Ep 442: The Internet Is Broken. Can Google’s Deep Research Fix It?
57:29 Share

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. There's a brand new AI tool that you're probably not using yet. It hasn't grabbed many headlines. It's not one of those sexy splashes like Sora or something like that.

but it's from one of the biggest tech companies in the world. One of the biggest companies in the world. It's from Google. So today we're going to be going over Google Gemini Deep Research, which I think is one of the best new AI tools ever.

that you're probably not using. And we're going to show you it live. We're going to take suggestions from our live audience. And then I'm going to end the show by talking about the bigger picture of what does this mean for the internet? What does this mean for writing? What does this mean for content publishing?

All right. I'm excited for today's episode. I hope you are too. Welcome to Everyday AI. My name is Jordan Wilson. I'm the host of Everyday AI. And this thing, it's for you. This is your daily live stream podcast and free daily news that are helping everyday people like you and me, not just learn about AI, but how we can actually leverage it to grow our companies and our careers. That sounds like you, you are a hundred percent in the right place. Uh,

So yeah, we do this every day as a daily live stream podcast, unedited, unscripted, the realest thing in artificial intelligence. But if you want more realness to help you actually grow, it's on our website, youreverydayai.com. There you can sign up for our free daily newsletter where we recap information

each day show providing you even more insights in the written word, but also you can go listen to watch and read more than 400 shows on our website, sorted by category. So go find what you need. We've already gotten the experts who have answered your questions. All right, before we get into.

Today's show, we're going to start as we do every day by going over the AI news. And hey, live stream audience, start thinking now. Go ahead and get it in. I need an example of a topic that might take you a long time to research. Not something that you're trying to learn, but a topic.

Right. So go ahead. And well, I guess a topic that you're trying to learn that would require a lot of research. So please get those in now for our live stream audience so we can go over them at the end. All right. Let's start with the news. NVIDIA has unveiled Jetson Orin Nano Super Developer Kit. All right. So this thing is a powerful AI computer priced at two hundred and forty nine dollars.

And it's aimed at enhancing AI in robotics projects for enthusiasts and developers.

So the Jetson Orin Nano Super, that's a mouthful, boasts a neural processing capacity of 67 tops. So that's 67 trillion operations per second for a little machine. That's wild. And that's a 70% increase compared to the previous Jetson Nanos 40 tops, providing a substantial boost for AI applications.

So, NVIDIA promotes the new Nano Super as an ideal tool for creating chatbots, visual AI agents, and AI-based robots, opening the possibilities for developers and hobbyists in the AI space. Next, Google. Yeah, we're talking Google deep research today, but Google also kind of released Gemini 2.0 in full.

So Google CEO Sundar Pichai on Twitter announced an experimental version of Gemini 2.0 called Gemini Experiential 1206, right? We love the model names here. So right now this is available to Gemini advanced subscribers. And according to Google, this model promises improved performance for complex tasks, making it a noteworthy advancement for users and fields.

So last week, as you all know, Gemini announced the availability of an experimental version of Gemini 2.0 Flash. That's like the mini model, right? So sometimes companies first release the big model, then they release like the bite-sized version. So Google did it opposite. So they released Gemini 2.0 Flash last week. Very, very capable. It benchmarked bigger than its, you know, quote unquote pro model. And so everyone's been wondering like, okay, are we actually going to get the full version of Gemini 2?

This is it, but it's also not technically it. So this is an experimental version. They're not calling it Gemini 2.0 Pro, but it is in the family of Gemini 2.0 models. So the newly launched Gemini Xp.

1206, if you're looking for it, that's EXP, offers significantly better performance in tasks like coding, math, reasoning, and following detailed instructions. So users can access this experimental model through the Gemini dropdown model on both desktop and mobile devices.

platforms and you do have to be on a paid plan to use it. And it is an early preview and may not function as expected. And it lacks real-time information and compatibility with some of Google Gemini's features. So yeah, the front end of Google Gemini finally getting some love. I've been saying for a year that Google is neglecting this and it's costing them billions. So yeah, now they're actually doing something on the front end.

All right, last but not least, Salesforce. We talked about this yesterday. We previewed it, but now the announcement is out. So Salesforce is planning...

an expansion in its Salesforce team with 2000 new hires to sell AI. So yeah, they're hiring 2000 humans to sell an AI that is supposed to sell better than humans confusing. Uh, so CEO Mark Benoist announced the hiring plan during a company event when they launched agent force 2.0. Uh, so they're doubling the initial target of humans hired, uh, that was shared just a month

prior. So Salesforce has already received 9,000 referrals for these positions indicating a strong interest in selling an AI that will eventually sell better than humans. Okay. So the hiring surge comes ahead of the release of the second generation of Salesforce AI, Salesforce's AI agent software set to release in February. Uh, so.

It's an interesting one, right? We're hearing this agent force is going to sell better than humans. It's going to cost, I believe, $2 per conversation. So it works with all your Salesforce data. And it's essentially like, yo, I'm going to text this customer for you. I'm going to email this customer for you. Cost $2 each time. But the thing that doesn't add up to me, I don't know about you all, but if they're trying to sell an AI product,

That is supposed to sell better than humans. Why are they hiring humans to sell the AI that I don't know? Confusing to me. All right, let's get into it. Y'all let's talk about Google deep research. Let me be honest. If you, if, if, if you're listening to the show, I don't how, how should I say this? I'm honest with y'all. If I like something, I say, I like it.

