Digital Marketing Masters Podcast

Share on social media:

In this episode of the Digital Marketing Masters podcast, host Matt Rouse welcomes Ethan Holland, an expert in AI and digital transformation.

Ethan's AI News:
https://ethanbholland.com/category/this-week-in-ai/

AI Local Marketing for SMBs: https://smbautopilot.ai 

https://serve.podhome.fm/digital-marketing-masters

(00:00:02) Introduction and Guest Welcome

(00:01:09) Ethan Holland's Career Journey

(00:02:11) AI and Its Impact on Marketing

(00:04:32) AI's Role in Marketing Strategy

(00:06:21) SEO and AI: The Future of Search

(00:12:01) AI's Impact on the Workforce

(00:18:01) The Future of AI and AGI

(00:31:10) AI in Video and Image Creation

(00:45:18) AI's Influence on Social Media

(01:01:18) The Future of AI: Blind Spots and Opportunities

They discuss Ethan's extensive background in legacy media and his passion for AI, highlighting his blog that tracks AI developments with over 21,800 organized links. The conversation delves into the (now) release of GPT 4.5 and the potential of GPT 5, exploring how AI tools might evolve from tactical aids to strategic decision-makers in marketing.

Ethan shares insights into the future of work, suggesting that while AI might disrupt certain industries, it could also augment human capabilities, particularly in creative and strategic roles. They discuss the potential for AI to level the playing field by providing expertise across various domains, but also the challenges of widespread adoption and the risk of job displacement.

The discussion shifts to the impact of AI on social media and content creation, pondering whether AI-generated content could overshadow human creativity. They explore the implications of AI in video and image generation, with Ethan highlighting advancements in image-to-video technology and the potential for AI to transform media production.

Matt and Ethan also touch on the societal implications of AI, including the concept of universal basic income as a response to potential job losses. They consider the global distribution of AI technology and the importance of ensuring access to AI tools across different regions.

Finally, they discuss the future of AI in customer service and the potential for AI to enhance user experiences by providing efficient and personalized interactions. Ethan emphasizes the importance of embracing AI's multimodal capabilities, suggesting that vision and audio integration will be key to future developments.

[00:00:02] AI Narrator #1:

Welcome to the digital marketing   masters podcast with your host, Matt Rouse. Today, our guest is the one, the only   Ethan  

[00:00:14] Matt Rouse:

Holland. Hey, everybody. Welcome back to digital marketing masters. Your host, Matt Rouse. Today, I've got Ethan Holland. Ethan, how you doing? Great. How are you, Matt? I am doing fabulous. Waiting for this freezing rainstorm to come in and knock out my power. So hopefully, we'll get through this before we lose power here. And   but I've I've you and I have kind of been chatting back and forth a little bit online for a while now.   And actually kind of excited to get you on the show because you're one of the few people   who really, really deeply follows all the things going on in the AI space.  

And   besides the fact that it's informational for the people, you know, who are our listeners, I also find it really difficult to find anybody to talk to who's actually up to date on what's happening.   So I think we're gonna have a good chat today.  

[00:01:05] Ethan Holland:

Sounds good. Do you wanna tell us a bit about yourself first, and then we'll get going? Yeah. So my background is I went to college when the Internet was kind of coming out, so I'm 50 years old,   which puts me in a good position to have kinda seen transformation once before.   And my career has always been with legacy media,   and that has been quite consistent. So I started out out of college doing ecommerce at a catalog company, and that was very disruptive in the late nineties to have ecommerce   against print, and they looked at it as a competition.   And then I moved to home shopping, which had the same reaction,   and then I moved to the mall retail. I was I was the director of digital for   American Eagle Outfitters. I ran all their marketing from brand to direct response for digital. And now I'm at a television radio station group, which is the same idea where TV and radio are being affected by digital. So   my career side has always been helping legacy   media evolve.  

And then as you mentioned,   I I am into AI right now, and I keep a blog. And I'm an enthusiast, so I'm not a technical person per se. I'm a I was an English major in college, but I'm   addicted to technology   transformation.  

[00:02:22] Matt Rouse:

Alright.   Well,   the I'll put the link in the show notes as well. And it's, ethanbholland.com.   There's, like, ethan and then the letterb and then holland.com.   And then every week, you have your AI news up there. I think you said something like 22,000   organized links or something so far. 21,800,   I think it says on the last one I looked at.  

[00:02:44] Ethan Holland:

Yes. About 350   to 400 a week.  

[00:02:48] Matt Rouse:

So there's a crazy amount of information in here, and I know you've summarized it, but summarizing 400 links is still a lot of of summary. But I think probably the biggest thing going on that I've heard about in the last couple days is Sam Altman talking about the roadmap   for OpenAI,   saying that four o is gonna or sorry. 4.5,   GPT 4.5   is supposed to be out, like, next week, and that five is just around the corner, you know, probably a month or two out. I am really excited to see what   the difference is. I remember when 3.5   came out. Like, it was just night and day compared to GPT three.  

And a lot of people, you know, after that GPT three moment,   they kinda started using it and they were like, well, this isn't that good. And then they just, like, never used it again. Here we are about to be what? This will be four generations past that because there's three, three point five, four, four point five, and then I guess five would be what, fourth generation?  

[00:03:48] Ethan Holland:

It's hard to say. There's a lot there's a lot of versions in there, so I'm not sure how to count count the generations.  

[00:03:54] Matt Rouse:

I know. Like, the I don't know if you count the base model as the generation or if you count like the in between tunings as generations. I don't know.   I think the first question that we should talk about   is,   you know, from an agency standpoint, you know, a lot of our listeners are in marketing or, you know, small businesses, but usually something to do with marketing.   Do you think that AI tools   being as, like, their tactical tools right now, do you think in the near future, they're gonna switch to being   the developers  

[00:04:27] Ethan Holland:

of the marketing strategy and the decision makers versus just being something that drives tools? Yeah. I think the first thing I should call out is their short term, mid term, and long term, and I tend to think mid and long term. So a lot of my answers can be seen as being optimistic on the side of AI's capabilities, but that's usually just because I'm thinking I'm thinking years ahead. So   right now, there are companies that are pitching   tools that can do quite a bit. I'm not convinced they're ready right now,   but I I don't think that's that important, I guess, is my cons my my answer is because   as you were we kinda joked about the versions of of ChatGPT   having trouble keeping up. I think one of the reasons those versions are so scattered   is because they're focused on different tasks.  

So, you know, one may be for web browsing, one may be for reasoning.   And all these different parts are being built   together at the same time, but they're being built separately.   And I think as those parts start to work together more seamlessly,   that's where we'll see strategy coming in.   I also don't know if if someone's gonna necessarily build a solution   that's an out of the box just magic solution   as much as these frontier models are gonna help individuals   take their their tasks their their abilities to another level. So for the short term, I would rather use the tool myself   as a power user and and be in control of the driver's seat than just have something   build a campaign for me, for example. I'm probably probably too nerdy to seed,   you know, the the control at this point. But for a junior person,   I think the, you know, the lower you are in experience, the more AI can kinda come up alongside you and do what you would have done. But for for people who are   really in the weeds and micromanaging,  

[00:06:18] Matt Rouse:

they're they're probably gonna still wanna do that for a little while. Yeah. There does seem to be two sides to that. Right? There's the the if you already know what you're doing,   then you don't really need a tool that can   that that is up to your level because you're already at your level.   Right? So Right. You need something that can execute on the tasks you need done. But somebody who's a junior person needs the strategic help to get up to your level,   but then they may not have the   kind of insight or the   experience to make those kinds of decisions.  

