July 28, 2025 • 52 mins
Kevin Surace, known as the father of virtual assistants, shares his insights on how AI is revolutionizing marketing and business operations. We discuss the democratization of content creation, the transformation of search, and why companies refusing to embrace AI are heading toward extinction.
• AI is making marketers 10-100x more productive by enabling rapid content creation and analysis.
• Content that previously took weeks can now be generated in minutes with equal or better quality.
• The future of SEO is uncertain as AI assistants increasingly answer queries directly without sending users to websites.
• People who use AI tools for everything are gaining massive advantages over those who are only "playing" with the technology.
• Corporate policies banning AI use are equivalent to companies refusing to use the internet in the 1990s.
• Creative professionals need to evolve from content creators to strategic partners who guide and refine AI outputs.
• Marketing agencies must position themselves as "robot overlords" rather than doers to remain relevant.
• The best implementation strategy is to start with tools like Microsoft Copilot before advancing to more sophisticated applications.
• Successful AI adopters use these tools for every aspect of their workflow, not just occasional specialized tasks.
--------
Guest Contact Information:
https://www.kevinsurace.com/
https://www.linkedin.com/in/ksurace/
—----------
More from EWR and Matt:
Leave a Review if it was content you enjoyed: https://g.page/r/CccGEk37CLosEB0/review
Free SEO Consultation: https://www.ewrdigital.com/discovery-call
One-on-One Consulting: https://www.ewrdigital.com/digital-strategy-consulting/private-consulting-session
—
The Unknown Secrets of Internet Marketing podcast is a podcast hosted by Internet marketing expert Matthew Bertram. The show provides insights and advice on digital marketing, SEO, and online business.
Topics covered include keyword research, content optimization, link building, local SEO, and more. The show also features interviews with industry leaders and experts who share their experiences and tips.
Additionally, Matt shares his own experiences and strategies, as well as his own successes and failures, to help listeners learn from his experiences and apply the same principles to their businesses. The show is designed to help entrepreneurs and business owners become successful online and get the most out of their digital marketing efforts.
Find more great episodes here: https://www.internetmarketingsecretspodcast.com/
https://seo-podcast-the-unknown-secrets-of-internet-marketing.buzzsprout.com
Follow us on:
Facebook: @bestseopodcast
Instagram: @thebestseopodcast
Tiktok: @bestseopodcast
LinkedIn: @bestseopodcast
Powered by: ewrdigital.com
Hosts: Matt Bertram
Disclaimer: For Educational and Entertainment purposes only.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:02):
This is the Unknown Secrets of Internet Marketing,
your insider guide to thestrategies top marketers use to
crush the competition.
Ready to unlock your businessfull potential, let's get
started.
Speaker 2 (00:17):
Hi, welcome back to another fun-filled episode of
the Unknown Secrets of InternetMarketing.
I am your host, Matt Bertram.
As everybody's been askingabout AI and LLMs and where the
future's going, I thought it'dbe great to bring on one of the
forefront experts in the fieldon it, I think.
Kevin, how many keynotes do youdo a year?
(00:38):
Something like 50 or 60keynotes.
Speaker 1 (00:39):
Yeah, between 40 and 50 typically.
I just got off a plane fromanother keynote.
Speaker 2 (00:45):
So we are lucky to have Kevin Serais right?
It's Kevin Serais, and he hasbeen in Applied AI.
He's been known as, like, thefather of chatbots.
Is that right, father?
Speaker 1 (01:00):
of the virtual assistant.
I invented the darn thing andit was almost 30 years ago, and
so Siri and Alexa and all ofthat had to license all of that
original technology to continueto survive.
Speaker 2 (01:11):
Well, I just feel honored to have you here and I
have some questions.
Like I said that, I had someLLMs help me create because I
thought it was appropriate, butI just want to open it up to you
of what's most topical rightnow.
What are you looking at?
What are you talking about?