If something stinks, I say it stinks, right? Even though I know, you know, there's people at these companies, you know, I message with them, I talk with them all the time. And sometimes I'm like, ah, I might feel bad if I say, you know, product from Microsoft stinks or product from Google stinks or product from OpenAI stinks. I don't care. All right. And if I'm being honest, Google Gemini on the front end has been absolutely terrible.

Up until like two weeks ago, because generally on the front end of Google, you're only working with models that are three to nine months old. All right. And to actually get a true understanding of what Google is putting out there in its Gemini chatbot, you have to be a little technical. You have to go to Google's AI studio and you have to go to or you have to go to its Vertex AI platform. So those are generally for developers. But I've said this all along.

People are making decisions by going to the front end of Google Gemini. And now the front end of Google Gemini. So if you go to like Gemini.google.com, right? That's the front end, like the front end of chat GPT, the front end of Claude, right? So we're not talking about the backend. So Google has been always crushing it on the backend, but on the front end, I've been scratching my head for so long. So finally, uh, Google Gemini on the front end has gotten some love. So you do have to be on a paid plan.

to get deep research. But we talked about now there's 2.0 Flash on the front end. Now there's this new 2.0 experimental 1206 and this deep research. So in the last week, three new modes or features or models have been added to the front end of Gemini. So Gemini is finally good on the front end or good enough. But I think deep research is a sleeper. So essentially, think of

perplexity, right? That's the easiest thing to say. It's funny, right? Because I think about Google deep research. The first time I used it, I said, perplexity is cooked, right? Like there's no way perplexity will survive this unless they drastically improve. And y'all, I love perplexity. I've been using it since literally day one, since it was released to the public. I've been paying for it every month. I'm a big fan. Although recently, I think some of the quality has gone down. I think the iterative ability to work with perplexity has gone down.

I think especially ever since, uh, at least for what I personally use it for ever since perplexity announced it's shopping features, it doesn't really, in my opinion, do a good job anymore comparing products and services. It really just tries to sell you them. So, uh, perplexity I think has gone downhill, but it goes to this thing of being an answers engine, right. Um, and, and how we use the internet. And I think it's already changing, right. Um,

And my barometer of that is my wife, right? Like now that she's starting to say, oh, let's perplexity this or, oh, let's chat this, right? But I think this isn't just in the AI circles anymore. We're all starting to turn to large language models to expedite the process of research, right? Let's be honest. There's a problem. The problem right now is the internet sucks.

It is an absolute terrible place. All right. And partially it's because of large language models, right? Especially over the past 18 months, so many big publishers, so many, you know, very popular websites online are losing traffic. So what that means is they have to make their website experience absolutely terrible, right? So let's just say big website, all right? Big content website.

I don't know. I'm going to throw out Buzzfeed or something like that. I don't know. Right. Let's just say they are making a million dollars and, or maybe that's not a good example. Uh, let's say, um,

It's like Cora. I don't know. Right. So somewhere that you would generally go for answers. Right. But all of these content, uh, companies out there, all these media publishers, they make money by visitors, right? They have display ads on their website. They get, I don't know, a penny partial, a penny when you go visit their website, uh, you know, and then when you click on something, they get money for that too. Right. That's how the internet works. So

So now that more people are using large language models and they're not going to the websites anymore, all these big content publishers are bleeding and they're dying. So they say, all right, well, hey, we may be down 30% users. So we just have to increase the number of ads on our website by 30%. So the internet browsing, the internet is a complete wasteland. It's absolutely terrible, but ultimately it is terrible.

the large language model's fault. It is AI's fault, right? Without getting into this too much, but essentially large language models, they just steal all the quote unquote steal. They use, they borrow all of the internet's information and you don't really need to go to those websites anymore, right?

So Google deep research does that at scale, right? So as an example, perplexity, depending on what you ask it, it might quote unquote browse, you know, five to 15 websites and say, Hey, we're going to save you the trouble. And we're going to pull these main points. It's an answers engine.

Okay. You don't need to browse, right? You don't need to have 37 tabs open and your computer's humming. You know, you don't need to do that anymore because we have things like chat GPT search. We have things like perplexity, but now I think we have the best one and it's not even close in Google's deep research. All right. And Hey, live stream audience, keep getting those suggestions in, but have you used deep research and podcast audience?