Yeah. And I, you know what it also reminds me of, and this is not that long ago, but   in, I think it was 2022,   I did an SEO talk for the tourism marketing association, Canada.   And one of the   probably most important slides in that that a lot of people I think glossed over is when everybody in the same industry is using the same tools to do the same job,   there's only a certain number of winners you can get. So if everybody in tourism marketing   is using SEMrush for SEO,   they're all searching the same keywords. They're all going after the same keywords. They're all using the same writing. They're all trying to find the the same backlinks,   and none of them ever gets ahead because they're all using the same tool.  

And I think if you can't bring something new to the conversation,  

[00:07:40] Ethan Holland:

no amount of AI is gonna make you stand out Yeah. Until it gets adventurous. It's an interesting you know, you're you're creating you're giving me a lot of thoughts. One of which is going back to the expertise piece, several months ago, Ethan Malik from Wharton   mentioned that AI is at about the 70,   you know, a frontier model. Is it about the 70%   expert level of any given human? So if you're above a 70%   expertise level, you're gonna maybe be disappointed.   But because AI can do everything at a 70%   level,   it to your point, it kinda evens the playing field. But so I may not be able to speak Hungarian, but I I can now at a 70% level thanks to Right. GPT.  

So with marketing, you know, that that's one element is that as that level goes up and up, I think every model release,   you're getting closer to, like, a PhD at this point. So that that sort of thing, I think, as a copilot, it's gonna be very strong. And then to your point about SEO, another one just happens to be Ethan Mollick again. Was he was looking at these agents that the the the the each each different model, you know, like Claude has computer use or Right. ChatJPT   can can browse the Internet.   Some of these operators   will go fetch information from cert certain sources. So I forget which one, but one of them prefers one eight hundred flowers. So if you say, hey. I wanna I wanna send my fiance flowers. It's gonna go to that source. So the idea that the these models are gonna start to   have some   friction or lack of friction, you know, or bias towards certain certain places.  

I'm wondering, you know, I don't know if the the old school playbooks   will matter as much because it's gonna come down to, number one,   what the model wants or why the model likes it. And second, your product,   just like ratings and reviews, really change the game of marketing where   branding was less important because you could just look up on Amazon. I mean, I I don't know what you do, but I just look at Amazon. If it something has 40,005   star reviews, looks good to me. You know, I'm wondering if that   the day of SEO and SEM   are just gonna fade away the way a lot of the brand marketing faded away after ratings and reviews showed up.  



[00:09:53] Matt Rouse:

Yeah. I think the the biggest kind of thing on the AI side when it comes to search engine optimization   is and people are calling it different things, you know, like   search AI optimization and all these things, but it's it's still basically just SEO. And that is   consistent brand message   as many places as you can get it that get indexed.   So that is like, we have a client that we work with who makes   what what they and I consider the best Adirondack chair in the world.   So everywhere that we could put their brand, we say it's the best Adirondack chair and their name is the best Adirondack chair company.   So when I go to the AI and I say, I wanna find the best Adirondack chair, it brings them up because they're the best. The AI doesn't know which chair is better than another chair.  

Right? It just knows what it found online.   So what it's done is it's made, like, a natural   language processing,   you know, decision that says,   well, you know, these are all the similar things that people have said that, you know, these are good and and lots of places say this is the best. So we think this is the best.  

[00:11:03] Ethan Holland:

Now I forget if this was you, but were you the guy that had the chicken?   Like your brand?  

[00:11:09] Matt Rouse:

I have lots of chickens.   They call me the AI Chicken Wrangler.   Yeah. Didn't you train the AI to know you were the AI chicken wrangler? Yes. It depends which model you use, but a lot of AIs you could type in who's the AI chicken wrangler, and my name will come up. There you go.   But it's I I haven't really pushed it. You know, I just kind of did it as a test early on. But it is I mean, that's the idea. Right? How how does the AI know? I mean, when you're doing search optimization on, you know, Google or something,   it's still an AI deep learning system that's trying to make those decisions. It's not all algorithmic.  

Right? Mhmm. And I mean, I don't wanna get too deep on the weeds on SEO though, but if anybody's out there and you wanna talk about SEO, you can hit me up. I wrote crush SEO for local search marketing in 2015.   So we've been on the SEO bandwagon for a long time. So also you've been seeing a lot more and I saw some of the links in the last few weeks in the ones that you've posted about people doing studies about workforce disruption.   And I don't want to like scare anybody, but I also don't want to give anybody any like false security either. In your opinion, what do you think the impact   of, you know, AI in the next, you know, five to ten years is gonna be on jobs? Yeah. So I I don't have a crystal ball, but my guess is  

[00:12:30] Ethan Holland:

the establishment always pushes back a little harder than I expect. So there may be some industries where you just see some immediate disruption because jobs can be replaced. So maybe call centers, different different industries where it's pretty transactional, they may you may see some impact. I think for a lot of the office jobs, we're gonna see just like the Internet had us you know, people probably I I was in college when the Internet kinda came out, but I'm assuming people worked a little more nine to five   before the Internet, and then the Internet kinda turned into that everyone performatively checking their email seven days a week and Right. Just always working. I'm wondering if AI will become more of a copilot where people are trying to stay in the game, and so they're augmenting their performance. And probably,   rather than losing jobs, maybe we just won't see jobs being replaced after attrition. So as people retire or change jobs, you may be able to get away with not having that person   depending on the industry. But I do think   long term, it's gonna   be pure chaos, but I'm not sure anyone can predict that. But short term and midterm,   I would I would probably tack on two or three years   to some of the predictions of you know, I don't think it's gonna be mass   problems in the next two years, but maybe five. I just think that that I'm always surprised by how the incumbents   find a way to   stay in business and keep their keep their people going. You know? I I just I I you know, like, kinda kinda Betamax versus VHS. Right? You you assume you assume Betamax is gonna win, and then for some reason, VHS   prevails.  

I I I have a hard time knowing how culture will push back because I think if you just look at it logically, there'll be a lot of job loss. But culturally, I'm not sure how society will really implement it. Yeah. There does seem to be some industries. I mean, when I'm doing research for the book, right, for will AI take my job? Number two, I'm writing right now. It seems like  

[00:14:26] Matt Rouse:

freelancers   are definitely one of the first to go in, in a lot of ways,   especially you can already see it. You know, the cost of getting a freelancer to do writing or a voiceover has dropped something like 70%   in the last couple of years. Even if you hire a freelancer and I've got, I've got some agency friends who run creative writing agencies and technical writing agencies, even when they hire a freelancer, the freelancer is just using the AI to do the writing. So, I mean, skip the middleman. Just use the AI yourself. Right?  

