Talking AI Live at Google I/O

0:00Matt Paige:Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters. I'm your host, Matt Paige, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI. We are live from Google I.O. with a special episode of Talking AI today. I'm joined by three of the top AI creators on the internet, Kushank Aggarwal, Marcin Teodoru, and Jay Enriquez. combined, these guys have a following of over 2.2 million people across the internet, which honestly just blows my mind a little bit, but welcome guys. Welcome to the show. Thanks for having me.

0:36Matt Paige:All right. So now we only have 30 minutes, so we're going to do, we're going to move fast. We're going to hit the biggest updates from Google IO and give you our unfiltered takes on what actually matters, what's hype, what people are actually going to use, and what could change the way we search, create, workshop, interact with AI. There was a lot that It was covered today, so we're going to try to get it in. But the thing that stood out to me is that this IO didn't feel like a bunch of random AI announcements and demos. It felt like Google is really starting to show us that Gemini is becoming this intelligence layer across everything they own.

1:08Matt Paige:Search, YouTube, Gmail, Docs, Chrome, Android, shopping, even glasses, which was really cool. And that raises the big question for today. Does Google's advantage come from having the best individual AI products or from having the distribution of billions of people. I don't know, Gishai, be curious your take on that to kick us off. Yeah, I would say compared to other companies, Google is not trying to bring people to AI. It's bringing AI to people. For example, they talked about AI mode having a billion users in the first year. That's only possible because everyone's already going to google.com for every single question they have.

1:46That's the distribution power.

1:47Matt Paige:Yeah, it's almost less about which AI tool is the best. And it's still critically important, but we're in this bubble. We talk about AI every day. Everybody's seen the viral chart of the different blocks. And there's this tiny little block of people that are actually using AI every day. But it seems like it's, if you can bring AI to this massive group of people, that's the strategy in a sense. I don't know. The Google search was the big one to me. I'm curious. I don't know, Marcin or Jay, it always felt like search was almost disconnected. They were trying to weave in AI, but this Google I.O., it felt like it's becoming a core, like AI is a core theme in it.

2:25Yeah, for sure. I guess one of the things that stood out to me is exactly what you were saying, Matt, around search, right? Because for the longest time with AI, one of the things that we were looking for is this everything app, or will we have an everything app? It will basically do all your shopping for you. And one of the things that stood out to me yesterday was when Google said that they were processing, like, was it like quadrillion? Oh, I got this stat.

2:48Matt Paige:9.7 trillion tokens per year per month. Two years ago. Two years ago. A year ago, it was 480 trillion. But then they came out with the 3.2 quadrillion tokens, which is like, y 'all know what a quadrillion is? Does anybody here know how many zeros that is? I think it's a million millions now. Yeah. We got to ask Jim and keep going on that point because the growth is just exploding with this token use actually. So we might find ourselves next year with Sundar saying another big number because once more AI is being served to more people on search, like search is the big thing. And it's very clear that directionally is where they're taking it.

3:23And so if you bring, and in fact, one of the things that you were talking about is bringing anti-gravity. Like where would have you thought about a genetic coding platform being brought to the masses more than search being integrated with something like anti-gravity, right? So like the billions of people who are using Google search now who can generate their own UIs is going to be big, not just in the token usage, but also in the experience of people and bringing them into the fold, let's say.

3:47Matt Paige:Most companies don't have an AI problem. They have a prioritization problem. There are hundreds of different places you could be using AI, but the hard part is figuring out which ones are actually worth investing in, which ones you're ready for, and what should actually come first. That's why we built Gen ROI at Hatchworks AI. Instead of basing your AI strategy on a few workshops or whatever happens to be in the room, Gen ROI uses AI-guided discovery to gather insights from across your entire organization. It helps us surface the themes, patterns, and connections that are easy to miss. Then our strategists work with your leadership team to turn those insights into a prioritized AI roadmap, weighted business value, feasibility, readiness, and what actually makes sense for your organization.

4:23Matt Paige:So if you're investing in AI, but still trying to figure out where it can actually create the most value, I highly recommend checking out GenROI. Go to hatchworks.com backslash GenROI or just hit the link in the show notes. Did y 'all catch that though? That was the other part that was so interesting on search and it was like other things too, but it was like this ability to build these generative UIs, custom dashboards. And if I remember correctly, it was almost like native in search. They had the whole thing with like docs kind of going after Whisperflow a bit. But any thoughts on that experience?

