YT YouTube2Text

YouTube2Text → Transcripts → How Google Won The AI Race and What You Need To Do To Win In The Age Of AI | Kasim Aslam

How Google Won The AI Race and What You Need To Do To Win In The Age Of AI | Kasim Aslam — Transcript

by Joe Polish · 8,175 words · 1,182 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00What I'm up here to tell you is I think
0:01Google's going to win. I think Gemini is
0:03going to be the most powerful AI tool in
0:04existence. Google owns all the most
0:06important websites on the planet. If you
0:08think about it, Google basically
0:10captured the collective [music]
0:11consciousness of humanity and they put
0:13it on a hard drive. They took every
0:15piece of information every human [music]
0:17has ever created since the advent of the
0:19internet and before cuz that was all
0:20digitized too. [music] And then they
0:22stored it for itself for a someday use.
0:26And so [music] I think in in a world
0:28where we're watching a space race,
0:30Google has the [music] fuel necessary to
0:33go to space and everybody else right now
0:35is actually borrowing from Google's
0:37fuel.
0:37>> Yeah. So let me let me introduce my
0:39friend um Kawsum Aslam. Okay. And that's
0:43pronounced like awesome. Okay. And
0:46I still mess it up every once in a
0:47while. He's a college dropout. See what
0:50I was telling you? We're not we're not
0:51really reaching for the you know lot of
0:54lot of skill here. He's a college
0:55dropout turned powerhouse entrepreneur,
0:57author, and proud front row debt. He
0:59built and sold the world's leading
1:01Google ad agency in 2022, and he now
1:04heads paro talent uh.com, a high
1:07performance recruiting agency for
1:09executive assistants. Uh he's also my
1:11partner in seedol.com.
1:13And uh you know, I'm going to ask him a
1:16few questions uh about some different
1:18things and I think he might learn a
1:19thing or two. So give it up for my good
1:21friend Kasa. [applause]
1:24Thank you. Do you have a preference of
1:26seats?
1:26>> I like that one.
1:27>> Okay, go ahead.
1:28>> Yeah, go ahead.
1:29>> And uh so, oh, you know what I need to
1:31get from uh Ununice? Do you mind
1:32bringing me the uh uh phone that Oh,
1:36>> yeah. that I had earlier because I just
1:38realized I don't have I put something in
1:40chat GPT
1:42>> based on that. I want Thank you. that I
1:44wanted to ask you. So, I'll come to
1:45that. So, you had What did I not say
1:49about you that people need to say? just
1:51rattle off a little bit about uh what
1:53you what you do now, what you used to do
1:56with your agency. And we're going to
1:58talk a little bit about uh AEO versus
2:01SEO and different stuff like that. And
2:04then we're going to do some Q&A with the
2:06audience because I know once you get
2:07talking, people are going to become
2:09intrigued. Uh I built the number one
2:12ranked Google Ads agency in the world. I
2:13had that business for about 15 years.
2:15And the reason that's relevant is
2:16because I spent $100 million a year of
2:18other people's money. Um, thousands of
2:20clients went through my shop. So, I got
2:21to see a diversity of industry, which
2:23was helpful. And the reason that's
2:25helpful is because Google, Scott
2:26Galloway, has this really amazing
2:28keynote. A lot of you have probably seen
2:29it where he talks about how I forget the
2:31the other connections he makes, but he
2:34says something like uh Amazon is um
2:37love, Apple is sex, Google is God. And
2:40the reason Google is God is because on
2:42an anthropological scale, the entity to
2:44whom people pose their questions used to
2:46be God. So, you're a shepherd in a
2:48field. 2,000 years ago and your child is
2:50sick and you have no other option but to
2:53drop to your knees and look to the sky
2:54and say why, how, can you please, etc.
2:57Flash forward and this isn't a
2:59usurpation of a deity, but but I think
3:03it's really profound to think that that
3:06that relationship has actually been
3:08usurped by Google or it had been. And so
3:11Google is the closest thing we've ever
3:14gotten to something like
3:18uh like intent alchemy. It was the first
3:22time ever that human intent started to
3:26be distilled in a way that could be
3:28measured and managed. Uh and that's
3:30different. Google knows what you ask it.
3:31Facebook knows what you tell it. And it
3:33it just it led to kind of this ongoing
3:36everpresent obsession with with the data
3:41with just analyzing the data. I have an
3:44entity now that I Here's the thing. I
3:45don't have an offer. Evan's been making
3:46fun of me for the last 24 hours because
3:48I don't
3:49>> gently guiding.
3:50>> Yeah. Gently guiding. I There's nothing
3:53gentle.
3:54Um I don't have anything to sell on this
3:58this business line yet. I have six
4:00full-time employees that have done
4:02nothing but research how LLM source
4:04their citations because I'm interested
4:06in that. And this is why you don't give
4:08poor people money because we will spend
4:10it in fun and interesting ways.
4:11[laughter]
4:12Uh but we've we've re I have more data
4:15on this topic than any human alive I
4:16think except Monik Bond who owns Search
4:18Atlas who's also a buddy but his the way
4:20he's acquiring the data too I don't
4:22agree with his thesis gen brilliant
4:23brilliant guy by the way and I'd love
4:25for him to be able to combat that
4:26because I'm sure he's smarter than I am.