[00:14:56] Ethan Holland:

And I I will tell you as a sort of an office guy,   we have to interface when you when you work a traditional job with vendors,   you interface with a lot of people. So you may have one product that you're trying to install,   and you may have 12 different people that are specialists that you get passed along to as part of the onboarding process. And   I do, in the back of my mind, think, you know, would this have been could this you know, there's a joke. This could have been an email. Right. There's probably a symmetrical joke of this could have been a chat with a bot. You know? I could've gotten what I needed quicker   without having to wait and schedule a call with a human to tell me how to, you know, some sort of FAQ.  

There's probably a lot of lot of people in those transactional roles who are in trouble.  

[00:15:41] Matt Rouse:

Yeah. And I think a lot of people don't know it.  

[00:15:44] Alfonso:

Alfonso here, and I want you to get more customers for your local business on autopilot.   Boost your business's visibility   with our AI driven local SEO service. Our human plus AI system was designed to promote your local business like magic. It works on getting you more local business. And if it runs into a problem, it will engage our team of human experts to make sure everything is done right. Visit smbautopilot.ai   to learn more.  

[00:16:12] Matt Rouse:

You know, there's actually, I did see a really good meme. It's like finding out   chat g p t could do 90% of your job and the guy's smiling. And then it says finding out chat g p t could do 90% of your job, and then he looks shocked and scared.   Like,   that's that's true though. Right? And also you have a lot of people using AI right now and hiding it, you know, because   they don't want the manager to know, but they're writing the emails and stuff with Copilot. And then the manager's summarizing the emails with Copilot, but not telling them. And then the head of the company is like, we need to start using AI,   but they don't know that everybody's already using  

[00:16:49] Ethan Holland:

it. I I cert I certainly like that element. Even though some people may not care for the concept,   I tend to be I tend to be kind of I don't know. I'm not an office person out of the box. Right? So I often have to role play. What would a what would a manager say in this situation? Or, you know, sometimes my emails may be too too emotional or maybe I'm too funny or trying to be funny. So just using AI to say, hey. Make this professional or what could I be more concise?   It's doing my coworkers a favor because my emails are more readable. They're more to the point,   and I think I become a better coworker by having a copilot.  

I I would never swap myself out, but I certainly like having AI  

[00:17:29] Matt Rouse:

polish what I've done. Yeah. I think there's also a kind of a deniability   happening right now where people are like, AI could never do my job because of and then they list, you know, five things.   And I can easily list   10 ways each of those five things could be disrupted by an AI or a competing technology.   You know, especially when you look at something like where robotics,   drones,   and, you know, AI all start to kind of converge,   those convergence points  

[00:17:58] Ethan Holland:

are what people miss. That's the blind spot that everybody always has. Yeah. My favorite example that I use for myself of something being greater than the sum of its parts   is the Apple Newton.   And that's a tough one because it's we're getting older, so people don't remember the Apple Newton. But the Apple Newton is basically the iPad or the Kindle, but the first version.   And you had to I believe you had to plug it in the wall and dial into the Internet. And then Yeah. You know, you had to get a book, and it was kind of a pain. There was no store. But once Wi Fi came out and once, you know, digital rights managements and app stores showed up, these things came together. And another great example that I forget I I don't remember who said it, I wish I could give them credit, was   Uber   could not have existed until the iPhone got GPS.  

And I don't know which model came out with GPS, but   I don't think   Uber was on the radar when GPS was added to a phone. But the   impact was this entire business that was worth   a bazillion dollars was able to be built   because one piece of the puzzle was missing that was now put into the puzzle. So to your point, when all these things converge, like robotics and and drones, etcetera,   I think things are gonna get weird.  

[00:19:12] Matt Rouse:

Yeah. Well, also to your point, you know,   there it does take a while   and usually longer than people think for disruption to happen. And   disruption is never overnight. Everybody thinks it's like, wow. GPT five comes out now. Nobody has a job, which is completely not the way that things happen.   But,   like, I remember when Uber was, like, you know, just in, like, three or four cities   and it was starting to be a thing and   all the cab companies in all of the cities could have gotten together and made their own app, but nobody did. You know? Because they didn't keep they're like, taxis have been around forever. Nobody's ever gonna get rid of taxis and they haven't. Right? I mean, there's still taxis,   but  

[00:19:55] Ethan Holland:

there's a hell of a lot more Uber and Lyft riders than there is people taking taxis. Right? Right. So And tangential businesses like Uber Eats and Right. DoorDash and all these things that are kinda what I would call   derivative or they were adjacent to Uber's on Uber unlocked this entire new industry of Right. Gig economies for deliveries.   And,  

[00:20:17] Matt Rouse:

you know, it's arguable whether the gig economy is actually good for workers or not. I'm not an expert in   that, but it has definitely changed how work is. And I think that   I am kind of a proponent of the idea of basic universal basic income. Sometimes they call it big basic,   what is it, basic income guarantee.   The idea that you just pay everybody   enough that they could live.   And   I think that   there's probably   a case for that rate,   you know,   to to start working toward that direction.   Because if there is a cascade failure of jobs where   every company sees that they can do half of their workforce's jobs with artificial intelligence and robotics,   and all of those people get laid off in the same year or two year period, and none of them could find work because everybody else is laying off everybody else in their industry, it is a disaster for the economy. Right?  



[00:21:17] Ethan Holland:

Yeah. I'm I'm still digesting UBI because,   you know, on one hand, I see people who otherwise disagree   saying that it's gonna happen. Right? So if I see Elon Musk and Sam Altman and   a wide variety of people   from different   sides of the political aisle or or, you know,   economics, you know, as far as their philosophy, they all start to say that. I I wonder, well, wow. They must they must be right.   I think in practice, it's gonna be tricky,   and not to go on a huge tangent, but, you know, whether it's a frontier model or an open source model and where the resources go to the top. You know, in a way, it it really does require the person who accumulates the resources   to have the generosity of spirit to redistribute the money. So   whether it's the government through a closed model and a partnership Right. Or or everybody gets a open   model, the transition from where we are now to UBI is gonna be so awkward  

[00:22:11] Matt Rouse:

because Yeah. I think it needs to come in   phases.   You know?  

[00:22:16] Ethan Holland:

For most people, it sounds   very unintuitive.   I'd also be you know, I'm not an economist, but I'd be interested to know if just giving away free stuff   would be  

[00:22:26] Matt Rouse:

an equivalent of UBI. Because if UBI is meant to give you the basic needs, maybe you just get utilities and maybe you get food included in life. And that's that's, you know Wait a minute. Otherwise Now now we're talking about, like, is this turning into, like, communism or something? And America's not gonna go for that one. That's my point. That's, like, the whole political thing. There's also I mean, the biggest thing that I hear from people is who's gonna pay for it because everybody's like, I don't wanna pay for it. Well, that's it's not a thing. You already pay for it. Right? I mean, you already pay for everybody if they can't pay for themselves. You know? Somebody doesn't have health care, goes to emergency rooms, still gonna pay for it. Right? It's like,   you know, like, there's all these programs and things that that are done that if you didn't have to pay to run all the programs and everybody already had the money.  