4:54Matt Paige:How does that habit begin to change for the rest of the world as they get this power? I think one of the great things is we're all builders here, right? And now literally 3 billion people get to actually see that by moment, where they're like, oh, can you make me a chart of a simulation of the solar system? And they see a build, they see that it's coding. So I think just giving people the ability to see that for themselves in search, which is kind of wild. And touch on your point. I had no idea search was the thing before coming here. I haven't used Google search in three years since AI, right?

5:25Since chat GPT moment. And then I'm like, once I looked at it, I was like, oh, wait, they have the infrastructure. They have 3 billion users. And now they're connecting the harness of Vibe coding to the masses. and it's kind of blown my mind because now I'm actually going to go in because I played with it. I was like, it's all real time. They have every app out there. They have every user and you can actually do it natively instead of using an API and trying to get like data from other sources. That's the part for me that just kind of blew my mind to see how that happened.

5:52Matt Paige:Does it feel like they've been kind of building this Death Star all along and now they're slowly like pointing it and they got search and they're just like ready to take off. But the point you mentioned, the democratization effect, I don't know your thoughts here, but like it's that moment i'm sure we've all had it to where you're like holy crap like i can build anything with ai and i wonder if this is going to be this like leap pad moment where people start to do more and then anti-gravity looks less scary and all the other tools that are out there because there's still that mindset that people think oh i i can't build i don't know how to code i think that would still be two different things to be bringing anti-gravity as just like something in the back and for the users.

6:32They don't even know it's anti-crackery that's when they are actually building something. To me, when they search, there's an intent. You just want to get to a result. You want to understand something. You want to get your question answered. And you're just able to get to the final result for the user a lot faster and everything is just happening behind the scenes. At the same time, I'm curious how Google is going to approach because their biggest revenue engine is search. If search is changing and you're getting away from the links and the AI mode and everything that's happening, how is it going to impact the small businesses who are relying on Google search results for everything for the business.

7:05So it's going to be tricky from that perspective, but from adoption of AI and just normal consumers, it's great because you bring the power of AI where people already are. Our moms are searching on Google. They're not going to go into Gemini for the most part. Unless your moms are using Gemini and like building things into it. So every time I talk to my mom,

7:23Matt Paige:she's like, you got to download that ChatGPT thing on my phone. She never does it, but like every point, but I'm sure that's with a lot of people, right? It's like they need that nudge. And maybe it is the fact that it's going to be native in there. That's going to do it. But you, you touched on something too. I don't know if y 'all caught this with YouTube and how that's evolving. I thought that was very cool. Cause it's like, and this, this is the shift from the query to the Tim blue links, it's almost going from like search engines, like action. Engine in a sense or something, maybe they should coin that phrase, but the cool part was it like takes you to the moment in the video that you need.

7:58Matt Paige:and it has contextual stuff there as well. Y 'all are all creators here. What are your thoughts on how that evolves discoverability and just that experience with users finding things via YouTube? I think for me, it was actually a moment where I realized actually to your point, which is local businesses. So people make content online, right? Short form is huge on YouTube. So now imagine being a local plumber and now you're making a piece of content saying, hey, I'm a local plumber in San Diego. And now through search, you're going to be able to find that exact video at that exact timestamp of the location, the reviews, all that stuff in a short form content.

8:34And obviously long form YouTube. Like to me, that was like the whole thing of like discoverability for people that never use it like us because we're creators, right? So we make the content actively. But for now, I think there's just going to be a whole new world for just the local businesses to do those types of that type of content and get discovered in a whole new medium. So it's very exciting, I think, in that sense. Yeah.

8:55Matt Paige:They had the example of your kid going from the balance bike to pedal bike. And they had the example of the dad with the towel wrapped around their waist to guide them. I was like, oh my God, that's genius. So I'm going to go try that when I get home with my four-year-old there. The next big thing, Gemini Spark. This, okay. So agents is obviously the biggest thing. So you knew Google was coming with something agentic in nature, but I'd be curious, all of y 'all's takes here, but I feel like everybody that comes out with something agentic, a lot of it has the high promise of it's the everything agent tool, whatnot, and then it falls flat, but I'm curious your thoughts on Gemini spar.

9:32Matt Paige:How do you think it's going to do? I don't know if anybody's tried it yet, but just the concept of it and bringing this agentic approach into Google. I think it has a huge potential, like directionally it makes sense for them to dive more into the space, but I feel like with a lot of these agentic platforms, it is very determined by the implementation of it. Like how good is it? Because I think right now it's not yet publicly available. It will probably be in beta in a few weeks or months. With a lot of these tools, though, the way that we use it is probably going to determine if it's really valuable or not.