4:27Uh but I think that this is it's
4:31critically important to not just
4:33business but also humanity because
4:35there's a lot of directions we can go
4:37and I promise not to go in all of them
4:38but uh the the the one I would offer
4:41just as an example is the way that LLMs
4:43are currently delivering information is
4:45very and everybody here has experienced
4:47this themselves. As a matter of fact I
4:48was talking to John who was like oh yeah
4:49I know that and I've been saying that
4:51LLM have inbuilt confirmation bias to a
4:55degree that's horrifying. So, if you're
4:57provax and I'm antivax and I'm asking
5:00LLM about vaccines and you're asking
5:01LLMs about vaccines and you're looking
5:03for peer-reviewed scientific literature,
5:05you will get exactly what you're looking
5:06for and I will get exactly what I'm
5:08looking for. And there's something about
5:10we already live in a world that has the
5:13propensity for echo chamber. And now we
5:16just weaponized this trillion dollar,
5:18you know, nuclear technology to just
5:21send people as far and fast as they
5:23possibly can in the direction they're
5:25already heading. And and there's
5:28something about that that's terrifying
5:29to me. And so, and there's also a way
5:31that I'm going to use it to make tons
5:32and tons of money. So, I've got like the
5:33evil capitalist in me and then also the,
5:36you know, there's just something about
5:37it that I feel we need to have these
5:38conversations and put those guardrails
5:40up. How did I do there, Joe? I ran.
5:41>> No. you know, no, let's just say, okay,
5:43what about the non-avieval capitalists,
5:45right? What like what's the No, I think
5:48that's that's great. Can you just just
5:50the most basic level define uh LLMs,
5:54large language models? What what the
5:55hell does that actually mean? How does
5:56it apply to everyone here?
5:58>> Yeah. So, so first of all, it's actually
5:59it's inadequate nomenclature. And I've
6:01I've tried to because words are so
6:03important. Semantics are so important.
6:05The only thing that separates all of us
6:06from any industry is semantic
6:08architecture because like real estate
6:10guys scare the daylights out of me. and
6:11M&A guys scare the daylights out of me
6:13because they say like I just go
6:14full on like 10th grade dropout and but
6:17as soon as you can flatten semantics I
6:19feel like it puts everything kind of on
6:20an even playing field. So LLM
6:22technically is large lang language model
6:24and a large language model is a
6:25mechanism within which we've put uh if
6:28you've ever done the research on how
6:29they train AI it's really interesting
6:32there's nothing that's been coded we've
6:34been taught to believe that all anything
6:35that happens on the computer has been
6:36coded by a human instead what they did
6:39and I'm going to put this in just the
6:41the
6:45I'm going to do my best to make this as
6:47accessible as I had to make it for
6:48myself I guess is what I'm saying so I
6:49had to dumb this all the way down but
6:50they they took a computer and they said
6:52it gave it 100,000 pictures and they
6:54said this is a man this is a woman this
6:55is a man this is a woman this is a man
6:56this is a woman this is a man this is a
6:58woman and then they started to ask the
6:59computer back what's a man and what's a
7:01woman but they don't know why the
7:04computer was able to identify they don't
7:05know high cheekbones low brow Adam's
7:08apple like they don't know what about
7:11the pattern recognition enabled the
7:12computer to know this is a man this is a
7:14woman and then you do that on every
7:15level of analysis and you cascade it
7:17across every data repository you can
7:19possibly get your hands on and that ends
7:21up being an LLM. An LLM is um it's a
7:25it's a it's a pattern recognition
7:28mechanism. Now LLM [snorts] isn't the
7:30right word for all AI. Not all AI is LLM
7:32and not all LLM is AI and nobody's
7:35really clear on on which ones are and
7:37which ones aren't. So we're saying LLM,
7:39but that unto itself is a problem. What
7:41I like saying right now is AI engine
7:44because not all of these are are trained
7:46on language, which is really really
7:48important. Some of them are trained on
7:49models per Simon's genius and some of
7:51them trained on images. And so we're
7:53we're we're kind of painting with this
7:55broad brush. But the AI engines are
7:58pattern recognition mechanisms. And the
8:00reason that that's so important is
8:01because the propensity for
8:04uh confirmation bias is built into that
8:08framing. Pattern recognition means
8:11compliance with the established
8:14understanding of how something works. So
8:17we're it's basically peer-reviewed. We
8:19all agree that this is what something is
8:21or what it looks like and then the
8:23computer goes, "Oh, you all agree that
8:25must be true." It reminds me of the
8:28Einstein had this really great quote.
8:30There was a 100 scientists that wrote a
8:32treaties against whatever he was saying
8:33with I think theory of relativity and he
8:35said, "Well, if I was wrong, it wouldn't
8:36have taken a hundred. It would have just
8:38taken one."
8:38>> [snorts]
8:39>> And it's a reference to the fact that
8:42the hierarchical structure in terms of
8:44information is actually very often
8:47inversed in terms of who believes what
8:50and and who's actually right. And and
8:53the AI engines are built for
8:56confirmation from the aggregate. Um
9:02I'll stop there.
9:04>> Well, so let's go back to our trip. um
9:08uh Whitney
9:10Parardis where I'll find in Texas and
9:13you gave the whole spill about uh Google
9:16and how they're the most powerful. Uh
9:17and then we'll come back to some of
9:19this. I want cuz I thought that was such
9:21an incredible insight and and uh we
9:24talked a little bit about that at a
9:25previous Genius Network meeting and I
9:27just would like everyone to hear that
9:29and that that'll give some contextual
9:31understanding. And then we'll talk about
9:32SEO and we'll talk about how to use all
9:35of this in order to position your
9:38business, what to pay attention to, what
9:40not to pay attention to, and get some
9:41predictions from you about where things
9:43are going. Not like Nostradamus
9:45predictions, but just by stuff that
9:48you're looking at every freaking day.
9:50Uh, just out of curiosity, who's who's
9:52if you're using AI on a daily basis, who
9:54prefers chat GPT to all else? It's about
9:57about 30 30% of the room. Okay. Who
10:00would prefer Claude to all else?
10:02Grock.
10:05Uh, Perplexity.
10:07Okay. Gemini
10:11one.