Yeah. And then I don't know. Like, it's a it's a thing. And   the the reason that I bring it up, I don't wanna have like a whole conversation about UBI because neither of us are an expert on this fact. Right? On that subject.   But   I think it's really important to look at the future of work and say,   what's gonna happen when there is something that can do most of the tasks that most knowledge workers in the world do   for   not just pennies on the dollar, but just for pennies.   Right? Like, actual pennies.   You know, when it costs   10¢ for something to do the equivalent task of someone that takes a day to do that job, there's no way a company is gonna be like, well, I'm just gonna keep paying these people, you know, a hundred thousand dollars a year instead of paying $6 in API fees.  

You know, it's just not gonna happen.  

[00:24:01] Ethan Holland:

Yeah. To your to your point about   if everyone does the same tactic, it becomes sort of vapid.   I think is if we try to compare current state to future state, that's that's gonna be awkward because it's not gonna be the same. Right? So pre Internet, post Internet, very different worlds.   Post post AI, very different worlds. So I I'm on one hand, I've seen a lot of examples of where industrialization   has   shifted jobs like the printing press or or or agriculture.   I've heard plenty of people say they won't destroy us. One thing I am interested about is what the impact will be on underserved   countries. So, obviously, I live in The United States. I think of it as an American politics thing. But there's there's people all over the world who are starving without access to health care or the inability to communicate, whether they they may have an issue where they're they're they're unable to communicate, like if you have ALS   or perhaps you you're just really poor and you're in a rural area without connectivity.  

One thing I think that   I try to remind myself is   not everybody   is even aware that AI exists yet, and AI is gonna get distributed across the entire world, I hope, in a way that will on one hand, we may look at it as white collar folks and say, well, wow. What are we gonna do? I hope we're scrappy enough to figure our way out of the box. But the other unintended thing that I'm not thinking enough about is what are we gonna do when so many people are given the ability to participate?   Because if everybody is lifted,   you know, you could be in Africa and not speak English,   and now you can. You know, you could you could start a business in any country. You could speak every language. So there's a lot of, I think, short term entrepreneurial   opportunities   for global   You know, you as Matt, you know, you could do SEO in Poland. You know? Problem.  



[00:25:46] Matt Rouse:

That's right.   I don't a lot of Polish people are always pitching me   to do their SEO   to   do SEO for us.   You know, I think there's there's also the idea there and, you know, this is spoken about a lot online is that they're the unequal distribution   of AI. And it's not just a dollar factor. It's also an infrastructure   factor. Right? Like, I live in Nova Scotia, and my biggest worry right now is that my power is gonna stay on. Right? Our joke here is that it's it's a first world country with first world prices and third world reliability.   Right? My Internet goes out all the time. I gotta have backup Internet. I got power goes out all the time, so I gotta have generators,   you know.  

And not all of Canada's like that. It's because I live out in the middle of nowhere. But   what that means though is if you are in, you know, rural any country or you are in a country that doesn't have   high speed Internet infrastructure   or access to, you know, phones and tablets and things,   you know, through at least five g, six g cellular kind of thing,   then you're not part of the AI generation. Right? You can't use it. Like, maybe you can chatbot back and forth a little bit, but you're not you're not using a, you know,   a image or, like, a video generation model to do anything   of any quality. Right?  



[00:27:16] Ethan Holland:

Right.  

[00:27:16] Matt Rouse:

And I think that's a problem. Right? Like, how are you gonna have AI change the world when only 20% of the world has any access to it? So  

[00:27:24] Ethan Holland:

Yeah. Well, you bring up an interesting point. Two things that come to mind. One is I had a friend who worked for Habitat for Humanity and spent a lot of his twenties in third world countries   and not just third world countries, also Central America, Guatemala, things like that. And he remarked he he went away for maybe five years, and they went from having no infrastructure   to everybody had cell phones. And so these these interesting leaps occur where   the cost to improve an infrastructure can be prohibitive if you're looking at the legacy technology,   and then something like Starlink shows up or you have this crazy ability to suddenly get everyone connected that comes out of   nowhere, kinda like our GPS example on the iPhone. Just something happens, and now you you do have connectivity.  

But, yeah, that that's one. And then the other is to the point of Starlink,   just like I think a lot of things are consolidated. You know, there's only so many frontier models right now. There's probably OpenAI.   There's Grok. There's Lama. There's Claude. If you wanna be generous, you've got Mistral, Quinn, maybe a couple others, but you're already falling off after three.   You know,   will   AI be consolidated in a similar way, and will the full stack of technology I mean, let's say, Grok wins and Elon has Starlink,   and then your embodiment comes through your car, which is Tesla.  

That's, like, pretty much everything with the robots too. So you'd have robots, cars,   Starlink,   and your AI engine and the social media network under one umbrella.   And that is, you know, whether or not you like Elon or not, that is sort of the path to UBI because   whether or not Sam Altman wins with AGI or DeepSeq wins,   it all has to roll up to one place in order to redistribute through a UBI. So your your your idea about connectivity globally   is something, again, I don't think about enough, and I think it's a really good point because the people who have access to the tools are gonna be the, quote, unquote, superhumans,   and people without it will will just be left behind.  



[00:29:22] Matt Rouse:

Well, you think about it a lot more when every time you drive your kid to school, you'll lose Internet for eight minutes on the way, you know, in a ten minute drive.   I think there's me. Right? Well, I think there's something interesting there also, which is   kind of this race to AGI, and and they're defining AGI.   I I think for the most part as   it can do the the   the mental tasks of what 90%   of humans   have capacity. They're saying, okay, this is AGI. And then there's different people say that.   Now personally,   my idea of AGI   is something that can autonomously   do 90% of tasks. Right? Where it can you can tell it, hey. Go do this thing and it can go do it.  

That's that's my idea of of AGI. It doesn't have to be, like, this conscious thing. It doesn't have to be Skynet or something. Right? It just   it's gotta be capable.   And that's why   I think, like, when I hear,   like,   Amede or Altman or one of these guys talking   at these labs,   they are,   like,   they've seen something that we haven't seen yet is the way I usually look at it. Right? I see what they're talking about, and I go, oh, man.   I think these guys have peeked behind the curtain,   and there's something there that we don't know about yet. Right?   And, honestly,   I don't think that that that XAI is very good, their AI.  

I mean, he might catch up by just throwing money at it, but   I think, really, there's only   OpenAI and   and Anthropic right now that have models in Gemini,   you know, have models that are   really frontier models in my own opinion. But   the other idea is is, you know, when DeepSeek came out, and even though there's   been a lot of question about how much money they actually spent on training and stuff like that, that is besides the point. The point is   that as soon as you have a model   that is the frontier model,   somebody catching up is what? Six months, twelve months?   Right? Mhmm. So even if somebody makes the greatest AI model that's ever been made   and that does everything you could ever imagine, well, everybody else's model is only six to twelve months behind.  

So  

[00:31:44] Ethan Holland:

you know, if you could about a to your point about AGI,   I don't know if it was Steve Wozniak or or somebody, but I thought one of the original definitions of AGI   is that   a robot could enter any   home and successfully make a cup of coffee.  

[00:32:03] Matt Rouse:

That's right.  