Speaker 1 (01:28):
to just kind of set the table for everyone Well,
look in the AI space, andespecially if you're in
marketing and maybe you're inmarketing if you're listening to
this podcast.
When I talk to marketing people, you have everyone from I don't
really use it, I don't believein it all the way to.
I'm doing everything with it,okay, and it broadly meaning Gen
(01:51):
AI, and that could bemultimodal.
It could be analysis of papersor documents, analysis of
spreadsheets, analysis ofperformance, analysis of ROI and
CAC and LTV and things likethat Certainly, creation of
content.
But I would open with thisright, if you are not the expert
(02:15):
, the expert in doing everythingyou can in your job not playing
, but doing well then someoneelse will be the expert and you
will be gone, because the peopleusing this technology,
including me.
I use Gen AI for every singlething, right.
So I ideate, I create content,I modify the content, I analyze
(02:39):
all kinds of documents, right,and I'm easily 10 times more
productive than I was 10 yearsago and actually critically
thinking more.
Speaker 2 (02:49):
I mean, the multimodal component is
fascinating because you canupload documents, you can take
pictures of stuff, it can helpanalyze that and now it
remembers, like your previousconversation, so it's building
that kind of body of knowledge.
We actually have an AI thatanswers our phone.
So we did end up kind ofswitching over to that.
(03:10):
I kind of felt like if the bigagencies or the big marketing
companies like, let's say,google and Meta and stuff like
that, you can't get anybody onthe phone unless you like know
somebody, I was like, well, whycan't everybody else do it too?
Right, and so so we implementedthat probably a year and a half
ago, and it was really quiteeffective of capturing the right
(03:30):
information, qualifying thepeople, passing it off to the
right department, and we had to,we had to work on a little bit
to kind of polish it for peoplethat, and we had to.
Hey, we're an AI, right, like,and and that's changing Some
people are wanting to talk toAIs more than humans.
So I don't know what.
What are you?
What are you most concernedabout?
(03:53):
Like, I know, as we weretalking in the pre-interview,
there are certain industriesthat are a little slower to
adopt.
There's people that I'm talkingto in other industries that are
using it as a play thing, like,hey, I made a little video with
my kids, I made some images,pictures.
But, like you said, marketingand there's another industry
that I'm not going to name thatis really risk-taking, on the
(04:16):
cutting edge and seems to driveeverything forward, and so I'm
seeing, from an AI governanceeven standpoint from
corporations, marketing'sleading the charge.
I mean, is that somethingyou're seeing?
Speaker 1 (04:29):
Yeah, well, here's the thing.
Look, when you look at this eraof Gen AI, a lot of what it
does is generate content.
It does a lot of other thingsfinance and other things but it
easily generates content.
So an example I'm in the techindustry and if I wanted someone
(04:49):
outside to write a blog postfor me, I would hire that person
and it might be $500 to $1,500,and it might take a week or so
and they'd write my next blogpost and I go, that's it for
that week.
I'd do it again next week, anddo it again next week, right,
and we'd go back and forth, etcetera.
(05:10):
Well, today, if I really wantto write a blog post, I start
thinking about writing all 52blog posts for the year today.
Yeah.
And I can ask a modern model.
First of all, go review myentire website.
From that entire website,generate 52 possible topics for
my blog post.
Then I start going to each oneand say now, generate a thousand
(05:35):
words for each one that alignwith other things we're already
saying on the website, butpromote what we're trying to say
without being too commercial,et cetera, et cetera.
So I can set those bounds.
So, basically, at the end ofthat I've got 52 blog posts and
I have that in under an hour,maybe 30 minutes, and then after
that I'm going to spend therest of the day editing those,
cleaning them up, throwing a fewout, replacing them, et cetera.
(05:55):
By the end of the day I've doneblog posts for a year.
I didn't hire anyone.
I'm done in a day.
That is at least a thousand Ximprovement over any other way
we've ever done this.
A thousand X, that is marketchanging.