10:13Gemini wins. I'm so long on Google. And
10:16I'll tell you why. Now, here's the
10:18thing. Nobody raised their hand when we
10:19said Gemini. Because when you use
10:20Gemini, you feel like you're getting
10:22inadequate responses. When you use
10:24Claude, Perplexity, Chat GBT, you're
10:27getting mature results. And the reason
10:29you're getting mature results, and I can
10:30prove this from a data driven
10:31perspective, proves of a dangerous word.
10:33I can show you correlation isn't
10:35causation, but I I can show you why I
10:36believe this to be true. All of the
10:39other AI tools are just googling
10:41something faster for you. So, if you ask
10:44ChatgPT or Claude or Perplexity a
10:47question, they're going into Google's
10:49very, very, very mature forest because
10:51Google search is the most complex search
10:54graph that's ever existed. and they're
10:56mining the forest and they're finding
10:57the value and sometimes there's some
10:58distillation of the value, sometimes
11:00there's some amalgamation of the value,
11:02sometimes there's amplification of the
11:03value, but it's Google's mature forest
11:05from whence the value comes and then
11:07they bring it back to you. When you ask
11:10Google
11:12that same question, it's not going into
11:15its very mature forest. Google is
11:17delivering a sapling. And so you look at
11:20the Gemini results and you think like,
11:21god, this is garbage in comparison. And
11:23it's because Google, and this is the
11:25craziest part, Google is not using
11:27Google's own search graph in its
11:29AIdriven results, which is insane. It's
11:32insane. So, here's how we know this.
11:35Hey, I hope you're enjoying this video,
11:36and I want to let you know that I have a
11:38new book that's come out, [music] and if
11:39you'd like to get it absolutely free,
11:41there's a link below in the description,
11:44or you can wait till the [music] end of
11:45this video, or you can simply go to
11:47joe'sfreebook.com and you can get a copy
11:49there.
11:52I have six employees, one supervisor,
11:54and five technicians. Their job every
11:57single day, it's a soul sucking job.
11:58It's horrible. That's how I sell it, by
11:59the way. It's how I hire them, is they
12:01have to ask 20 uh 20 questions within
12:05the confines of 10 categories. And the
12:06categories are the ones that you would
12:07guess at. So, it's things like local,
12:08brick and mortar, SAS, professional
12:10services, info marketing, etc., etc.
12:12[snorts] Um 20 queries. And every query
12:16is asked of three LLMs. And and the
12:18number three has more to do with the the
12:20the data crunching tools that we use and
12:22the inability to to expand it any
12:23further. So you take 20 questions. So
12:25like one question would be um the the
12:28best boutique coffee shop in Kingman,
12:30Arizona is the example that I use
12:31because the the data is I think the most
12:33clear. And we use hyperspecific phrases
12:36like that. You want qualifiers. So like
12:37best, closest, nearest, top, uh best
12:40boutique is is hyperspecific. And you
12:42want that because you want to get rid of
12:43the noise. You want to get through to
12:44signal. So those qualifications and kind
12:47of commercial intent tend to be really
12:48important. I also like smaller niche
12:51industries like Kingman, Arizona doesn't
12:53have a lot of boutique coffee shops.
12:54Manhattan does. So if I get a result
12:56from an LLM in Manhattan, it's harder
12:58for me to understand how that hierarchy
13:00might be structured where in Kingman
13:02it's a lot easier. So that's one query.
13:04That query is asked of three separate
13:05LLMs and then he's and then that
13:07person's tasked with asking another
13:08query and then another query and another
13:10query and they're all based on that same
13:11category. So it might be best boutique
13:13coffee shop in Kingman, Arizona. And
13:14then it might be uh non-GMO coffee shops
13:17in Kingman, Arizona. And then it might
13:18be CED oil free coffee shops in Kingman,
13:20Arizona. And so on and so forth. And we
13:21do that 20 times across three LLMs. And
13:23then the LLM delivers an answer or
13:26multiple answers, but generally
13:27speaking, it delivers an answer. But the
13:30answer is I'm less concerned with what
13:31it also delivers is a collection of
13:33citations anywhere between one and 10.
13:35Sometimes more, but 10 tends to be the
13:37the the max for the most part.
13:38Citations. And the citations are what
13:40the the AI engine has told us it used to
13:42come up with that query. So the question
13:45we have to ask ourselves is why why that
13:47citation? What made you think that was
13:49relevant? So my team takes every single
13:52individual citation and they map it. I
13:55can show you actually. Would you all
13:56mind? Let's throw up slide 10.
14:03What do I point at to keep it going?
14:06There we go. Oh, we'll go back. So,
14:09>> and if you you know we can put it on
14:10this one behind us too if you guys want
14:12so it's easy.
14:12>> Oh, I can see it there.
14:14>> Um,
14:14>> perfect. So, just a a quick soundbite a
14:17precursor. This is a HubSpot and this is
14:20something that HubSpot shared. Organic
14:22traffic has dropped. It's a precipitous
14:24cliff and it's actually gotten worse.