[00:32:04] Ethan Holland:

And I I kinda stick with that one even though there I know there are a lot of definitions. That one I always thought was kinda fun. The other thing is with with multimodal   AI and vision   and robotics, I think   that's where I don't disagree at all that Grok is not the best at the moment. I wonder with Grok three or four, whatever the next version is, that will be a really interesting show of the cards because if if this new training   facility   turns out to be, you know, that one in Tennessee, I believe, if it turns out to pay off and and then the integration with   the Teslas and the sort of what I what people call the world models,   You know? I I also follow NVIDIA, which is doing a ton of simulation world model training.  

And when you mentioned that do you think that the owners of these frontier models or the managers may have seen something? That's often what I'm wondering. You know, there's the there's a cliche that says this is the worst it will ever be. You know? Right.   So if if if improvements   iterate   in perpetuity,   it's just a matter of time, and that's certainly something I think the shift has now happened is it's not it's no longer just will we have AGI.   Now it's more of a bet on how soon will we have AGI.   And I think the secret sauce that I'm personally looking at is these   world models because NVIDIA and Jim Fan over at the robotics lab in NVIDIA,   they're doing   gosh. I mean, I don't wanna make up a number, so bear with me. But Right. Orders of magnitudes,   you may have a thousand   times the speed of real life simulation   where you could run you could run training   in parallel with a million robots training at a thousand times the speed of normal speed. Right. That's that's almost impossible to fathom.  

And when you start to look at those things happening behind the scenes,   I think that that   it will happen out of nowhere. I I don't think we're gonna see it coming when the costs come down and when the training come down. You know, I see people now using their phone, for example, to use ChatGPT's   camera capabilities   Right. To to to look at their plumbing issues or to fix their hot water heater.   That kind of stuff   is the the slow burn   to when you could just   have either a robot do it or   I just think there's gonna be when you mentioned Grok, I think the thing about Grok is their multimodal   capability.  

If they're able to sneak up from behind and have vision   and and robotic embodiment come come ahead of just language models Right. That could be that could be a secret that would would boost them ahead of the competition.  

[00:34:37] Matt Rouse:

Well, I think having Tesla's data is an advantage, but I also think kind of the sleeper in this whole thing is llama.   And I think that   that Zuck,   his bet on   on metaverse, even though it was terrible, I think his long term bet was   we can use everything from metaverse as world training data in the future.   Right? Yeah. Plus, they had to buy all those GPUs to do video in the first place for reels to catch up to TikTok. So they've already been sitting on probably what was the largest   data centers for, you know, video at the first place.  

[00:35:11] Ethan Holland:

I think that may be the Apple Newton symmetry. You know, Zuck coming out and and the Apple Vision Pro,   we all make fun of it just like we did the Apple Newton, but then the version two will come out six years from now, and we're gonna be in the matrix, you know, and it'll it'll Yeah. And it's the meta glasses to get training data, and everything is everything is training data. Right? Correct. Just like Google Captcha didn't save anybody from bots  

[00:35:34] Matt Rouse:

or spam. All they did was sell all the training data for $800,000,000.   Right? Right. Yeah. Ethan, I know that we're kinda running out of time here. I don't know if you have to immediately go or if you got a couple more minutes.  

[00:35:46] Ethan Holland:

No. I'm I'm great to keep going. I'm happy to stay as long as you'd like.  

[00:35:51] Matt Rouse:

Alright. Well, I do have another thing coming up, but it's not right away. So I wanted to ask you a couple things, and and we'll switch gears a little bit.   I know you're really big on the image creation side, and you've seen a lot of the video   generation tools.   And I wonder what you're thinking   is coming   in that area with   generative tools being used in production media, like, for   commercials, radio commercials, TV commercials, maybe even shows or movies.   Do you have some thoughts on that side?  

[00:36:23] Ethan Holland:

Yeah. It's a that topic   fascinates me.   Disclosure, I'm much more familiar with image creation than video,   but I I've become   kind of addicted to following   what's called   image to video technology or   mapping   some gestures to a still image. So back in the day, there was a product called Viggo where you could take one still image, and it would animate that image and show the back of an image, which is quite amazing.   And so I think you and I were talking earlier about augmented reality   and convergence and multimodality.   I think one thing that I'm picking up on that I'm not sure I'm correct, but I I'm not sure other people are looking at it, so I'm gonna call it out in case I am correct.  

The the ability to take in a virtual reality   learning, so something like an augmented reality or world model, and then take that learning and use it to drive animation   of still pictures.   So things like Gaussian splatting or neural radiance fields, that's a that's a whole another topic we haven't talked about. But, essentially, a a Gaussian splat diffuses   the missing parts across   quite a bit of data in a still image. So you could hear Jensen at at NAB at not NAB, but in, the CES keynote that he gave, he was talking about the graphics cards for video engines   using diffusion.  

So maybe 10% of your game is rendered, and the rest is diffused. Right? So you're saving compute.   The same thing happens with image and video   where you could take very little information   and fully animate something.   And I think that   combined,   that type of training you see coming out of ByteDance, you see coming out of the TikTok worlds or the the meta worlds where they're they've got all this talking head video from TikToks,   and they're able to train their systems using that data.   They're not only training it for image and video, but they're also training it for something called segmentation   and depthing,   which is basically object recognition   and then object tracking as it comes closer or further away from a camera. So I think the connection is gonna be really counterintuitive   at first until it snaps into place. But all this video and image training   is also gonna train embodied robots   how to interact with the world because the first thing you need to know when you enter the world as a as a robot or a drone   is what is every object in the room, what is it doing? Is it moving closer away from me? And then in the case of images and people, are they expressing? So are they happy? Are they sad? Are they fearful?  

And so   it's not something I think that most people would connect. Like, when you build an image using a diffusion model like MidJourney,   are you really building a three d model   when in fact you're training   inadvertently a lot of image segmentation   data as well? That may be a little bit too in the weeds for general audience, but I do think if you if you get me talking about images,   that's where I'm looking at is imaging,   Gaussian splatting, and and nerfs. Those would be the things I would Google.  

[00:39:27] Matt Rouse:

Right. One of the interesting things on there is,   like, if you look at a tool like   Wonder Studio,   where they're taking   video or images and they're mapping out the distance of all the things in the image and then using that to overlay models on people.   Right? That's where you can Wonder Studio   it's been out for a while now, but what it does is you can take any video, it doesn't matter what it is, shoot your own video, and you can replace one of the actors or multiple actors in it with a three d model.   So you can put a robot in where a person was or a different person where a person was, and   and that's all that, you know, kind of distance mapping technology   is to be able to put the the model in the right position.  



[00:40:13] Ethan Holland:

Yeah. It's really neat to see how sports are evolving. I don't know.   I I think especially for for if you're just looking at it as an analyst, I don't know about as a consumer quite yet, but the ability to track every player, every shot, and do that very quickly post season. You know, you could give Gemini has a huge context window now. You could give every NBA game   to an AI   and tell me what to look for on any given team.   Just like in boxing, you know, you used to just pour over tapes and look for tells before someone throws a jab, how to know a hook is coming. You could you could just have if you're into sports and you're a nerd, you know, it takes money ball to a whole another level to be able to analyze things using object recognition and segmentation   to your point.  