I always keep my email, my LinkedIn, hit me up. Let me know if you've used deep research. All right. So like I said, here's what it is. It is an AI powered research tool by Google inside of Gemini. So you get it on the front end of Gemini and then you go to the dropdown menu and it automates in-depth web research and report generation. So I kid you not the very first time I used it, it visited 169 websites,

All right. And it pulled the important information from 169 websites and it did deep research, right? And it was freakishly accurate. All right. So it's designed for tackling complex topics efficiently. So right now who can access deep research? Well, uh, it's available to Gemini advanced users. All right. And right now it's only on the desktop, which kind of stinks. Um, and it's,

Uh, it's not yet available on mobile or the mobile app. So not a huge fan of that. I would love to be able to do this. Uh, here's, here's a little tidbit about me. Y'all, um, most nights, right? If I can't sleep, whatever, I'm usually using perplexity or chat GPT to do some last minute research for my show the next day. Right. Um, there's a lot more planning that goes into that, but you know, I might be up and thinking, oh man,

I need to know a little bit more about quantum computing, right? What's a qubit again, right? So I'm generally using these a lot. I would love to be able to use this on my phone, but I can't yet. All right, but mobile access is planned for 2025. So what probably a lot of you all are thinking, well, how does this actually compare to perplexity and chat GPT search? So maybe if I can make this go a little faster and stop rambling on, maybe we can do some head-to-head research.

But essentially, Perplexity focuses on real-time, accurate, cited information. All right? That's a good way to think of it. ChatGPT Search. So that's a new product from ChatGPT that was released, I believe, the last week of October. So ChatGPT Search is just more of a broad conversational search where I think Perplexity –

doesn't handle kind of open-ended things very well, I don't think. I think you have to be a little more targeted with how you use perplexity. Think of how you would traditionally search on the web. That's kind of those type of queries, right? So not as good as these open-ended queries or multi-step researching unrelated things. I think that's what ChatGPT is great at. But deep research is just something else, right?

It's a different kind of animal. It is literally like having a human sit down and research a very complex task that in general would require a human to read online all day, many hours, right? So that's where deep research really shines. Just being able to go beyond a simple answers engine and provide you an answer. But it's going to give you in-depth information

comprehensive, very impressive, fully cited research. I love it. So here's what it's good for. Collecting and analyzing extensive data, marketing analysis, competitor benchmarking, and automating complex research workflows.

Here's how it works. So it essentially uses advanced reasoning for deep dives, right? So you give it a query before you start, all right? Because it takes a while. We're going to get to that, some of the pros and cons. Before it starts, it gives you a plan. It's like, hey, based on, you know, whether you do a simple query, whether you do a very advanced query, right? It essentially is like, all right, well, here's the plan. Let me know if this is it. Love that, right? Yeah.

I wish perplexity would do that because sometimes perplexity just gets things wrong.

Uh, it analyzes multiple sources for insights. Like I said, the first time I did it 169, I was scratching my head. I was like, wait, what? Uh, I would say most of my queries get anywhere between 40 to 200. Um, my biggest one, I think it went to 250 websites, which is wild. Right. And then it produces structured detailed reports that you can either copy and paste in the window there inside Gemini, or there's a button to open it up inside of Google docs.

So here's the downside may have gaps in accuracy or depth. Um, also what you can ask it. It's confusing because a lot of times you're going to get essentially a content moderation error. And it's like, Hey, I can't help with that. Can't help with that. Can't help with that. And it's

And it's like, okay, well, why? Things that you wouldn't even think. But in my research so far on deep research, and I would hope that Google could give a little more context on what's allowed and what's not allowed, because even simple things, right? So in my experience, if you ask deep research as an example to compile a bullet point or a bullet pointed, bullet pointing to unrelated things, or creating a top 10 list,

or, you know, looking at three different topics that may be unrelated on the surface and trying to compare and contrast them. So for certain things that are maybe unrelated, for multiple step research that's not really aligned, or sometimes just...

haphazardly, right? Generative AI is generative. I've run the exact same copy and paste kind of prompts inside Google deep research. And half the time I'll get a content moderation, half the time it'll work and crush it, right? Right. So keep that in mind. This is literally brand new. It's been out for less than a week. And it is extremely impressive technology. And there's also potential bias, right? That's the other thing you have to keep in mind. Scrapes the web. Guess what?

The internet is a wasteland. The internet is full of bias. And these models are new. So sometimes you might see something from a chat GPT, from a perplexity, from a Google Gemini deep research. And you're like, this is full of bias. Guess what?

That's what's in the information it's ingesting. The internet is full of stereotypes. It's full of hate. It's full of misinformation, right? People always lose their frigging marbles, right? Let's say you get a hundred answers from chat GPT or a hundred bullet points or a hundred sentences, and there's one thing that's incorrect, right? And you're like, oh my gosh, hallucination. Models can't be trusted. I don't know. Give me the smartest human in the world. And if I can ask it subjects from any range, it's going to get most of them

Wrong. That's the thing I don't understand about, right? Something you get from Google Gemini, ChatGPT, perplexity, et cetera. It's not fact written in stone. Literally at the bottom of any large language model, it says, hey, this could be wrong. Always check, right? Human in the loop. You got to check. You got to look. You got to be smart. You got to use your head, right? Look at the sources. But still, I think Google deep research can take tasks that would take humans more than a day

Let me say that again. Google deep research can take on complex multi-step researching tasks that a human would do. And a human might take more than a day and it can do it in a couple of minutes. All right, let's look live y'all. Also get, get those examples in. There's a lot of comments coming in from, uh, from YouTube. Uh, so thanks for, um,

our YouTube audience. Uh, but yeah, please get an example in. So I'm going to launch one example live and this takes a couple of minutes. All right. Uh, so let me just say that it takes a couple of minutes. So I'm going to do this example. And while this example is going, that's the fun thing I like about doing these things live and, uh,

Trying to also read, you know, a couple dozen comments is it's a little tricky. And sometimes it takes me a hot minute. All right. So let's go ahead. So podcast audience, I'm going to do my best to describe what we have going on here. All right. And live stream audience, let me know if you can see here. So I should have mentioned this earlier. I have Gemini Advance on my personal Gmail account and also my Workspace account.