That is a complete change andanyone can do what I said.
You know, with just a few hoursof experience of understanding
(06:18):
how to ask for what you want andget it right.
We've never seen that kind ofdemocratization of content
generation, good or bad.
You actually don't need 10,000hours of marketing knowledge to
actually execute that plan, editthose things and have them
ready to go out.
You know, probably a juniorperson could do that.
(06:40):
They could do a pretty good jobat it.
Speaker 2 (06:42):
So that was, I think, one of the reasons why Google
added to the kind of frameworkof EAT expertise, authority and
trust.
They've added from theexpertise to experience, like
having experienced that layer,and they've kind of brought back
authorship of like who's sayingit right as a lawyer, a doctor,
or is it me or is it you?
Like it's going to weight thosethings differently.
(07:04):
I also think like the nextlayer is what I'm more concerned
about and where I'm focusing mytime is Now the ideation and
the cleanup phase.
Like I've seen some workflowswhere it can continue to ideate
that you can, you know, uploaddifferent documents.
It can pull in Right, I do itall the time.
Speaker 1 (07:22):
Here's five documents I've already written.
Combine them, clean it up andcut it to a thousand words.
Boom.
You know not that I can't do it, but why would I?
I wouldn't do a better job thanit can.
And, by the way, it's anenglish major phd, right, I'm
not right.
It learned literally a trillionenglish phrases.
Speaker 2 (07:41):
I think it's pretty good at it yeah, well, I'm
looking at like so the agent,like so, with these questions
that I got and we'll have to getto some of these because I
think they're pretty good I putin your website and I said go
call the website, summarize thewebsite, give me some questions,
that would be helpful.
And it broke it apart into somegreat categories and it
(08:01):
referenced what we said.
Said that agent um, you can nowuse automation attachments to
attach it to wordpress and say,okay, those blogs that you wrote
, go, ideate it.
Maybe there's a human, uh,person, checking it in the
middle, like a subject matterexpert checking it.
Okay, okay it.
And then that agent continuesto work autonomously to go post
it on the website based upon aworkflow for the next 52 weeks
(08:25):
for the next 52 weeks.
Like, or you're like, um, youknow, and now I'm seeing you
know the spam get out of controlof people putting these things
on autopilot and going.
I want to blog a day about allthese topics, on all these like
it, like what?
What are the?
The rate of content creation?
Uh, google can't even keep upwith.
It is like unindexing stuff andand we need so much compute
(08:47):
powder to do it.
Like, are there bigger issueswhen you look at it at a higher
level, of things that you'reseeing from a trend?
Speaker 1 (08:54):
well?
Well, certainly so.
So, so let's just, first of all, we've been through this
already and I'm going to use thepodcast as an example.
Okay, it used to be 20 yearsago.
There were, I don't know, 20 or30 radio broadcasts that
mattered on a national basis.
Maybe there were 10, thereweren't many, right?
But if you wanted to broadcastand be heard nationally, you had
(09:16):
to get on a national radionetwork or build a radio network
, right, an independent radionetwork, and have a syndicated,
and people did that.
So, like I said, maybe there'sa dozen, maybe there's two dozen
, it wasn't many.
In the last decade, aspodcasting became practical and
very inexpensive like for acouple hundred bucks, I've got
everything I want and it soundsgood, and it looks good and
(09:36):
everything else.
We have approximately 45million podcasters.
We've democratized the abilityto broadcast.
Now, there's a long tail onthat.
The vast majority of them haveprobably less than 100, you know
watchers or listeners a week.
And then there's the Mr Beast.
You know that's making 65million a year, right?
So we democratized access tocreate this content and push it
(10:01):
out.
Now, that didn't destroyanything.
Didn't destroy anything.
In fact, it allowed us tocreate these niches that
literally only a hundred peoplelisten to a week, but it's your
most.
You think it's very importantwhat that person is saying, or
their guests or or whatever itis.
It's fascinating.