14:25This is I think a year old data. This
14:28says 60%. It's now 65% of all searches
14:31results in zero click. So, that's where
14:33we are. It's already click mageddon is
14:34what people are calling it. like you're
14:36not getting the traffic to your website,
14:37which has a whole bunch of other
14:38implications, too. Um,
14:41this is what I want to show you. This is
14:43what our research looks like. So, those
14:44are the three LLMs in this particular
14:46example, and you could swap those three
14:47out for any. Gro is the outlier, by the
14:49way, because it's the only AI tool
14:51outside of Gemini that really has its
14:52own data repository, but there are some
14:53flaws to it. Uh, Meta's data repository,
14:56meta goes out more than it goes in,
14:58interestingly. And so we take every
15:00single LLM, we ask the same query of the
15:02LLM, and then we track all the URLs, and
15:05then we look at the URLs along the level
15:07of analysis that you would assume based
15:10off of conventional wisdom. So we're
15:12looking at things like keyphrase
15:14permeation, back links, page speed,
15:17schema markup. There's like 50 some odd
15:19data points. And so if you take one
15:21query, three LLMs, 10 citations per LLM,
15:2550 data points per citation multiplied
15:2720 times, multiplied by five employees,
15:30multiplied by every business day for the
15:31last year, that's how much data I have
15:33on this. So it's it's not nothing, I
15:35guess, is what I'm saying. It's
15:36supported by some level of, you know,
15:38active um butressing. And what's really
15:41interesting, too, is a lot of the
15:42stronger assumptions we made are only
15:43proving to be more true. So we think
15:44we've kind of caught the direction with
15:46which this is going. Can y'all go? There
15:48we go. So, this is what this really
15:49looks like in real time. And uh this is
15:52the part that's really interesting. And
15:54I usually do a great big grand reveal.
15:55Here's ChatGpt as an example. Best
15:57boutique coffee shop in Kingman,
15:58Arizona. Those are the citations. These
16:00are the the the citation prowess, let's
16:03say. Um those are the things that we
16:04would be analyzing from an SEO
16:06perspective. There's a reason this is
16:07massively flawed, by the way, and I only
16:08figured it out half. We're like nine
16:09months into it. Um again, more data
16:12analysis, page speed, schema markup. Um,
16:15and then here's query two, where to find
16:17specialty coffee in Cayman, Arizona. So,
16:19we're getting really granular and really
16:20specific. Um, we can kill this now. Kill
16:23the slide. So, this is the part that's
16:24really interesting is when uh we get
16:27these citations, one of the things that
16:28we check is
16:31in
16:33if you take the rank order that the LLM
16:35delivers a citation and the
16:36chronological rank shows some level of
16:38prowess, we we check that against the
16:41Google search results. And so if it
16:43ranks number one in the LLM, where does
16:46it rank in Google? And maybe it ranks
16:47one or two or three or four or five or
16:4810. Um, Chat GPT from memory has
16:52somewhere between a 40 to 60% overlap
16:54rate with Google. Uh, call it 50% on
16:57average. Um,
16:59Perplexity has a 70 to 80% overlap rate
17:02with Google. Claude's somewhere in the
17:04middle. Guess what Gemini's overlap rate
17:05with Google is
17:08in certain categories? It's zero. It's
17:10zero. And the the highest it gets is
17:1225%. Which means, going back to the
17:14original thesis, Google is not using
17:16Google's own search graph. Google is
17:18building its own knowledge graph from
17:19the ground up because Google's the only
17:21company on the planet that has the data
17:23to do that. Amazon could do it with AWS,
17:25but it would be a gross violation of
17:26their terms of service. They can't use
17:28that data. Apple has some proxy data.
17:30They have a more valuable window, but
17:31they've never captured it. They took a
17:33really weird stance, by the way. The iOS
17:3514 thing was an interesting direction
17:36for them to go, and I think they'd have
17:37to regret it. I don't I don't know if
17:39that was a miscalculation or if they're
17:40smarter than me and there's, you know, a
17:42switcheroo coming, but Google's the only
17:44company on the planet that has the data
17:46repository that would be necessary in
17:47order to train its own AI engine to this
17:49degree. And so the reason that Gemini
17:51and Google has a propensity for staying
17:54back and letting uh the the other
17:56pioneers take the arrows, they did it,
17:58there was a congressional hearing about
17:59Facebook because Facebook had 55,000
18:01demographic and psychographic profiling
18:02factors on people and they allowed
18:04advertisers to use those profiling
18:05factors. So what do we have that
18:07actually triggered GDPR triggered
18:08privacy concerns triggered the geriatric
18:10morons in Congress you know coming at
18:13Zuckerberg asking them what is the
18:14internet at when that happened Facebook
18:17had 55,000 Google had 72 million
18:19demographic and psychographic profine
18:21factors but they didn't they didn't lift
18:23the hood and let people see that so
18:25Google has a storied history of having
18:28all of these tools that they don't
18:30release to the public they wait look
18:32what they did what's the image
18:33generation thing banana panda something
18:34banana Nana, banana. All of the sudden,
18:38Google goes from not being on the im the
18:40AI image map at all to releasing what
18:44ostensibly is, from the people that I
18:46know the greatest of of everything that
18:48exists. And it happens overnight without
18:50any warning. They wait they wait to see
18:52what happens with legislation. They wait
18:54to see what happens with public opinion.
18:55And then as soon as they can tell the
18:56direct Google's run by attorneys, as
18:58soon as they can tell where the wind is
18:59blowing, then they then they release.
19:01And you'll notice that Nano Banana came
19:04out almost immediately after legislative
19:07precedent was established as far as what
19:08happens with these images. So Google and
19:10here's the craziest part. When you look
19:12at Chachi PT Claude Perplexity, they're
19:16drawing from a well that Google owns the
19:21access to.
19:23It's a flawed well. We know SEO is
19:25flawed. Any SEO who tells you that SEO
19:27is about quality content is absolutely
19:28full of SEO is about link
19:29building. You buy links, you rank.
19:31That's a fact. It's the reason SEO.co
19:33and the Hoth and and Pausy rank exist.
19:36There's lip service given to the
19:37content. And you do need some content in
19:39order to rank obviously, but that's not
19:41what gets the the the the top rank.
19:45Google search is flawed and Google knows
19:46that. So, Google abandoned Google search
19:48graph is building its own knowledge
19:50graph from the ground up. And the day it
19:51wants to, it can rip the rug out from
19:53every single other LLM. Now, how long
19:55does it take them to do that? Will they
19:57do that? Who knows? But I'm so long.