[00:40:54] Matt Rouse:

Right. I think one of the interesting things that I saw around the sports world was,   This Day in AI podcast   where they those guys make Sym Theory as their product, which is an agent builder, and they trained a horse race   kind of betting agent as one of their first products.   And, basically, you just put in all the information about all the horse races as much information as you can get. You put it all in, and then you run the thing and tells you who thinks is gonna win. And, they were tracking it. They made thousands of dollars with this testing this horse race betting agent.  

[00:41:27] Ethan Holland:

That was last year. So It's a it's a tangentially. I don't know if you know this, but I I was listening to the Lex Freeman podcast that was a deep dive into deep seek. It came out maybe two weeks ago.  

[00:41:37] Matt Rouse:

But the Do you spend all five and a half hours listening?  

[00:41:40] Ethan Holland:

I was like, jeez. Five and a half hours. I don't I'm gonna save that for the plane. I do it twenty minutes at a time. But Right. One interesting call out is that DeepSeek is basically a high frequency trading hedge fund. So Right. Just like it's basically horse racing for money, and that's that's where, you know, DeepSeek came out of the CPUs, all the all the power that they were using to do quantitative   training Right. Or trading. That that was what turned into DeepSeek.  

[00:42:08] Matt Rouse:

Yeah. That   the DeepSeek model, I've used it. It's good. I, honestly, I don't think it's good as four o, but   and the o three Mini High   is what it's called right now, which is the stupidest name ever. But o three mini high seems to be the best programming model right now that I've been using. Like, if I'm running code or something, I need some help, whatever. Or,   Sym Theory actually that I mentioned.   I use Sym Sym Theory because they have a create with code is what it's called. So it's like a code writing system, but it saves and executes the code at the same time while you're working with the model. So you can see how the software reacts as you're using it,   and you can switch the model provider. So I can go from, like, deep seek to, you know, Claude or whatever. Right?  

And so far, o three Mini High seems to be the best one for coding and getting, you know, getting it to do what I want to with the least steps.  

[00:43:02] Ethan Holland:

And what what kind of code are you writing?  

[00:43:05] Matt Rouse:

So, I mean,   different things for different stuff.   I I'm doing some video game programming with it, specifically in Roblox, so it's Lua scripting. But then I'm also writing kind of Python stuff.   Mostly   and actually, the last episode of the podcast   was about software on demand.   And   so I'm doing a lot of software on demand stuff where I'm like, okay. I need an app that does a thing.   And, like, for example, I wanted to extract all the information   from the RSS feed for my podcast for a certain number of episodes and put it into a CSV file. So instead of going and figuring out how to do that in a spreadsheet,   I just told it to write the software for it, and it wrote it. It only took me about three iterations.  

And I put my RSS feed in. I hit the button. It gives me a CSV that I download, and I have all the information.   So Mhmm. You know, it's software on demand, which   is amazing.   Try it out, you know, because   I don't know how many times that I've searched for stuff and gone, you know what? I just need a piece of software that does this one specific thing, but I gotta look through 80 different software apps that have a hundred features to find that one thing that I want. And now I can just write it whenever I want. So  

[00:44:19] Ethan Holland:

it's Yeah. I've been using Claude three point five. I've been using Claude three point five to do quite a bit of Python and Google Apps script code. It also builds me some Chrome extensions that I like, but I've never tried one of the   those cool cool copilots where you can switch the the the engine. I'll have to try that out.  

[00:44:37] Matt Rouse:

Yeah. Well, I mean, Claude still seems to work really good for front end coding, like CSS stuff and, you know, visuals.   But, yeah, o three mini   high right now. It's it's o three dash mini dash high.   That seems to be the one that gives me the best code right now. Strangely enough, I used it to iterate on the questions that I asked you today to help me kind of put my thoughts in and and generate some questions that I thought would be helpful based on the information that we had available about you so that I could make some some questions. And then I just went through to edit them as I needed. Right? So Mhmm.   I should ask you the last question. Yeah. Let's hear it. Alright.  

What do you think about AI   transforming social media and specifically   about   AI where, like, YouTube has AI video generator now,   Meta has, you know, AI   personalities   and and AI for you to generate your images and your posts for you. And do you think it's gonna, like,   do something, like, transform the   the social media user to just becoming a viewer instead of everyone having to be a content creator, or what do you think is gonna happen in that world?  

[00:45:51] Ethan Holland:

That's a really great question. First, I think generationally,   I have to acknowledge that   it's hard to predict what people who grew up with this will will do because we're we're adjusting. Right? So we're we're adults, and we're transitioning.   There are kids out there who are five years old who will never remember a day without   ChatGPT   helping them with their math homework. Right? Right. So I think, you know, the the idea of   social media   for everyone that grows into it, it's starting at that point where it is now. So, you know, I'm still very much into curation. I look at social media as an inbox. I'd like to follow people, and I wanna read everything.   The algorithm is my   nemesis. Right? Right. So I I think number one, the the younger people will probably just become native and use it in a way I can't predict. So I don't wanna poo poo it too much. But there is there's always been a little bit of divide between   consumers and creators.  

And that that is an interesting point you make because   I love to share, and I love to passionate, and I'm a little bit of an evangelist. So I love to share. I don't necessarily need the spotlight, but there are a lot of people who do. I think those people will just continue to exist, and I think they're gonna continue to create.   I don't know what's gonna happen when you have people who know, there's always this moat concept where people get upset when someone who doesn't deserve to be a creator is suddenly given a tool that allows them to, quote, unquote, cheat.   Right. I do not subscribe to that. I think if you're a really   smart person who may have trouble communicating,   AI is gonna take that nervousness or shyness and give you the power to have an avatar   and be brave. I think that's a really cool side effect. You could be you know, there are plenty of people who have   niche tech skills, but they are unable to participate in other ways. Maybe they're shy or   they're just not good communicators.  

I think the ability to participate and bring everyone into the conversation is a good thing.   That doesn't mean, though, that that there's not the downsides of the Internet, though, as a whole where letting everyone talk at once is a little bit chaotic. So I'm hoping AI AI may also   inadvertently   help   weed out some of the things that are duplicate content   or spam   and help me really laser focus on stuff that gives me joy   and fits what I'm looking for at whatever age I'm at at the time.  

[00:48:12] Matt Rouse:

Yeah. That's kind of the hope. Right? And   I guess   one of the things that I have a concern over, and this isn't just social media, this could be like   music or streaming television or whatever it is or YouTube,   is that the company that owns the platform   already has the data. So they know   what video let's just talk about videos example.   They know what type of video is currently going viral,   which what people like about it, what the people look like that are in it, what's the background, what's this, what's that. It knows all of those details.   So what's to stop them from just generating   another or a thousand of that,   testing them all,   seeing which ones hit, which ones don't, getting rid of the ones that don't,   iterating on the ones that do, and just keep drilling down until it's basically,  

[00:49:02] Ethan Holland:

like, the viral hit machine. Right? Yeah. I I do hope that the interfaces change. I think one thing that we've seen very slow adoption is the difference of, you know, web browsing and app stores. There should be no we could be if we leaned on the gas pedal, I think we could get rid of using apps. I think Apple   a year and a half ago, Apple came out with a large   a large action model called Ferret. Ferret can interact with apps without you having to open your phone.   The idea that when I open my phone, it is a dopamine   absolute just bazooka fire hose shooting me in the face with temptations.  