So we have been through this.
We've like seen this happen,you could argue.
(10:21):
But we did that with video andYouTube.
Before YouTube, you know, therewas no ability to share your
videos basically anything withanyone.
Now there are millions ofvideos uploaded every day.
Every single day Doesn'toverload anything.
Google's built the servers tohandle it, right, and they're
(10:41):
indexed and they're, uh, closedcaptioned and everything else.
It's fascinating, right?
So we've seen this happenbefore.
So now we're generating animmense amount of marketing
content.
That's mostly ai generated,right, because it's just so good
and so dead internet theory is?
Speaker 2 (10:57):
I've kind of gone back to that.
Speaker 1 (10:58):
I don't know I know, you know that's dead because all
of it is just ai talking to AIand no humans care.
But?
But that would be like sayingthe dead podcast theory.
But it isn't really right.
There are 45 million podcastsand someone's listening to
virtually all of them.
Some at the very far tail it'sjust their family listens, right
, but nevertheless, that's whatit is.
(11:20):
And the same with YouTube.
I mean you could say it's thedead video theory.
There's a ton of YouTube videosthat have only been watched two
times I mean probably millions,maybe hundreds of millions and
then there's some that have beenwatched hundreds of millions of
times.
So you democratize the abilityto create this content.
It creates a long tail, butnothing seems to blow up.
(11:40):
We get used to it as humans andwe go interesting.
Now youtube is starting to seeand we're going to see a lot
more over the next two years asthis technology advances lots
and lots of ai generated videos.
Oh yeah, well, what I like aboutthat is it's still, in general,
the thought of the human.
That is, I came up with a topic, a subject, maybe a script,
(12:02):
maybe an idea, maybe whatever,and I generated a short or long
or whatever long form video orwhatever it is.
I came up with a topic asubject, maybe a script, maybe
an idea, maybe whatever and Igenerated a short or long or
whatever long form video, orwhatever it is, and I'm now
using tools that allow me tolook as good as Hollywood.
I couldn't do that before.
Right, I could look as good asan iPhone, which is incredibly
good, by the way but I couldn'tlook like Hollywood.
I couldn't do those effects.
Well, now I can.
(12:23):
I can have a rhinoceros run onthe moon if I want.
There's no problem to do that.
Hollywood's been doing that for25 years with CGI, but now
everyone can do it with AI orsoon, and so we'll have a
proliferation of that.
But still, you know, if youthink about it, there's a human
behind it somewhere.
So, yes, we're generating ablog post today.
So, yes, we're generating ablog post today.
(12:43):
And people used to generate ablog post today.
It was really hard.
You had to go around thecompany and find the next person
to write something and line itall up and manage it, and now I
can really just have AI do it,and I have to come up with some
ideas, but they've come off myown website, so it's fine.
Speaker 2 (12:58):
So, kevin, on the early adopter curve, which I
think we're kind of like in thatlike general adoption is kind
of starting to happen, thepeople that are pushing the edge
like so, like music, likeyou're heavily involved in music
AI in music.
I remember there was a guy thatI don't know if it was art or
music or something, but hebasically won the award and then
came out and said it was all AIgenerated Right, and even in
(13:21):
podcasts you got the notebook LMRight.
You can upload a document andhave two ais talk and eventually
, that's right, that's I do it.
Speaker 1 (13:28):
By the way, I use it all the time.
By the way, that is, use ofnotebook lm real quickly to
generate a podcast is amazingfor companies.
Let me tell you why.
I can take our latest pressrelease that no one's going to
read and create a 15 minutepodcast that everyone can listen
to on the way to work, and theywill.
I know I use this as an exampleall the time, so it looks like a
(13:48):
toy.
And then you go actually thisis quite valuable and in fact I
can put out other content.
That way, I can take our ownpress releases from any company
and say you might want to listento it as a podcast instead, and
these two people talk about itand they interact and the whole
thing's written, you know,scripted by AI, off of only the
data that was in that, in thatpress release.