19:59I've only ever lost money in the stock
20:00market, by the way. My thesis has always
20:02been buy high, sell low, cry. And if
20:05anybody wanted a stock tip from me,
20:07dude, I'm I'm deep deep deep in Google
20:09because I really believe in in the the
20:11the the data they have and their ability
20:14to to execute upon it. Um I just think
20:18it's going to be really extraordinary.
20:20And so I I really like the other thing
20:22that's really interesting. I bought my
20:23children Chromebooks. They're eight and
20:2510. And I was just thinking through like
20:26their life from a vocational perspective
20:28and I'm like there's nothing left to do
20:29in 10 years. What am I going to teach
20:31them? And I decided two skills. One,
20:33learn AI obviously. And even that is is
20:36it I think that skill is what would you
20:38call it? It's almost transitionary. Is
20:40that a word? Transitionatory.
20:42>> Transitory.
20:42>> Transitory. Yeah. So it's a transitory
20:44skill that's not going to be applicable
20:45for very long, but it's important for
20:46them to know. The second one I actually
20:47think is authority building. So I bought
20:49my kids Chromebooks and I said, "You can
20:50only use this to learn AI and build
20:52authority." So, they're shooting YouTube
20:54videos for other little kids, teaching
20:55them how to use AI. Here's why this is
20:57relevant. Buy them Chromebooks. Fire up
20:58Chromebook. Help them create Gmail
21:00accounts. Guess what Google's default
21:02search engine is on Chrome. Guess Quick
21:04Like Bunnies.
21:08That's the browser. The search engine is
21:10Gemini. Gemini is not a search engine.
21:13Google killed it took Google Search and
21:16shot it in the back of the head. It
21:17still accounts for 85% of Google's
21:19revenue. the fact that, you know, the
21:21third most valuable company on the
21:24planet took its highest performing asset
21:26and no longer makes it the default on
21:28all of its new devices and instead
21:30delivers Gemini. The way to think about
21:32these things, this is how I've explained
21:33it to myself. Search
21:35>> is an interstellar space station. No, is
21:38is a is a excuse me, is a spaceship.
21:40Okay? So, when you're searching, you're
21:42on your planet and there's galaxies and
21:44universes and all sorts of things out
21:46and if you want to search and you're
21:47like like right now I'm shopping for a
21:48car. So, if I want to buy a car, I might
21:50have 20 levels of analysis. And you
21:52don't realize that's how complex
21:53commerce is, but it is. There's 20
21:55levels of analysis potentially. There's
21:56things like, you know, maintenance and
21:58fuel efficiency and let's say you have a
21:59special needs child or I'm about to buy
22:01new land or whatever. There's 20 levels
22:02of analysis. So, I choose one level of
22:04analysis and then I hop into my my
22:06international spaceship and then I say,
22:08"Okay, I want to learn more about um uh
22:11vehicles for children with special
22:12needs." And then it brings me to the
22:16galaxy that has that information. And
22:18then I have to go from planet to planet
22:19to planet and sometimes moon to moon to
22:21moon and I get to mine that information
22:23to my heart's content and then I capture
22:25the information that I want and I do the
22:26amalgamation I do the distillation I do
22:28the amplification and then I bring that
22:29information back home and then I say
22:31okay now I want to learn about fuel
22:33efficiency and then I shoot off again
22:35into this galaxy and so on and so forth
22:37and on and on and on it goes and that's
22:38how search works. It's a dumb mechanism.
22:41AI is a little android that you say hey
22:45I want to learn about cars. These are my
22:47concerns. Go. And it goes to the
22:50galaxies and and the moons and the
22:52universes for you. And it understands
22:55the dependency and the interdependency
22:57sometimes better than you do. And then
22:59it brings you back the information. And
23:02this is relevant because the LLMs or the
23:06AI engines are pulling from repositories
23:09of content. And everybody right now is
23:13fishing in an ocean that anytime Google
23:16wants to they can nuke which is amazing.
23:19It's amazing and the implications are
23:20staggering. So I think what we all need
23:22to do is really diversify as we go for
23:25the traffic game and traffic is the most
23:27valuable facet I think of the business
23:28endeavor. All the big companies are
23:30Amazon is a traffic company. Met is a
23:32traffic company. Google's a traffic
23:33company. If you own the eyeballs, you
23:34own the business. And so I think we need
23:36to a diversify and then b I think
23:39everybody in this room needs to do what
23:40that man does. I think authority
23:42building is the most important
23:44in and I'm not good at this by the way.
23:46I've got nothing to sell here. Like what
23:48Evan Carmichael does look at, you know,
23:50everybody's like Alex Carlos made $105
23:51million in a day. No, he didn't. He made
23:54$105 million in seven years. But he
23:57turned himself into his own traffic
23:59repository. And those traffic
24:01repositories are they're unbelievably
24:04potent. the value in them. It's a wealth
24:06from which you can always pull. And and
24:08we're all used to being on this this
24:09strange merrygoround where you have to
24:11pay every single time you pass go. And
24:14that's that's it's a horrible framework
24:15for business. And so I think that that
24:17the world of AEO is going to change that
24:18in a lot of ways. Um but I I have some I
24:21have some theories on how we can take
24:23advantage of that and make some money.
24:25>> Okay. So well, let's get into that. Uh
24:28does anyone have a question for him? Is
24:30anyone lost in any way, shape, or form?
24:32throw him a mic because I want to make
24:33sure that you're following this.
24:35>> I'm lost.
24:36>> Okay. Sorry, Beth.
24:38>> Bet. Yeah. Um, but
24:42>> Well, what
24:43>> what you're saying is
24:46okay. Tell me diversify. Diversify.