It is not available for many workspace accounts, right? Even though I pay, I think $30 a month for Gemini on workspace. And I paid $20 a month for Gemini advance on my personal Gmail account. Right.

I don't have it. So if you're wondering, if you're on a workspace account, so, you know, if, if your email is, you know, john at company.com and you go to your company.com version of Google Gemini and you log in, it's not there. Don't worry. I don't know why. So many of these things don't get rolled out to Gemini advanced workspace accounts. So you might have to be on a personal Gmail. That's where I am right now. All right. So I'm going to go ahead and put in this prompt. I'm going to read it here.

But first, all right, so I just went to Google Gemini on my personal Gmail account in the dropdown window, right? So now you have 1.5 Pro, 1.5 Flash, 1.5 Pro with Deep Research, 2.0 Experimental, and the new 1206 2.0 Experimental. All right, so I am in the 1.5 Pro Deep Research. I just put a big prompt in there. All right, I'm going to read it. I'm going to read it here in a second.

Okay. And now it's essentially asking me to confirm. Okay. So I'm going to go ahead. Sorry, I got multiple windows going here. So I'm going to go ahead and confirm this and click start research. All right. If you're listening at home, if you're on our live stream audience, go ahead, click your stopwatch and go. All right. So let me go ahead and read this prompt that I put in. All right, here we go.

Hey, this is Jordan, the host of Everyday AI. I've spent more than a thousand hours inside ChatGPT and I'm sharing all of my secrets in our free Prime Prompt Polish ChatGPT course that's only available to loyal listeners like you. Check out what Mike, a freelance marketer, said about the PPP course. I just got out of Jordan's webinar.

It was incredible, huge value. It's live, so you get your questions answered. I'm pretty stoked on it. It's an incredible resource. Pretty much everything's free. I would gladly pay for a lot of the stuff that Jordan's putting out. So if you're wondering whether you should join the webinar, just make the time to do it. It's totally worth it. Everyone's prompting wrong, and the PPP course fixes that.

If you want access, go to podpp.com. Again, that's podpp.com. Sign up for the free course and start putting chat GPT to work for you. Here's the prompt I put in y'all.

And then let me know how long this would take you. If you had no clue, if you didn't listen to the Everyday AI show or read our daily newsletter and someone asks you this, ready? So here's what I just asked Google Gemini Deep Research. I said, OpenAI and Google have collectively released more than a dozen noteworthy new AI products and updates over the past two weeks.

from OpenAI's Sora and projects to Google's VO and Gemini 2.0, there's been a lot. Please research deeply and detail every AI release from OpenAI and Google over the past two weeks. For each, please tell me what the update or release is, what it does, who it's for, when it's available, and the three main competitors for each product and service. Woo!

Imagine, imagine if you haven't really been paying attention, right? The last like 10 days in AI, it's by far, Google and Gemini have both been releasing updates that would normally be their biggest update of the month or of the quarter or maybe of the year. They've been doing it every single day for the past 10 days.

So let's say you were just on vacation and you know AI, right? And you walk back and you're like, wait, what? This would take you? I don't know. This prompt, we'll see how it does. Or let's say you are new to AI and you're like, all right, I'm starting now. I want to get caught up. I want to know Google, what Google and OpenAI are up to. Take you multiple days, if I'm being honest, right? I am asking a ton here, all right? I'm not just saying, what did they release? I'm saying,

I need every single AI release from the two companies. Not just that. I need to know what it is, what the new product or update does, who can use it or who it's for, when it's available, and the three main competitors, right? So as an example, I know that Google released their VO2, which is amazing, by the way.

I think it's already hands down better than OpenAI Sora. All right. If I had access to VO2, we'd be doing a show on it. Yeah, I got to. It's funny. I'm messaging and chatting with all these like VPs at Google. And sometimes I just forget like, hey, yo, can I get access to this? Right. Anyways, this would take me a while. Right. It would take me a long while. So let's see how Gemini is doing.

So now I am back in the deep research panel. It's already been to 39 websites. All right. So I can scroll through. So the way this works for podcast audience, right? It's kind of this new dual pane, right? So I have the chat on the left-hand side and Google deep research says I'm on it. I'll let you know when your research is done. In the meantime, you can leave this chat. All

And then it's essentially, it has a little icon with kind of a status indicator. And as soon as it goes all the way around in this little circle, that means it's done. But I can see in real time what websites it's going to. So it's going to searchenginejournal.com. It's going to techradar. It's going to openai.com. It's going to moz.com. It's going to cloud.google.com. It's going to YouTube. It's watching or analyzing YouTube videos, about a handful of them.

