you know, went or chose open AI aside from it being the best model, right? The ability to create GPs and to be able to share them with your team, right? This is something companies hire us for all the time. Hey, teach our hundreds of employees prompt engineering, teach them chat GBT and create some GPs for us that our team can use. It saves a ridiculous amount of time. It is such a huge advantage. These GPTs have been such a huge advantage for OpenAI, and they had first mover advantage. They've been out since November, so Google finally, you know, release them after about four and a half months.

So here's a side by side. Ready? So they are free to use both companies. However, you can only use five Google gems if you have a free account. If you have a free Open AI account, you can use millions, all right? For both of them, you have to have a paid account to build. But even with a free account, you can only use five if you are using the gems. These are just the five pre-built ones from Google.

Uploading files. You can only do that in GPTs. Yes, I said that correctly. That brings me to my next point is live documents. So that's one of the big differentiators. And that's why I just demonstrated that live because it kind of falls short. This could be such a game changer for Google gems if it's. could work consistently at a high level because yes you can upload static documents inside of

Gpts you cannot upload individual documents inside of gems instead you give it access live access to your Google Drive Google Docs and Gmail so a big differentiator there right it's actually a downside of Gpt's but if it doesn't work consistently inside of gems and if you cannot have really finite control over it how good our Docs, right? Because hey, if it's going to start pulling in things from other folders, other files that you don't want it to, that's problematic. We'll do some more testing and try to see if you can only send it to certain files, but by default, that is not an option. Right. So let's say there's, you know, five very specific PDFs that, you know, your company is constantly using.

That's great to create a GPT on. To do that inside of, of right now, gems, it might not work. because you can't steer Google Gems to just use those files. As you saw, it's seemingly going to pull in what it thinks is relevant information from everywhere. All right. So instruction following. I'd say Google Gems is good, but Open AI GPs are better. I'd say they're great. Model used.

That's important. Advantage GPTs. The GPT4O model is better than the older Gemini 1.5 Pro model that you have available right now. How you build them. So you can build manually in both, but you can build conversationally in GPTs, which I think is a huge advantage, especially for beginners. You can just chat with the GPT builder like you would a normal GBT. You can have a, you know, a 100 message conversation to build a GPT. It's simple.

So advantage GPT. Here's a big one. Third party integrations. Right now, there are none, none inside of Google Gems. you only have their official, you know, three to four integrations. Inside of GPTs, it's pretty unlimited. Because not only are there GPTs, you know, I think the last count was like four million, but you can use an API, right?

So you can tap into almost anything with an API key with a little bit of coding. So you don't need to. But that really opens up a GPT to so many possibilities in working with your company's data and with your company's current systems by how. having that API access inside of a GPT. That's essentially an advanced setting. You don't need to use it. Next, sharing.

That's huge. That's huge. Right now, as of today, there's no sharing capabilities inside of gems. Huge miss inside of chat GPT creating the GPDs. There are. You can make them public. You can put them in the directory, in the store. You know, there's a GPT store.

You can share them individually with one person. you can share them with a team. If you're on an enterprise account, you essentially have your own store with all your employees, GPTs, right? So sharing is huge. Public store, only GPTs. Conversation starters, only GPTs, right?

That's essentially little prompts, starter prompts for you know, for you or the user to use. And then also, can you use in multiple chats? Only GPDs. And that is huge, y'all. And I'm going to go ahead. I'm going to try to keep this tight. I'm going to try to explain what that means.

All right. So when I am in, when I am in Gemini, all right, in a normal chat, I cannot hit the at button and talk to that example gem that I just did, right? In chat, GBT, I can. And that is a huge benefit. Here's why. Let's say you have a repetitive task that you do all day. And it really requires five steps.

You could build five separate GPs that handle each of those five steps very well. So maybe step one is research. Step two is, you know, reading a PDF. Step three is is parsing the PDF and grabbing certain information out of that PDF. You know, maybe step four is is turning that information that you parse from the PDF into a spreadsheet. And let's say step five is, you know, turning that information into a PowerPoint. Let's just say, right? You can do that. You can build five different GPDs that do each of those steps very well.