Speaker 2 (14:09):
So, kevin, where do marketing agencies and marketers
?
Because that's mainly wholistens, and I know you, you
have a company out there thatdoes like uh it, it checks code
and and you know you're you're.
I want you to speak to some ofthat, cause we have a lot of web
developers that listen to thiscall.
A lot of agencies listen tothis call.
What do they need to be doingto prepare to continue to keep
(14:31):
pace with the changes that arehappening?
Because if it's democratizedand everybody's more productive,
it just raises the bar right,it does.
Speaker 1 (14:41):
That's right.
We talk about this a little bitin that when you use Gen AI,
you use it not to play and notto toy, but you use it as part
of your workflow all the time,and it could be with agents or
non-agent, it doesn't reallymatter.
What happens is we'remultiplying your brain by 10 or
(15:01):
20 or 100x.
So all of a sudden, we'veamplified your brain.
It's your brain, you kicked itoff.
It doesn't actually do anythingby itself.
Even agentic doesn't do it byitself.
You've set it up.
You set it up, you program it.
You say go execute these thingsand it's on my command and I
can turn you off tomorrow, right, so we've multiplied your brain
, we've amplified your brain,and and and so this is really
(15:25):
critical, because if we'veamplified Matt's brain and I'm
competing with Matt and I refuseto use this stuff Matt is now a
hundred times more productivethan I am, and the clients will
begin to expect that.
Because that's what happens theclients, the clients begin to
expect that the agencies use thestate of the art, whatever it
(15:46):
is right, whatever the state ofthe art is at that time, and and
so, look, there was a time whenthe internet didn't exist, but
marketing agencies did exist.
We go back to mad men.
We can go before mad men, right, I see the mad men post.
Yeah, so the fact of the matteris we've been marketing for a
long time.
We've been marketing beforethere were newspapers, right, I
mean this is incredible.
(16:07):
And as new technology come,look when the internet hit, if
you were a market agency thatsaid, well, we don't do that
internet stuff, you know, westill just do print ads.
Well, eventually your businessdried up, just went away, right,
so, this isn't new.
So if you're not, you know,Jenny, I's been with us a couple
of years, two and a half years.
If you are not at the forefrontof that, it's okay.
(16:29):
Your clients will find someonewho is.
Speaker 2 (16:33):
Yes, that's true, like I I think I I just got back
from a future of SEO conferenceand the stuff that I'm like no,
that's possible.
You know there's a lot of kindof roadblocks and how to figure
it out and what's the right.
You know what's the right LLMto do it.
But I mean they're they'reabout three months, four months
ahead of where I want to be andI look at that going.
(16:56):
I got to catch up Like it'sonly three or four months, but
that's.
I mean that that's a that's somuch time, like that's so much
time in this world of howeverything's progressing.
I mean Google just rolled outall this stuff with youtube and
(17:17):
I mean they can make, like youwere talking about, the videos
that hollywood can make you canmake that for 40 bucks, like you
can make scenes for 40 bucks,like you can make, uh,
commercials, that are 15 000commercials for you know, 100
bucks and I I've, I've, uh I.
Speaker 1 (17:26):
I demonstrate some of the some commercials that have
already been made fully with aiand these are for very large
companies who easily would havespent a million or more on that
commercial.
Not 15K, they would have spenta million on the commercial.
Speaker 2 (17:38):
Yeah, yeah, yeah.
Speaker 1 (17:39):
Probably 20 million placements right, or 50 million
or 100 million and they wereable to make it for
substantially less than that.
Now, they could have made itfor $40, but they hired a team
anyone.
They probably spent a hundredgrand on it instead of a million
.
But the point is no question,you are going to see Hollywood
grade commercials shot with nocameras, no sound, no lighting,
(18:03):
no actors, no actors.
I thought it was interestingwhen SAG was on strike last year
and they demanded that theyhave AI in their contract.