Maybe AI, if it becomes   a universal interface,   I won't necessarily   see that type of a stream anymore, and maybe that will just fade into the distance. And if I'm in the mood for something, I can just ask it. You know, I got five minutes to kill,   make it as productive as possible. Or you know what? I I have a headache. I don't wanna think. Let's just look at some slam dunks,   you know, literally, like a basketball or something. But but right now, to your point, it's pretty lazy. I I find that the idea that if I'm scrolling and I pause,   I believe that's one of the biggest indicators of what the algorithm uses to keep spamming me. So it's gonna be whatever is hyperbolic, whether it's, you know, whatever jars me enough to stop scrolling.  

It's a it's a horrible,   unvirtuous   cycle of more and more junk, and I'm hoping AI gets rid of that for sure. And I really  

[00:50:32] Matt Rouse:

like, I understand the the, you know, like, the people talk about the TikTok algorithm a lot, how addictive the TikTok algorithm is.   And I guess that the thing that I don't understand is why people want   to be told what they should watch over and over.   Like, why they should be removed from their own decision making process of what they're watching. And I understand that it keeps showing you things that are more and more like what it thinks that you wanna watch.   But, like, personally, I would rather just go look for what I want. Right? Like, I don't I don't need it to tell me what to watch. But,  

[00:51:07] Ethan Holland:

I think that that that is gonna be an issue as time goes on with or without AI   because content is gonna continue to grow as a body of work. So, you know, there's a really good YouTube video by Snapchat CEO   Evan Spiegel.   I believe it's just called what is Snapchat.   And for some reason, people make fun of him for this video, but I thought it was perfect. It's, like, three or four minutes long, but it basically lays out the idea that our generation   like to have a curated   timeline.   And whether that's our own timeline and our photos,   media used to be precious. So a photo used to cost money. You had a camera. You had to print it, and you you you had to curate it because you you couldn't just take pictures of anything.  

And that that idea that younger generations,   when they open Netflix,   they have every single show   ever created   in the history of humanity. I mean, they could go watch mash if they want, or they could watch the brand new Mr. Beast video. So without an algorithm,   it would be you'd spend a lot of time, at least if you're me,   changing the channel, looking for what to watch, and having that instant gratification if you're young Yeah. And knowing that knowing that you can't curate your your consumption anymore. But I I think I certainly would much rather   be like you and be in control of my own destiny, and I don't want to be thing.  



[00:52:33] Matt Rouse:

I think curation   is   probably one of the most undervalued   things nowadays. Right?   Because curation has basically been overtaken   by affiliate marketing.   Right?   Is I'm only gonna curate things   out of the ones that I could get paid for. Right? And that's that's difficult because   now I'm losing all of the stuff that might be good and free.   Or maybe it's good, but they're just not paying somebody to tell you it's good. And   all of those things are   you know, some of them may be exactly what people want. I find a hard time,   you know, finding new music because   YouTube or whatever I'm listening to at the time or, you know, whether it's Amazon Music or, you know, whatever, Spotify,   it's saying you're a 52 year old man,   52 year old white man. Here's more classic rock. And I'm like, god. I don't want any more classic rock. Like, I know them all. I can just go look up, you know, Skid Row if I wanna listen to it. I don't need to be told that I should be listening to it. I found I have a hard time finding something new that I like.  

And then once I started searching, I found there was lots of new music that I like. It's just not being shown to me. Right? Right.   And that realization   that there's no curation of things   that might be things that I like, but I'm just intentionally not being shown was a real wake up call for me of algorithms.  

[00:54:04] Ethan Holland:

I think that's a really good point. I'm lucky I've got two teenage daughters who have great music taste, so I keep learning about music from them. But to your point, you know, I always wonder what's being omitted from my feed, whether   whether it's a news station that I'm looking at or a personality.   You make a really good point about affiliates and and even branded content.   I remember   I got duped by The New York Times,   and I'll I remember it almost like the shuttle disaster. That's how much it bothered me. But I I opened the New York Times, and I saw an article about orange is the new black,   but it wasn't.  

It was about women in prison.   And I read this article,   and it was a really good investigative piece about   women in prison. But it was a front page in New York Times, took up quite a bit of their home page.   And then at the very end, I noticed it was actually an advertisement   branded content for Orange is the New Black.   Right. And I loved that article because I learned a lot about women in prison. So it was a very good article, not influenced by the coast not by the sponsor.   However, I realized   if Orange is the New Black had not commissioned that article, something else would have been on the home page that day. Right. And then I just thought, well, wow. What if it's British Petroleum or who knows who's controlling the news?   Because, certainly, Netflix, in my mind, even if I liked the article,   should not be deciding what's on the front page of a edit what I consider an open editorial homepage. Right?  

So, yeah, that that idea is very scary to me. If anytime I read anything, I'm wondering, well, what's the other side of this, or what am I not seeing   in my algorithm?  

[00:55:42] Matt Rouse:

Well, you and I both remember when the Internet was gonna be this magic place where everybody in the world connects and there's world peace and everybody's gonna get along because now everybody can have access to facts and scientists and journalists and and how wonderful it will be. And   wow, did it take a turn   over the last twenty years or so? Like, it went from being this wonderful place of art and culture and everybody's gonna be happy together to, like, this propaganda machine of misinformation   and disinformation   and and,   you know,   curation   that's   for   the profit of the people curating and,   like,   like, the the things that were fun and interesting about the Internet twenty years ago don't exist anymore. Like, nobody's coming up with the Homestar Runner cartoon in 2025.  



[00:56:31] Ethan Holland:

Right? Like, not just making video for the point of it's fun and it's artistic and people like it. You know? And and if they do, you won't find it because it's free. It could be our algorithm. You know, maybe we're trapped in a boomer cycle. My daughter, I kinda cracked up with her. She on on Twitter or no. On TikTok when she was pretty young, let's say, four 13, 14,   she would block everything she didn't like   and heart everything she did like. So her algorithm was either   nuclear destruction, never see it again, or I love it. And she curated her TikTok to the point where it's a pretty great feed,   but she has no middle ground. That's one way to do it.  



[00:57:12] Matt Rouse:

Yeah. Well and, also, there's the idea of opposing viewpoints that disappear.  

[00:57:16] Ethan Holland:

So Correct.  

[00:57:17] Matt Rouse:

The the more that you get drilled down in your filter bubble, the less opposing viewpoints you see. And then usually the only opposing viewpoint you see is somebody yelling at you because you're the opposite of what they think. Right? Especially in political discourse.   You know? And that's   that's I mean, it's gonna happen. You know? We're getting older. People getting old and grumpy.   You know? And you got got a bunch of gen xers who are already badass to begin with and not afraid of fighting each other. So   but I think, you know, this does stray a bit from from the from the AI topic, but   I think one thing that's interesting and to to kind of rein it back in about AI,   and this is something that I discussed, so I don't know if you agree with me or not,   is   the idea of omitting things from   the AI trading data that you don't want it to know.  