10:04Is it going to replace Cloud Cowork for a lot of people? Is it going to replace the people who are using OpenClaw? Because when the product managers of Google developing Gemini Spark is thinking about that, they're probably thinking about those authentic platforms as well. And so with a lot of these Google tools, even though there's a lot of, let's say, combination of things into Gemini, they also mentioned a few new things that seem to be very Google-esque, like Gemini Spark is there. I think they mentioned this new tool called Google Pix, which are clearly slightly experimental and are very Google-y.

10:34So maybe Gemini Spark is one of that, but time will tell. It's only based on users trying it out, whether we will see it survive.

10:42Matt Paige:The one bit of feedback I got for Google is they have so many dang products. It's tough to keep up with, but all of them are so cool in nature as well. But I don't know any other thoughts on the agentic nature with, especially with it being in the Google ecosystem. I think that's like a key thing we see from a lot of companies we work with is that intelligence layer. Every company we talk to is getting huge value, but it's in silos. The individual's like doing some cool stuff for clock code and somebody's doing something else over here with open claw, but they don't have the shared context. In my perspective, Google is approaching all their product releases a bit different to other players in the space.

11:18Everyone else is building for early adopters, you know, CloudCo, Codex, for people who are at the frontier of adoption. But this Google I.O. felt like it's mostly meant for people who are not at the forefront of adoption. They are a little behind, and that's like where the bulk of comes in, the majority of people where they are. And I like the idea of Spark. And I like the fact that it's actually constrained to just Google products with MCPs rolling out soon. And the reason I say that is with any product, for example, Cloud Cowork, Cloud Code, Codex, you feel like you have endless possibilities.

11:50A lot of people feel scared and intimidated because they don't even know where to start. They don't even know what they can do. But the fact that you have a limited number of tools you can connect to, which means it's easier for them to communicate those use cases. for example with the daily brief that they launched within the Gemini as one of their agents so pre-packaging all of those things would make adoption easier in my opinion because for most people 80 % of the things you're going to be building is going to be the same you need something for emails you need something for project management you need something for communication so those things are the same regardless of what people do for work and only 20 % of the things which are custom for everyone's workflows so I think it's going to ease adoption for people and feel comfortable with the concept of agent.

12:30The other thing is people feel scared of agent. I think calling it Spark actually changes the narrative a little bit. Instead of saying, we're not saying, okay, let me just go put a query and get page ranking and see which page ranks were. So you just say, I'm going to search it. You go for intent based. You kind of change the narrative and agent is not an intent. So I think maybe we'll see how it plays out, but it could be interesting play if it really works. Let me just spark that quickly. I'm just going to spark. Let me just spark it and I'll get the result. You get to the intent-based forces getting stuck in the, oh, I need to build an agent.

13:03Building an agent just seems very heavy for a lot of people. Feels a lot of work, even though it's just a prompt. Yeah.

13:08Matt Paige:So A plus on the Spark name. We'll see if they turn that into a verb. I would like to add a share note on that. So one thing I thought about too is it definitely isn't going for the tech people, right? It's obviously going consumer. But the one thing I found interesting was they finally in the AI space, Google is starting to standardize things. So we obviously saw agent payment. They're the first ones. And what did they say on stage? We finally all agreed on something. The shopping cart stuff, that universal shopping cart. Everybody has the same point. So I think they're going standardization and becoming like the layer for everyone.

13:40And I think that's going to include agents. I think that's going to be the first thing that they're going to roll out. And basically, because right now you can't connect, obviously, OpenCloud. You can't bring stuff over. You can't take Hermes stuff. I think they're going to somehow figure that out because they're going to get a billion people using it versus 600 ,000 of us in the AI weeds learning OpenClaw. And then I think once they get enough data, I think they're gonna actually be like, okay, this works, this doesn't work. Let's standardize this. So then everybody can communicate each other.

14:04That's my thought on it. As soon as I saw it, I was like, oh, that's, I think where they're going.

14:08Matt Paige:I like to the agent, the daily brief they came out with. That's one of the first ones. I always start with a client. So you can get like instant value, connect your stuff in there. But on the enterprise side, obviously this is not the GCP conference, but they rolled out their Gemini agents, And Gemini enterprise agent platform, trying to remember one of the names, right? But they had a concept of an agent garden, which is the same thing. It's these agents that tons of people are using and they're packaging it and making it easier. It's such a big point on the ease of getting started and whatnot.