It's going to wikipedia.com. It's going to slash dot. It's going, right? It's going to so many of these, so many of these, right? Data camp, microsoft.com, Tom's guide, right? This is so many. All right. So here's another thing. Sometimes, all right, it's done. Was it, was anyone keeping track out there? How long was that? It's done. I don't know. I think it was like three minutes. So this is not fast. It's not supposed to be. All right. So,

Uh, let's go ahead and click show you what you can do here. And let me tell you what deep research is and what it isn't. All right. If I go in now, and if I see a quote unquote problem in this research report, deep research is not good at iterative prompting. So what that means, uh, I'm not going to be able to be like, Oh, like go back into Gemini and say, Oh, make it funny. Redo this. Right. That's not what it's for.

What it excels at is doing one-off deep research. Okay. And if you need to modify this content, improve it, um, you know, change the formatting, things like that. That's not what deep research is for, right? You can copy and paste this in Google. Uh, you can also click open in docs. All right. So let me just do that quick.

Because a little, I wouldn't even say this is like a cheat or a hack, but something that I don't think a lot of people are doing is, well, you can open it in Google Docs, right? And then you can click the little Gemini guy right there. Although I do have to say,

Not a huge fan of the Gemini implementation in Google's products. So Google box, Gmail, uh, sheets, et cetera. Not great. Uh, but you can go in. If you need to work with this doc a little bit, you can open it up inside of Google box. All right. So let's look at how it did. All right. I'm going to try to be quick here.

All right, so it starts. OpenAI and Google unleash a torrent of new AI products. So we have a nice intro paragraph here, and then it goes into OpenAI's offerings. All right, correctly. It said OpenAI has been making waves with its 12 days of OpenAI event. Correct. All right, I love this. All right, so it actually created a chart. I didn't ask it to, but it created a simple chart and said export to Sheets.

So let me just click export to sheets, right? I didn't tell it to. It just decided, although this chart stinks, right? It's like one thing in there. It says ChatGPT Plus, ChatGPT Pro. Yeah, it didn't complete the sheet. So this is experimental, but pretty good. All right, so let's see how it actually did. It said Sora. All right, so it gives me a good update on Sora. Then it goes into OpenAI01. So something that you can do here,

is there's always going to be these little carrot dropdowns where it says, learn more. Love that. The UI UX of all Google Gemini products. It is chef's kiss. I love it. All right. So you can click this learn more button there, and then it gives you the source. All right. So if you're reading something in this report, and if it's like, is that right? Or is that up to date? Is that accurate? Right. I can click learn more. Um, and then, um,

I can click it. Right. So this first, uh, this first iteration was a YouTube video and it's like, all right, well from someone, uh, good. Okay. Yeah. I know who that is. All right. The AI advantage. So it's like, all right, that's a trustworthy source. Right. Uh, but you might want to research a little more. And if you want to learn something more about one of these things, there's little dropdowns essentially after each paragraph where you can learn more. All right. So it got Sora, it got our open AI. Oh, one, uh,

Let's see, chat GPT updates. It got projects, voice and video in voice chats, correct. The Santa voice, correct. Chat GPT search with voice, correct. API updates, that just came out hours ago. Nailed it. All right, so this is good. This is up to date. I'm trying to think. There was one or two things I don't think I saw in here. There was an update to Canvas.

Yeah, but for the most part, it hit just about everything, right? Very, very good. On the first part there, I would grade Google Gemini deep research at an A minus. All right, so let's see. It says Google's, now it's going into Google. Google counter moves. All right, so Gemini 2.0, got it. Multimodal output, native tool use, enhanced performance, deep research. There we go. This is meta, right? Talking about the very thing we're covering. VO, right?

implications and future trends and conclusion. So oddly enough, all right, and we have another little chart here where it's comparing Sora and VO. Great, right? All right. Video quality, Sora is 1080.

VO is 4K. Length, Sora, 20 seconds. VO, not specified. So it did a really good job of creating almost this rich multimedia-esque environment, right? So think of the future. I can see something like this, you know, creating images with Google's Imagine. It might create videos with VO, right? It might create interactive websites. So we've covered other Google tools that do something similar to that.

All right. So, but let's see. I did notice. Let's see. I did notice it didn't get everything, right? Because I asked for competitors. So let's see.

Okay. So, okay. I'm going back and I'm reading it a little more and it did get the competitors. So like for Sora, it says some of the top competitors include Runway ML, Caber AI, and Pika, Luma Labs, Dream Machine, Kling. Yeah. Okay. So that's good. So let's see if it got competitors for each. It gave me competitors for Open AIs 01. So it says DeepSeek, Nova, Quan. Correct. Well,

Let's see, VEO, same thing, OpenAI Sora. Okay, good. Pico Labs, Runway. So it looks like it only gave me competitors for brand new products that were released, not product updates. So I could have maybe been a little bit more clear in that. But overall, I think it did a really good job, right? So let's go ahead. I'm going to try to find some examples here.