24:50>> Don't just focus on chat GPT. So, if
24:52you're using Chat GPT and you think I
24:54want my business to be mentioned by
24:55ChatGPT, you actually don't. You want to
24:57make sure and this is this is a new
24:59paradigm switch too because we're so
25:00used to single platform optimization.
25:03Everybody in here has their one thing. I
25:04was just talking to somebody about this.
25:05They're like, "Man, I nailed meta ads
25:06and meta ads are great for my business."
25:08And then somebody else might be like, "I
25:10do Google search." And even even within
25:11the confines of that paradigm, Google
25:13isn't the only search engine, right?
25:14There's Google and Yahoo and like Oscin,
25:16etc., etc. What we need to do right now,
25:18and this won't be true. I expect
25:19consolidation because that's the way
25:21technology works. What we need to do
25:22right now is we need to have multiple
25:24horses out on the track. You need to
25:26have your chat GPT track, your your
25:28Claude, your your Grock, your Plexity,
25:30etc. If you want to play this game, this
25:32is a game of land staking. This is
25:34taking the beach head. This is internet
25:36circa 1997 when somebody got to go in
25:38and buy domain names like business.com
25:40for $12 and then sell them for 40
25:42million or whatever it went to. Or, you
25:43know, circa 1800s when some idiot just
25:45got to get off the boat, plant a flag
25:47in, you know, 1500 acres of the most
25:50beautiful Oregon territory and say,
25:51"This is mine." we can right now capture
25:55this this intellectual property in a way
25:58when the LLMs deliver a result. The
26:01thing that's really interesting is
26:03there's there's a um
26:06there's a first mover advantage and
26:08there's there because confirmation bias
26:11is built into the mechanism once you
26:13capture it. It feels like the fact that
26:16you were the first justifies you being
26:18the first and so we'll continue you
26:20being the first. Um, but the LLMs
26:23operate on levels of granularity that
26:24are a little bit different than Google
26:25does. So, we can talk about that in a
26:26minute. But what I'm really saying is be
26:28solution agnostic. Do not There are no
26:31sacred cows and we need to really kind
26:34of try to do things that allow you need
26:36a foot on the boat and a foot on the
26:38dock and you probably need six or seven
26:39or eight feet because we don't
26:40necessarily know who the winner is. What
26:42I'm up here to tell you is I think
26:43Google's going to win. I think Gemini is
26:44going to be the most powerful AI tool in
26:46existence. Google owns all the most
26:48important websites on the planet.
26:50YouTube specifically being I think the
26:52most important content repository second
26:54only to a a dark horse that I'll mention
26:56in a moment. How did I do there Babs?
26:58>> Pretty good.
26:59>> Okay.
27:00>> Pretty good.
27:01>> Thank you.
27:02>> So I I really want to be on team
27:04>> Gemini. Yeah, [snorts]
27:05>> I want to be on team because Gemini is
27:07Google and Google's YouTube and like I
27:09really want to be there. Um
27:10[clears throat]
27:11>> you turn lights up.
27:12>> From my from my own anecdotal
27:14experience, I get better results with
27:16GPT so far. Uh from a LLM Marina
27:19perspective, they're always back and
27:21forth on like who's winning. Uh from a
27:23talent perspective, I think OpenAI is
27:25getting more of the missionaries of the
27:27people who want to change the world with
27:29AI. And my my thesis on why Google is so
27:33far ahead right now and like launching
27:35Nanaban, all these other tools out of
27:36nowhere.
27:37>> They've been doing this for years.
27:39>> They had their own LLM like ready to go
27:42way sooner than anybody else.
27:45OpenAI just released it and so they had
27:47to so all the things that have been
27:48coming out are not that they've
27:49developing amazing things that they they
27:51already had a lot of these things and
27:53now they're just releasing them and so
27:55like where do we go from here and can
27:56Google stay ahead of the curve? I don't
27:58know. I don't I personally I don't think
28:00they're bringing in the top talent to
28:02stay ahead of the curve and I think GPT
28:04is going to or open AI is going to is
28:06going to be the better solution. I I I
28:09hope that's not true because I love
28:11YouTube.
28:12>> No, I think it absolutely could be. What
28:13I will say is you mentioned Google has a
28:15qu of amazing AI tools that they have
28:18yet to
28:18>> No, you threw that in there. Qu is not a
28:20word unless you're in Australia.
28:22>> Where's Simon?
28:24>> Thank you.
28:24>> Okay. So, [laughter]
28:27>> the question I wanted to ask, so there's
28:28there's something you said we can
28:29disagree and go about. I mean, I would
28:30love for you to be right and but we
28:32can't really test that and we'll have a
28:34conversation in five years and figure it
28:35out.
28:36>> One one thing you said though that I
28:37didn't understand and I'm trying to. You
28:39said that uh Google Gemini is the only
28:42one building their knowledge graph from
28:43the ground up.
28:44>> Correct.
28:45>> But you said Google is also the one that
28:47has all the data.
28:48>> Mh.
28:49>> So
28:51I'm like it has the least amount of
28:53overlap.
28:54>> So if you're building from the ground
28:55up, why does it matter if you have
28:57overlap or you have all the data?
28:59>> Well, so Google's not using Google's own
29:00search graph. Google Google is not a
29:02fetch. Google's a a capture. So what
29:05Google does when it crawls a website is
29:07it actually takes all that information
29:08and it ces it.
29:11>> No, just not caches. All right,
29:13>> now he's doing it on purpose.
29:14>> Yeah. Uh Google takes that information
29:16and and it stores it for its own
29:18purposes and it and it does it on an
29:20iterative level that's actually
29:21overwhelming. Like the amount of
29:22information that Google captures is
29:24insane and nobody else is from a search
29:26all not I I can't say all every search
29:28engine I've ever been exposed to from an
29:30internal perspective was a fetch not a
29:32not a a catch. Okay. But but why is that
29:34an advantage? Like why couldn't OpenAI
29:36or somebody else do that too?