14:38Matt Paige:So all right, a few minutes left. I think the big thing, this is like the wow moment, especially y 'all are all creators here, but we had Google Omni, which was very cool. We had picks as well. Let's start with Google Omni. So this is their world model. Maybe Jay, what is a world model? What's the value there? I know Demis has some interesting perspectives on that. Past just, oh, I can create some really cool videos with me and them. Yeah, I think you can think of Omni as just a video model. But I think if you step back a bit directionally, you can see Google approaching this sort of path to AGI a bit different versus OpenAI and Anthropic.

15:15Because if you talk about OpenAI and Anthropic, like what's their solution to get to AGI? It's just solve coding, right? And everything will take care of itself. But something unique that Dennis has always talked about is that you need some sort of world model in order to achieve AGI because he's thinking about something broader apart from just computer use and computer work. He's thinking about like robotics and being able to simulate worlds, right? And so I think Omni is just sort of a very infancy stage product into something that is going to be folded probably into Gemini in the future, maybe.

15:46But I think based on how they are speaking, that seems to be it. Because right now, if you were to compare it as a video model versus something like CDance, for example, I don't think it matches up. It might have different use cases. I saw a take where it's like Nano Banana for video where you can edit things out and place things back in and can do it quite well. But if you are, let's say, wanting the best scenes for your video production, it might not be the best thing. Although, of course, time will tell because, for example, the API is not yet available, but the costs can be debated. But I think it's a much larger play than just being a video model.

16:18Matt Paige:I'm not betting against Demis and I'm not betting against Andres Carpathic, which any takes on him moving to Anthropik, that was wild. If anything, OpenAI is going to have to step up their game, if anything. Yeah, totally. People like, it's not just even the talent, they're losing the ultra, losing the marketing play with it. Yeah, exactly. So on the Omni, has anybody created a video yet? Because this one's actually available to everybody now, right? Yeah, I spent like an hour on it. It's pretty quick. So to do like a 10 second video, it took about 45 seconds, which I thought was actually pretty impressive.

16:50The Nana Banana kind of vibe was really good. So I just took a pre-made scene that they had. And you know, those pre-made scenes are always perfect. Yeah. And then just did a replacement, put my face in it from a low res image off of my website. And I was actually very impressed with that part of it. But when I did do, to your point, try to make a new video, I was like, hey, make me a video of Wes Anderson style girls spinning in slow motion in front of the Eiffel Tower with balloons popping up and the guy's kind of looking at me. I'm just pushing it to the max, right? And it was not that great, personally.

17:19But again, it's day one, right? And they're only going to optimize it. And they said it's all running on 3.5 flash, which was their big announcement. So once it goes pro, I think it's going to be like C-dance once to C-dance two moments. Yeah, it was interesting too. In the video space, yeah.

17:33Matt Paige:On the three, at one point I said, they mentioned that if Enterprises moved to 3.5 flash, they could save 80 % on their token costs. So you can see that's a big thing they're thinking about right now, but it's funny. They have the Nano banana booth. I don't know if any of y 'all tried that, but I, they had the preset things you could do. And I did a nineties one. They turned me into a mixture of Danny Tanner from, what is that full house and Tim Allen. It was very bad. We'll link that in the show notes, but the picks product, I don't know, Kishank, any thoughts on that? I think it's the editability was a really cool feature.

18:05Matt Paige:And with, I guess, I'd be curious because chatGBT images 2.0 is amazing. And I didn't see the nano banana enhancements of the model, but it was almost like this editability that came out. Which is usually the final step if you're using an AI tool. When you are using an AI image generation or video tool, you need to get into the editing. You just cannot have or expect a prompt to do this. Very fine tweaks, which is where I think the editability is the next stage. versus just trying to build a better model. Because you can try to build a better model, but even if it produces a great result, you may have to tweak something.

18:40You may want to move text from here to there. So I think editability, which is the usability, comes in to meet that, to finish the last mile, is where companies should be building products. I think Cloud Design did a good job at it because everything you design within Cloud Design has the editability layer. Yeah, it's a product. Which turns into a product overall. Yeah, I agree. I think it just becomes more like an actual tool for users, right? Like Photoshop. You can go in and just change one line. And that's always been missing. And so my designer's been using, obviously, design tools for so long now.