The difference is you can keep that all in the same context window because you can at mention multiple GPs in the same conversation inside of chat GPT. And then you keep the findings or you keep the context in. So as an example, you talk to GPT1. You get all of this information. It stays in the same window. Then you X out of that one and you click at GPD. GPD 2. And GPT2 already has the information. It has the working memory. It has the context window of what you accomplish with GPT1. So on and so forth all the way through your process. So it's just a very easy way to bring in this kind of large language model AI automation throughout your built-in workflow. You don't get that in gems. Gems only work in their individual gem mode.

So what that means is even if you wanted to break down, you know, that same process into five little gems. Number one, it might be hard because you can't upload a single document. That's beside the point. Right. But then essentially, after gem one, you have to copy and paste the output. Then you have to go into gem two and explain, hey, gem two, I just got all of this information. Now go do your task, right? Then you have to copy and paste that.

You have to go into gem three, do the same thing over and over. It's time consuming. It's not intuitive, right? And that's the whole point of having something like a gem or a GPT is so you can work with them in multiple steps of your workflow journey and save time, which you don't get right now. All right. So let's wrap this up. Here's my final thoughts. I know it's not Tuesday, but I'm coming with a hot take anyways.

Google gems are kind of useless as they are right now, if I'm being honest. Because there is no new functionality here. The only functionality is you can save this to the sidebar, right? But here's the reality. You can't attach certain files, right? And all of these things that you could do inside of a gem, you already had those capabilities, right? Because inside of a chat right now, inside of Google Gemini, I can already, right? I don't want to make this longer, but sure.

Let's do this. Let's do this, y'all. Let me just tell you how, in theory, useless this is. So I'm going into, I'm going to go into the one that I did an example on. All right. So I'm going to go into my gems. I did this Google Doc text test. All right, I'm going to go to edit it. All I'm doing is I'm copying and pasting this. Then I'm going into a new chat. Okay. And so I'm showing you, this is going to do the same thing. I think I'm too. far zoomed in here. So I'm saying, look in my Google Docs. Okay. And now I'm doing this exact same thing. Okay. So all I'm doing is I'm putting the instructions that were in the gem in the chat.

And then I'm putting my, my little prompt. And you're going to see. So it's the, you know, give me my hot takes. Look in my Google Docs. All right. This took 10 seconds. Now this is the exact same thing. This functionality has already been there. There's nothing new. You're seeing it live here. I'm going to get the exact same results. This is not a new feature. It's not new functionality. This was already there. The only thing now is you can quote unquote save what was in this conversation into a gem. So it just adds it to your sidebar. However, I could just save this chat, call it the exact same thing, and it would go back and it would have the exact same functionality. There's nothing new here, y'all. So right now, if I'm being honest, this is pretty useless.

At least this new, oh, gems, gems, no, this is just, this is a Google marketing thing. They slapped a name on it. Zero new features with this gems. Aside from, oh, you can save it, you know, it gives it a pretty, like gives it an eye. and puts it on your sidebar, okay, until you can mention multiple gems in the same conversation, until you have new features inside of gems that you cannot do in a chat, right? Like you can do that inside of GPT, custom GPs, right? You have these conversation starters.

You have, you can tap into third party APIs, right? You can upload documents that stay with that, that, that specifically that stay with that, that GPT. There's nothing new with Google Gems. It's marketing. It's an icon. That's it. For me, right now, pretty useless.

Sorry. I don't think Google's going to be sponsoring the show anytime soon. But we test things out. We give you the truth. And that's the truth. All right. Well, there you have it. That's what Google gems are.

That's what's new, which is actually not a lot. And how they compare to GPs from OpenAI, which Open AI kind of wipes the floor. GPs in almost every single category, aside from being able to access live documents, which we saw, gems don't do that well. Aside from that, one little caveat, GPTs are still far, far better,

and it's not even close. So I hope this was helpful. If so, let me know. You know, it takes sometimes three, four, 10 hours to plan these shows that we bring you every single day. It takes you three, four, ten, seconds to share this, email this to someone, text this to someone. If you're listening on LinkedIn, click that repost button. It means a lot to us. If you're on the podcast, please leave us a rating.

Please tell someone about everyday AI. Also, please go join that. Thanks, a million giveaway. It's going to be ending soon. And please join us back tomorrow. And every day for more everyday AI. Thanks, y'all. Meet Firefly AI assistant. Now live in Adobe Firefly, the Allman One 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. And that's a wrap for today's edition of Everyday AI. Thanks for joining us.

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.