So if someone uses theirlikeness, they get paid.
And I'm going, guys, you'vealready missed the boat.
We're not going to use any ofyour likeness.
With all due respect, therewill be films in the future and
(18:24):
commercials and everything elsethat are made strictly with AI
characters.
And then people say to me no,no, me, no, no, we won't want to
watch those.
Really, you want to talk aboutShrek they're already.
Speaker 2 (18:35):
Well, they're already watching them, like the
influencers online, and and someof these yes, people, like some
of them, claim that, oh, I'm ahuman, but half of them are not
human and you can't tell thedifference no like you, can not
tell the difference cannot tellthe difference that the, the
technology has come along so farand so fast.
Speaker 1 (18:52):
You cannot tell the difference and it's game
changing and I think it's great.
I mean, look, it's here rightTrain left the station.
Now you brought up one thing isthe future of SEO.
Let's talk about that, becauseI think the market is in quite a
turmoil for the moment.
The turmoil is this Every day,day, more people are using ai to
(19:14):
make the recommendations forthem, whether it's a local
restaurant, a local dry cleaneror a car to buy or whatever.
It is right.
We have spent 30 yearsunderstanding seo and and yes, I
know that the google algorithmschange and we reverse engineer
them we try to figure it out.
You know we've all been throughthis right, but we understood
(19:35):
about backlinks.
We understood about how you'rechatted up on other websites.
We understood about buildingall of these other sites so that
you're referenced on the sitesLike this is the tradition of
what we do.
All of a sudden, you know thechat GPT isn't recommending you
as the best dry cleaner in townand there is no algorithm that
changes that, either quickly oreasily.
(19:56):
That's true.
What do you do with that?
So we're going to figure it out, but but the whole idea of
jumping to websites may be gone.
People may not do much atwebsites.
They may take therecommendations of the AI and
just execute on that.
Speaker 2 (20:13):
So Kev, that's actually Google like two weeks
ago just launched like all thisnew stuff and it's pushing it
towards what you talked aboutYouTube and like agentic AI and
and really the the the customerjourney is is like I don't want
to go do all the research, Idon't want to click on all the
websites, like just do it allfor on all the websites.
Just do it all for me, aggregateit, tell me what I should do,
give me okay ChatGPT like sixlinks, and it's going to be
(20:37):
different based upon the wholeconversation.
What that knows about you,that's right.
It's very, very personalized.
Speaker 1 (20:45):
Matt, it's already part of Google search now,
because they have let AI answerthe search for you the overview.
Search the traditional searchand more and more people are
choosing the AI search because,did you remember?
Well, you remember it's stilltoday.
I've got to be on the firstpage of the Google search, right
?
I got to be on Google results.
I got to be on the first page.
Got to be on the first page.
Now it doesn't matter if you'retop on the first page.
(21:05):
There is no page.
Speaker 2 (21:12):
Yeah, ai answered the question.
Well, you know, on the buyingcycle, they're going to go head
to head against amazon nowbecause, well, right,
everything's happening onyoutube or everything's.
And I think the ai overviewsare like slowing down the llms
or trying to slow down the lmsto go to a new platform.
But I'm seeing a lot of peopleswitch over that don't even go
back to google unless they'regoing to buy.
Now, right, they, they're doingall the research and ideation,
(21:33):
uh, in the LLM.
And then then they're like okay, let me go verify some stuff,
let me just look at the websiteand make the purchase.
Traffic's been cut in half,actually by 60% on average
across the board.
And the buy cycle uh, google'sjust saying just buy it now with
us, like you don't even have toleave Google, right, you don't
(21:53):
have to leave YouTube.
Like it's very disruptive to alot of people.
Speaker 1 (21:59):
Especially to you who's in this marketing game,
and you go.
Everything I've learned over 30years doesn't work anymore.
Speaker 2 (22:06):
And that's why I'm bringing people like you on to
help make sense of all this andpeople that are listening.