All right, let me see. So y'all, give me a second. Michael, not quite transitioning from everything ChatGPT to Google. Not just yet. You know what? Like three months ago, I would tell people do not touch Google Gemini front end with a 10-foot pole. Absolutely terrible, right? There was even problems early on on the front end of Google Gemini where it couldn't even access the web. It said it could, but it couldn't. I did dozens of videos. For a long while, Google Gemini was straight up

I want to be careful with my words here. I'll say Google Gemini for many months was a waste of time using it on the front end, on the backend, Google AI studio, top of the class, Vertex AI, outstanding Google Gemini in the front end. I told people don't touch it. You can touch it now. It's okay. The stove is no longer hot. You won't get burned. All right. Water's good. All right. Let's see. I'm trying to scroll through here. Y'all. So, uh,

If you did get something in, I'm looking through the comments here to see if we can find something that's easy enough because I do have to kind of type it in. All right. But I wanted to get something. Okay. I got – there's some fun ones here. There's some fun ones here.

Scrolling through here. Let's see. Three ideas from Denny. All right. Let's see what we have. I'm going to read this out. Sorry, y'all. It's a little bit hard to do this live and to also, I should start using like a hashtag or something so I can more easily scroll through these couple dozen comments and get a good example here. Let's see. Don't have a bedroom laptop. Don't have a bedroom laptop, Lee. Let's see. Here we go.

All right, some examples from our audience. So Allison from YouTube says, I write haunted history tours for destinations around the US. So I do a lot of historical and academic folklore research. That's fun. Jackie saying, I am working on research for the optimal mix of faculty type for a university in roles for each type, being 10-year term adjunct. Okay, I like that. Sabbatical life.

Tiny house communities in each state. How many and where they are. That could be fun. Uh, Ben is saying deep research topic. What are the contrasting internal and external forces driving the useless and non-informative naming of new models? Uh, and how can each company find a more customer friendly? Uh, that one's good. Uh, Michael saying drones in New Jersey, LOL. I would, I would like that. Um,

Let's see. Denny says three topic ideas for deep research, the economics of universal basic income, the science of magic tricks, the science of climate change. Oh, okay. I like those. All right. I think Denny, I think I'm going to choose one of yours. It's probably one of the easiest. I think I'm torn between two. Okay.

All right. I'm actually going to choose this one from sabbatical life. All right. It's, I think it's a little easier for me to say like, this is right. This is wrong. All right. So I'm going to go into new chat.

Again, I am in Gemini deep research and I'm going to type in now. I'm going to say, please, we're going to do the tiny house communities because this is actually, I think it should be a little difficult to research in 50 states. All right, let's see. So I'm going to say, please research. I like to say, please deeply research. Still doing some also FYI, a little tip.

Even saying things like that, encouraging it to go deep. I've done like, you know, kind of like some split testing on this. So you can run a simple prompt and be like, oh, give me, you know, tiny house communities in 50 states. But, and you're not going to get as deep of research as if you tell it to research deeply. I know that sounds counterintuitive or redundant, like, oh, it's deep research. Why would you encourage it to give you all the details, et cetera, et cetera. It actually helps. All right. So I'm going to say, please deeply research tiny house communities.

I love typing live. Tiny house communities in all 50 states in the US. How many are in each state? Where are they? Okay. And then I'm going to say, please, please take your time and make the research comprehensive, bringing in as many sources as possible to give me a

holistic view of tiny house communities. I've kind of always wanted like a tiny house, like somewhere on a beach somewhere. So I'm curious. All right. So let's see how, uh, let's first see if it will do this because like I said, sometimes if you give it too many specific directives that I guess it deems aren't necessary, it won't do it. All right. But it's working. So, uh, it's saying here's essentially the plan. All right. So, uh,

Essentially, here's what it does. It takes your prompt and then it breaks it out into steps. So here, let's see if I can click start research. I'm going to copy these. All right. Luckily, I copied them. So here's what it's doing.

So it gave me eight steps from my little prompt. It broke it down to eight steps. It said, number one, find tiny house communities on, in each of the 50 States to find the number of tiny house communities in each state. Three, find the locations of tiny house communities in each state for find information on the different types of tiny house communities in each state. Five, find information on the regulations and zoning laws, right? I wasn't even asking about that, but that's cool. Six.

Find information on the cost of living. Okay, that's cool. Seven, find information on the pros and cons of living in tiny house communities. Eight, find information on the environmental impact of tiny house communities. Pretty good research plan to me. All right, so let's see. It's already hard at work and it looks like it's about a fourth of the way done. All right, so I'm going to hit pause.

On that for a second. And I'm going to do a very, very quick comparison. Yeah. Unfortunately I have to switch a Chrome pile, a Chrome profiles here, and I'm going to go back very, very quickly. All right. And do that same comparison that I did for open AI and Google, because it's going to take probably about three minutes for this deep research on our tiny houses to change.

So I'm going to do the same thing, our previous query inside perplexity and chat GPT and quickly compare the results. So remember, I am not in the right mode. All right, there we go. So remember, I'm saying OpenAI and Google have collectively released more than a dozen noteworthy and new AI products and services, products and services.

products and updates over the past two weeks from open AI Sora and projects to Google's VO and Gemini 2.0. There's been a lot, please research deeply in detail. Every AI release from open AI and Google over the past two weeks for each, please tell me what the update release is, what it does, who it's for, when it's available in the three main competitors for each product service.