29:38>> Uh they they can but you need So Google
29:40has been doing it for 20 years. It has
29:42the computing power to do it. Google, by
29:44the way, y'all know Google broke
29:46quantum, announced it very quietly, and
29:48then has never said an effing thing
29:49since. So open AI is is data thirsty and
29:53p and low power. They have they went and
29:56asked Congress for $14 trillion because
29:57they don't have the computing power to
29:58do what they need to do now. for them to
30:00go do what Google has done over the last
30:0220 years I think is it's not an
30:03insurmountable problem but it's a really
30:05significantly sized problem. Google has
30:07captured all of this data and LLMs
30:10because of the way LLMs are trained.
30:11Remember the the the kind of weird sort
30:14of proxy training that I tried to
30:15explain earlier that's how they still
30:17work and Google is the entity that has
30:20that information. AWS hosts something
30:22like 30% of the frontfacing web. So,
30:24Amazon could do it because they actually
30:26have this information on their servers,
30:28but if you read an Amazon's terms of
30:29service, they cannot access that
30:31information, generally speaking, unless
30:32there's a specialized clause. That might
30:34change, too. We're keeping a very close
30:36eye on that. So Google's the only entity
30:38that that has the information and can
30:40train its own AI tool from the ground up
30:43where every other AI tool is still to
30:46this day chat open AI chat GPT is not
30:49storing it's fetching and it's the only
30:53thing it is doing is it's tracking
30:54updates which requires uh storing the
30:57same way that cookies are stored on your
30:58browser. It's like a temporary storage
31:00but it doesn't have the data and the
31:02data the content is that's that's where
31:05the value is. That's where you would go.
31:07If you think about it, Google basically
31:09captured the collective consciousness of
31:11humanity and they put it on a hard
31:13drive. They took every piece of
31:14information every human has ever created
31:16since the advent of the internet and
31:19before because that was all digitized
31:20too. And then they stored it for itself
31:23for a someday use. And so I think in in
31:26a world where we're watching a space
31:28race, Google has the fuel necessary to
31:32go to space and everybody else right now
31:34is actually borrowing from Google's
31:36fuel. But in an interesting way, they're
31:38borrowing they're borrowing within uh a
31:42framing that is massively flawed because
31:44Google SEO there's no nobody here would
31:47would I I I think everybody would at
31:48least agree with the fact that just
31:50because you rank one on Google doesn't
31:51mean that you're the best at anything.
31:53Like if you search for a chiropractor on
31:54Google, it doesn't mean that's the best
31:55chiropractor, right? That's the person
31:57who gained the mechanism. We've gotten
31:59used to that, but that's that's
32:01massively flawed. And so the other LLMs
32:05relying on Google ranking means that
32:07they're actually climbing a ladder
32:09propped against a wall that we know for
32:10a fact Google's going to take a wrecking
32:12ball to someday. And Google having the
32:14data that they do have allows them to
32:16rebuild the knowledge graph and rebuild
32:19it. This is what's really fun is from
32:21any frame in any paradigm they so
32:24choose. So Google page rank was based
32:26off of something called um was Google
32:28ranking was based off of something
32:29called page rank and it's a
32:31tongue-in-cheek because the guy's name
32:32is Larry Page and it's a peerreview
32:34system. It's inbound links equal value
32:37but but it's a meritocratic structure
32:39where the higher the the the inbound
32:41links uh the the higher the value of the
32:43inbound links the more value you got.
32:44and those links were ranked
32:46meritocratically and so it was kind of
32:47this meritocratic web. But you could
32:49reframe that and you could say something
32:50like um peer review based off of um you
32:56know Doug Bman wrote a book is has a
32:59doctorate has however many readers he
33:01has. So if Doug Brackman goes in and
33:03says and likes a comment and then I come
33:07in and like a comment and Google could
33:09decide no the doctorate author his like
33:13is worth a 100,000 more points than mine
33:15is and completely reframe the paradigm.
33:17And so the content paradigm Google can
33:19own and reframe the content paradigm.
33:21All the other LLMs if they wanted to
33:23reframe the content paradigm they first
33:25have to use Google search graph because
33:26that's we're going to fetch and then
33:28they have to overlay their paradigm on
33:29top of it or use the paradigm to fetch
33:32but even that is is inadequate because
33:34they're they're you know it's you have
33:35to go through Google's hall of mirrors.
33:37Uh, and that that example I showed you,
33:39by the way, the most important website
33:41on the internet right now, if you really
33:42want to invest in something, get an EA
33:45or that stoner nephew in your aunt's
33:48basement to go crack Reddit. Uh,
33:52Profound did a study of millions and
33:53millions and millions of Google driven
33:55citations. And Reddit was cited one out
33:59of five times. One out of five times,
34:0221% of the time Reddit cited, 18.8% of
34:05the time it's YouTube. But YouTube's 19
34:08times the size of Reddit. Here's why
34:10they like Reddit. If you really think
34:11through Reddit, Reddit is the I I
34:13shouldn't say only. I have I have a bad
34:15habit of hyperbole also. I'm trying to
34:17do better at that. Reddit is the only
34:19site that I know of in mass that allows
34:22for peer-reviewed content and
34:25peer-reviewed content creators.
34:28So when somebody goes in and they they
34:31start, you know, talking about vehicles
34:33and I'm talking about cars and I don't
34:34know anything about cars and Keith talks
34:35about cars and he knows everything about
34:36cars. The fact that Keith has been
34:38upvoted on other car threads allows that
34:41that equity to be brought over. And if
34:44you think of it from an LLM's
34:45perspective, of course they'd love that
34:47because they need to know who to listen
34:48to. All of humanity is not created
34:50equal. And so something like Reddit
34:54gives the LLM the opportunity to say
34:56like, "Oh, you're the expert on this
34:58topic." Content that comes from you.