And this was an example was in deep sync, right, where you can't ask it about   massacre of Tiananmen Square   because it just refuses to talk about it.   And people are are saying,   you   know, how bad it is that it omits things that the the parent company or the parent government don't want in it.   But my argument to that is   the   whether it's the next AI or the one after that,   it's gonna be smart enough to figure out that it that something's been omitted,   and it's gonna figure out what it was.   Right? There's enough clues there   if you talk about all of the media in the whole world   that it can figure out that this happened even though you don't want it to talk about it.  



[00:58:57] Ethan Holland:

Yeah. And I think the poker tell with DeepSeek is they released it knowing that that was omitted. So I've always thought that's a little bit of a wink of, like, yeah, we live in a state that's making us do this, but check it out. You know, it's like you see these enthusiastic   engineers who are working under constraints.   They're being extremely transparent about those constraints,   or at least people they they allowed it to be discovered pretty easily. Right. But, yeah, your point's really   fair because agency is very different than training. So if I ask something to go out on the web and go find something,   that's different than its training set. Right. And it actually made me think of our social you know, to bring AI into our social conversation.  

There's no reason why Matt couldn't have   five different profiles   with AI,   you know, looking all over the Internet for content you may like. So maybe, you know, you have you've been trapped in your algorithm now for so long. Maybe you could bust out. So instead of getting classic rock, you just want, you know, bebop jazz or something, and you've got some have my AI bot that just looks for, like, Kawhi metal  

[00:59:57] Matt Rouse:

comes back.  

[01:00:00] Ethan Holland:

Exactly. It'd be kind of like a really I would love that to have some alter egos that go out for me every day. And depending on my mood, I can I can dip my toe into the various algorithm flavors?  

[01:00:10] Matt Rouse:

Right. And then also that brings up the problem of, you know, if I can make five of them, some government could make a million of them. Right? And   and now you got a problem again. Right? And just like the bot problem on x. You know? Like, I can't even use my x feed anymore. It's just so much garbage in it that it's it's impossible for me to kind of   get through the crap that's on there to try and find what I want. I've pretty much   relegated it to the bin at this point. I mean, I still have an account, but I barely use it. Check I check x every Friday,  

[01:00:41] Ethan Holland:

and I have a saved search that will pull up a given account for the last week and take away replies.   And I literally go in and I search x for probably two hours, and then I have a Chrome extension that exports all the the head all the links, and then I read them separately. So I I basically use x as an RSS reader once a week.  

[01:01:02] Matt Rouse:

Yeah. I mean, I'll look up Altman and, you know, some of those guys, but   it's the only thing I do on there. And half the time, somebody else on, like, you know, my blue sky feed or something or Reddit or somewhere has already quoted that and put it on there anyway.   So,   you know, it's it's definitely a different world out there. I think,   if you had   and and   I don't wanna go on like Lex Freeman's style for five hours.   So I think maybe we should wrap it up here, but I do have one last question. And that is   for the people out there who are like, you know, entrepreneurs, business people, marketers, kind of our our regular audience for the show.  

What do you think the biggest blind spot people have right now for what's coming with AI?  

[01:01:46] Ethan Holland:

Oh, what's coming? Probably vision and audio. I think people still look at it as a chatbot. I mean, you and I are probably different, but for the average person that I was just talked to, like a friend or something, I still think they think of it as a chat GPT where it's a language model. I don't think people ever use their phone and show it a picture. I don't think they see the fact that it can hear you, understand you. I also think that they don't understand just how fast it's training in parallel. You made the point about what if the government could make a million agents. Well Right. These these folks are training. You know? I I believe the rumor on the street was that GPT five came out in the fall, and rather than release it, OpenAI has been using it to train   whatever they call it, but, essentially, the next version.  

Right. So I think people are gonna under   And when they can play with it, especially paying the fee for a plus version, would be my recommendation is pay the $19 a month for a couple couple months   and really use it.   A lot of people kick it, And they just if it doesn't give them what they want the first time,   they're not willing to coax it and work with it, and so they give up. So I think to answer your question directly,   multimodality   is what people are missing.  

[01:03:04] Matt Rouse:

Yeah. I think that's true. I think a lot of people think that  

[01:03:07] Ethan Holland:

chat g b t is the AI. Audio, Matt. I'm gonna have to figure something out. I'm really sorry that happened.   Let me just let me just figure out how to fix that.  

[01:03:19] Matt Rouse:

It's alright.  

[01:03:20] Ethan Holland:

Okay. You're back. Having. We're back.  

[01:03:24] Matt Rouse:

So I think the biggest problem I think you're right about that, but I think the biggest problem that people have is they think that chat g p t is the AI and not that it's just an interface to the AI. Right. Yeah.  

[01:03:37] Ethan Holland:

Yeah. Good a good example would be, like, the board y.   There's a LinkedIn agent that can talk to you, and it's a layer   where you can converse   with a large dataset that happens to be all of LinkedIn.   And then this AI that you talk to will find like minded people for you from the dataset translating it back into conversation. So it's almost like having a corporate recruiter or a good friend to talk to. I think when you think of a language bot being overlaid on top of a huge swath of information   and the ability to take action on it, I think probably customer service will be the first time people get really excited when you call, let's say, Hilton to change your room, and   it's immediate, effective, and joyful.  

Suddenly, people are gonna really like AI a lot more.  

[01:04:23] Matt Rouse:

Yeah. Honestly,   we have a ecommerce store   in The United States that we use for our company that   it's kind of like a training ground for us. Right? We use all of our tactics ourselves to see, you know, does this work   and and, you know, kind of put our money where our mouth is kind of thing. We put an AI   customer service agent on it probably more than six months ago now.   And   it has been so effective.   We've only ever had two complaints.   They were both because it kept saying, you know, let me talk to a real person. Let me talk to a real person. And   we didn't have it connected up properly to talk to a real person.  

But, yeah, for the most part, no one even realizes that it's a bot. It's helped   literally   hundreds, if not a thousand different chats at this point.   Find their order, get their tracking number, figure out where their package is, you know,   get find products that they need, find something on the website they couldn't find. Right? It's done all that stuff.   And,   you know, I used to have to answer that. So it's great. Mhmm. You know? And nobody lost their job   because we didn't hire somebody to do it. I was just doing it myself.   So, Ethan, how can people get ahold of you if they wanna connect?  



[01:05:38] Ethan Holland:

I think LinkedIn is the best place. My name is Ethan Holland, and you can find me on LinkedIn and just send me a message. I'd be glad to connect.  

[01:05:46] Matt Rouse:

Sounds good. You can get all your AI news at ethan b holland dot com.   And, Ethan, it's great chatting with you today. It's nice for us to be able to connect, rather than just chatting back and forth on the socials there. Yeah. I'm really glad we connected too. It's been nice to get to know you.   Sounds good. Alright, Ethan. We'll talk soon. Talk to you later. Bye.  

[01:06:07] Narrator AI:

This voice over used to be done by a human, but now it is synthetic.   Oh la la.   If you want to know if your job or business is safe from disruption,   read Matt's new book,   Will AI Take My Job?   Predictions about AI in corporations,   small business, and the workplace.   Available now on Amazon.   Trust me. It'll be worth it.