All right. So I quickly did the same thing for chat GPT and perplexity and let's see. So obviously what you notice is chat GPT and perplexity normally do these in five to 20 seconds, right? Where Gemini deep research takes a couple of minutes. In my experience, depending on what type of query you ask, how in-depth it is, anywhere from two to eight minutes, right? And I did a lot of testing when it was first released. So maybe the servers were a bit overwhelmed.

So let's compare here. All right. Let me try to zoom in so our live stream audience can see. So ChatGPT, let's see how ChatGPT did. OpenAI's releases, Sora, what it is, functionality, target audience, availability, main competitors, ChatGPT Pro, same thing. Good. That's all it gave me though. That's all it gave me. ChatGPT only gave me Sora and only gave me ChatGPT Pro, even though

I said, there's been more than a dozen between the two. And I asked for information on each of them. So although ChatGPT did a good job with Sora and ChatGPT Pro, it missed like eight things. Let's see Google. Gemini 2.0. Got it. What it is. Functionality. Target audience. Availability. Main competitors. Cool. VO2. Same thing. What it is. Functionality. Target audience. Availability. Main competitors. WISC.

All right. That came out yesterday and that's it. So it gave me two from chat GPT and it gave me three from Google. Not that great.

All right, let's see perplexity. So what perplexity is, and FYI, I'm obviously on the paid tier of all of these. So I saw that perplexity actually used its reasoning. So it has kind of a reasoning mode. And also perplexity uses a large language model to write its content. But the reasoning mode is perplexity, but I have it tapped into Claude Sonnet 3.5.

So let's see how this did. And I am keeping my eye on our tiny house research. It is about halfway done. This one looks like it could be a big one. All right. So here we go. Sharing perplexity. Let's see how perplexity did. So perplexity made a little table. All right. So it has the release name, the description, the target audience, the availability date, and the main competitors. So let's see how many it actually did. So it got Sora, Gemini 2.0,

Chat GPT-01, Chat GPT-Pro, and that's it. Not that great. Not that great, right? So both Chat GPT and Google clear, or sorry, both Chat GPT and Perplexity failed, right? All right. And I'm curious. I'm going to go ahead and

I'm going to run the same prompt very quickly, just in normal Gemini, because yes, Gemini does have access to the internet, obviously. So I'm wondering how just normal Gemini advanced does not Gemini deep research. All right. So in this instance, there we have Sora, what it does, who it's for availability, competitors, et cetera. Chat GPT-01. Good. Chat GPT-Canvas. Good. Projects. Good. And then Google. Let's see. We got Gemini 2.0.

And it has what it does, who it's for, availability, competitors, VO. All right. So none of the quote unquote base large language models did a great job, right? I'll give them maybe a C plus B minus where I think overall Gemini Advanced probably was an A minus or sorry, Gemini Deep Research was probably an A minus. All right. So overall, you can see even these internet connected large language models, they only visit, right?

They only visit a certain number of sources, right? So let's see in Gemini or sorry, in Perplexity. I'm tongue-tied today, y'all. I can see at each step how many looks like Perplexity went to about 20 sources. I can scroll down on ChatGPT, click sources. It actually went to more. It looks like ChatGPT, looks like about 20, give or take. Gemini, Perplexity.

Uh, Gemini, I can see sources and related content, scroll through them. Oh gosh, there we go. Uh, okay. So they each went to, it looks like about eight to 20, whereas Gemini, uh, Gemini on that one went to a ton. All right. We are finally done with our tiny house. I'm sorry. I know this one's a little winding, uh, but let's, let's go ahead and quickly, uh, look at our tiny house example.

I'm curious. I don't know. Are you guys curious about tiny houses in every single state? Let's see how it did. So zooming in here, I'm trying to think where, how many it went to. Let's see a lot. It went to 92 websites. Imagine going individually to these 92 websites. Let's say you were looking to move to a tiny home, right?

Quitting the rat race, going to find me a piece of land in one of these 50 US states. This could have taken you weeks, right? Because all of a sudden you're on, you know, ctiny.homes or you're on atkincottages.com and you get other ads and all of a sudden you're back to, you know, quantum computing research and watching everyday AI YouTube videos on accident, right? The internet is a very distracting, terrible place. So now...

Advanced research or sorry, uh, Google Gemini advanced deep research went to those 92 houses. It got me a comprehensive guide here. All right. So it has a couple intro paragraphs and then it's starting to break down a tiny house communities by state. Okay. Interesting.

So it kind of admitted, Hey, this is kind of tough. It said while pinpointing the exact number of tiny house communities in each state can be challenging due to ongoing developments and variations and definitions. This guide offers a comprehensive overview based on the latest information. Each state's unique approach to tiny house living is highlighted showcasing the diverse landscape of this growing movement. If it did this, I'm guessing it just did a couple of States. If it actually did every state, that would be very impressive. All right. So we have Alabama, Georgia,

And we have a couple, it looks like five tiny home communities and a little interactive chart that it dropped down. Alaska, Arizona, Arkansas. Okay. I was going to say, there's absolutely no way it did this for every single state. So it gave me a sampling. All right. So

Uh, it says this is a, at the end of this, I actually haven't gotten this yet. So thank you sabbatical life for this idea. It says, this is a partial response as the full response exceeds the token limit. I can continue providing information on tiny houses in the remaining States. So generally, like I said, um, it's usually not good at continuing research or iterative prompting. So I'm just going to say, uh, in your response, you said this, uh,

And I'm going to say, your last state completed was Arkansas. And I'm going to say, please continue where you left off. I have no clue if this is actually going to work. This is already turning into a marathon podcast. So if you want to know if this actually worked, can Google Gemini get to all 50 states and give us – it looks like it's –

Starting to, we'll see. All right. I'll let you know. We're going to, we're going to tease. We're going to tease the newsletter. If you want to know how this turns out, if you really need to know information about 50 States inside of, or for tiny houses, go to our website, you are everyday AI.com sign up for the free daily newsletter. All right. So let's wrap this up. What's our take on this? Number one, keep in mind, this doesn't mean hallucinations aren't going to exist.