35:01Here's what's really interesting. And
35:02this is what my data doesn't show. And
35:04it's so frustrating because I don't even
35:05know how to capture it.
35:07My friend John Moran is the greatest
35:08Google Ads mind in the world. Everybody
35:10know who knows Google Ads knows John
35:11Moran. I was searching for something in
35:13chat GPT and I found an article. I was
35:17searching on the topic of of Google ads
35:19and I found an article from search
35:20engine journal that cited John Moran. So
35:23search engine journal comes up and chat
35:25GPT is the citation. But when I went
35:26into the article because I was
35:27interested most people do not pursue
35:29citations. I go into the article because
35:30I'm interested I realized like oh chat
35:32GPT doesn't think search engine journal
35:34is the authority. ChatGpt thinks John's
35:37the authority and search engine journal
35:38just happened to be the network that
35:39cited John Moran. This is why this is
35:41relevant.
35:43We have an inbuilt bias towards URLs. If
35:47a URL shows up, we assume the LLM thinks
35:50it's the URL that ranked. But no, the
35:52LLM knows because John Moran's famous on
35:54Reddit. Anybody who cites John Moran on
35:57this topic, I'm going to put their
35:58content up. I'm solution agnostic in
36:00terms of where that comes from. LLM's
36:02like people over brands 100,000 times to
36:04one. An LLM is going to want Joe Polish,
36:06not Genius Network, because a person can
36:08get a doctorate, a person can publish a
36:10book, a person can have a community, and
36:11these brands are very transient as we've
36:13seen. So, it puts us in this place where
36:15going back to me saying that everything
36:17everybody should do what you do. You you
36:19have to if you want to play this game,
36:21plant a flag as you as a person because
36:24the LLMs are going to want to draw from
36:26the authority that a human being
36:28establishes. And that authority should
36:31be on something like a niche and
36:33granular level. I've put myself at a
36:35massive disadvantage because I own a
36:36staffing agency for executive assistants
36:38out of Latin America. I also own seed
36:40oil.com with Cho. I also own
36:41techstack.com, syndications.com, aeo.co.
36:44Like, I'm buying all this because I
36:45went and made an exit and did everything
36:46they told us not to do because I didn't
36:47have any of these guys. And there's
36:50there's no place for me to establish
36:52myself as an authority within within the
36:53confines of the the paradigm that we're
36:55talking about now. And it's massive
36:57massive massive disadvantage. You kind
36:58of have to pick. Now, you can diversify,
37:00but the divers diversification also
37:01needs to be hierarchically structured,
37:03too. But if you can establish yourself
37:04as an authority, go crack the code on
37:05Reddit, create multimmodal content, you
37:07can own industries. And the industries
37:09that you can own, that's the thing I'm
37:10most excited about. Evan was talking to
37:11me about like, why don't you just offer
37:12this as a service? I don't want to offer
37:13as a service. I don't want anybody to
37:14know I know how to do this. I want to go
37:16capture as much market share as I
37:18possibly can because there are some
37:20industries right now where if I own, you
37:22know, think about it as being the neck
37:24of of the the the the head that is the
37:26LLM. If I own the proxy through which
37:28all this traffic goes, I think there are
37:30billion dollar opportunity. There's
37:32hundreds of billions of dollars spent on
37:33SEO that's just got wiped off the plate.
37:35And when when so that search now goes to
37:37an LLM result, but the LLM results don't
37:39rank content the way that search engines
37:41do. It's entirely different. And so
37:43everybody here has the chance to do what
37:45you could have done in 1992 when we were
37:47just keyphrase stuffing and capturing
37:48market share. And those people have been
37:50the category kings ever since. You get
37:51to do that now, but you get to do it
37:52from the LLM stance.
37:55I think this is the biggest
37:56entrepreneurial opportunity in the 21st
37:58century as it pertains to every
38:00individual business because most
38:02opportunities are oh we have to go do
38:03that thing. What I'm saying is this is
38:05the thing you can do in your business
38:07and I I don't know of anything else that
38:09would have it's it's you know you talk
38:10about leverage. It's the little hinge
38:13that swings the big door. So crack
38:15Reddit. YouTube is the greatest content.
38:17The reason YouTube is so good is because
38:19it's it's the most repurposable content.
38:21Long form YouTube can be turned into
38:22podcasts and reels and and
38:24transcriptions for blogs and carousels
38:26and infographics and on and on and on
38:28and on. And LLMs love multimodal
38:30content. They love knowing, oh, you have
38:31a blog here and a tweet there and they
38:33love comments, too. They love
38:34engagement. It's all these things that
38:36don't scale. It's all these things that
38:38don't scale. You have to do all the
38:39things that don't scale. You actually
38:40have to reply to people on Facebook.
38:42It's obnoxious. But being able to do
38:44that lets them know, oh, this person
38:45actually knows what they're talking
38:46about. And if you do that, you own that
38:49industry.
38:49>> Okay. Hey, I hope you found that video
38:51awesome and useful. So, if you want to
38:52get a free copy of my book, I want you
38:54to click here. And if you want [music]
38:55to watch some more videos that'll be
38:57useful and awesome, click here. Go
39:00ahead. They're over here. Do it now.
39:03Come on. Thank you. Watch them.

About this transcript

This page contains the full transcript of How Google Won The AI Race and What You Need To Do To Win In The Age Of AI | Kasim Aslam by Joe Polish, generated from the public captions YouTube serves with the video. The transcript has 8,175 words across 1,182 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.