All right, just because it goes to 92 websites or 169 websites or 290 websites, that doesn't mean it's going to be 100% error-free. So no matter what you're using Google Deep Research for, always keep in mind it could have a lot of errors.

All right. So even as an example, most of the reports that it generates are between, at least for me and my experience so far are between six to nine pages. It does a really good job. It gives you all the sources so you can make sure, and you know, human in the loop, we still have to do our due diligence on, is this all actually correct? Is this factual? All right. I think this can take research projects that would generally take a half day to three days down to a couple of minutes. Let me, let me repeat that.

Your everyday research tasks that you probably don't even know that you're doing. If you work with this a little bit, it's going to, I believe, cut the initial research time down, I'd say easily by 95%. This is what we humans do, right? Whether you know it or not, you do...

You do one of these two things. If you're a knowledge worker here in the U S if you sit in front of a computer, a company pays you for your knowledge, right? Unless you're constantly interfacing with humans. But if for the most part, you're in, if for the most part, you're interfacing with the internet, which is a lot of us, that's not all knowledge workers, whether you know it or not, you're probably spending hours a day researching, reading, analyzing, creating content, right? Whatever that content may be.

95%, I'd say very easily. It's not a complete replacement. So there's that part. Or let's say you generally don't. Let's say maybe that's a couple of hours a week. Well, guess what? You can apply that same number of hours a week and get 10X the results.

Okay. So even if you are someone that is not spending, right, uh, hours a day, researching a certain topic, learning, uh, trying to decode something, trying to, uh, you know, study and analyze your competitors, all of those things. Even if you're not doing that, even if you're just doing more topical, lighter research on the internet, you can spend that same amount of time and go 10 X deeper. There's no reason not to, but here's, here's my last take as we wrap up.

What does this mean for the future of the internet? What does this mean for content publishers? What's this mean for me, right? Yeah, I'm being honest. A lot of my queries that I ask, it looks at our website, youreverydayai.com, right? It looks at our YouTube videos. Generally, how this thing works, how you can continue to watch and listen as I go on these long rants, people go to our website, right? Those people sign up for our newsletter, right?

A lot of those people that sign up for our newsletter will listen to the podcast. Then when we have great advertisers and sponsors like Microsoft, I can say, hey, we have this many people listening. We have this many people reading. So what happens when these answers engines like Perplexity, like ChatGPT, like Google Deep Research start taking that, right? I'm a human creating content for other humans. I love AI. I have nothing against this. But what happens? What's next, right?

I hope what this means is going to be an ad share model. I mean, it has to be right. It has to be in the long run because I'm paying $20 a month to use this.

And I don't know, I think it maybe should be something like Spotify, right? Maybe if you're everydayai.com in the same way, you know, on our YouTube channel, if you sit through those little 15 second ads or whatever, I maybe get, I don't know, a quarter of a penny or something, right? But you know, it adds up. But is it going to be the same thing for deep research? What's going to happen? Because now all of a sudden, and especially for larger content publishers,

they're losing and they're going to continue to lose as tools like deep research perplexity chat gpt search get more popular what happens who's going to pay these humans creating this great content for us to enjoy i'm not sure but i think it's a radical shift so let's see how it plays out all right hope this one was helpful y'all i know it was a long one we went all over the place

Took me like three minutes to read all your comments and suggestions. I appreciate it. Was this helpful? Let me know if you're still on the live stream. Sometimes I like to just, you know, do these just to see who's still listening. Let me know what your favorite coffee to drink is at home. I'm big on the Nespresso. I hope this was helpful. If so, please click that repost button. Yes. I hope that this is your cheat code. Shout a shout out. Someone, someone on LinkedIn said something super, super nice about,

Yesterday, where was it? I should have had this pulled up and ready. So yeah, Nathan, Nathan, thanks for these kind words. He reposted this yesterday. When I ask you to repost, especially if you're on LinkedIn, YouTube, Twitter, whatever, if you do, I really appreciate it. That's how this thing grows, right? Because these AI models are stealing all our traffic, right? But he said...

Jordan Wilson from Your Everyday AI is my cheat code, right? He said, hey, everyone's always asking me how I know everything about AI. This is my secret. So if we are your secret, thank you. But tell other people about it. Please subscribe on the podcast. Leave us a rating, repost, and go to youreverydayai.com. While you're there, sign up for the free daily newsletter. Recapping today's show, thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all.

And that's a wrap for today'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 youreverydayai.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.