I mean, we've been doing thispodcast for 12 years and the
rate of change, like I have thisgraph I'm about to do a webinar
to talk about some of thisstuff but there's a standard
deviation change to 2024 overthe last 15 years, of all the
(22:26):
changes that were kind of likeone a year, like that pace sped
up.
All the changes that were kindof like one a year, like like
that that pace sped up.
And then from 2024 to 2025,it's like three standard
deviations higher of all thechanges in their testing stuff,
like daily and like everything'sgoing crazy, and even like
traffic is a big metric.
That's not a metric anymore.
It's like and almost when youget into the top positions or
(22:48):
you get into the AI overviews,people also ask your traffic
disappears because people aregetting their answer on Google.
They don't have to go to yourwebsite for it.
So it actually is like inverse.
It's really interesting.
I just did a post about that onLinkedIn is revenues up,
impressions are up, traffic'sdown, like it's.
It's wild and, and so I'mtrying to make sense of all
(23:10):
where, where this is going froma futurist standpoint, like
where's the puck going to?
Like what are the things thatwe need to be doing?
I know you have somecertifications on your site Like
what do people need to bethinking about to prepare for
this?
Because marketing is going tobe is being affected first.
Right.
Speaker 1 (23:30):
Again, we went through this change when the
internet came and all of asudden it was brand new.
There was this entire new realmthat everyone in marketing had
to learn, that knew nothingabout it the year before, yeah,
and and and.
We went from yellow pages andphone books and libraries and
and print ads and all of that tojust doing searches on Google
(23:52):
or Ping or Yahoo at the time anda few others right,
unbelievable, unbelievablechange.
We did get through that change,but from that change created
entire new capabilities inmarketing agencies.
Yep.
And you've honed it and honed it, and honed it, and now we're
going through as big of a change, probably faster.
Faster because this technologyLLMs is built already upon
(24:17):
smartphones and internet, whichalready exist.
So we're riding this third wave, but it's riding on top of the
other waves, right, and that'swhy you went from zero to 500
million users or so on ChatGPT,in, or two Um, and in fact, llms
is the fastest growing newcategory in the history of
(24:38):
humans.
So what?
Speaker 2 (24:40):
how do you like?
This is something I know thatyou talk about in your talks
like what was that confluencepoint of things coming together
in kind of the AI space for thisto just go vertical?
Speaker 1 (24:52):
right, oh sure, easy, easy to talk about that.
So, so, first of all, deeplearning.
We solved deep learning in 2012.
We could do relatively deep, afew layers deep before that, but
in 2012, because of the, thealgorithms and the math and
cloud, google published aseminal paper that said actually
(25:14):
, we can go as deep as we wantnow unlimited depths.
So that means, instead oflearning 10 words, 100 words or
phrases or 1,000 phrases, youcould learn unlimited phrases
and you could build a neural netout of that.
Right, that was seminal,seminal paper and it was put out
there for the world tounderstand.
Well, by 2017, google again hada great idea.
(25:37):
They said you know, translatingEnglish to French sucks when
you do it word by word, becausethe French sentence is mangled
right because of the pronounsand the nouns and the verbs and
they're all out of order forwhat a French sentence should be
structured like versus anEnglish sentence.
So what would happen if welearned the phrase and then we
translated phrase by phrase?
That's a transformer model.
(25:58):
They published that document in2017.
All right, that's eight yearsago.
And out of that document,people said I wonder if, instead
of translating and, by the way,that's when OpenAI was founded.
I wonder if, instead oftranslating and, by the way,
that's when OpenAI was founded,based on the transformer I
wonder if, instead oftranslating, we just learned
basically the whole Englishlanguage and then you could
(26:19):
basically ask a question.
Instead of translating, we'dcome back in English, but we
would pose the answer alignedwith your query, right?
And so OpenAI had ChatGPT 1,1.0, 1.5.
I was using it like on 2.0 orsomething, and finally, at 3.5,
they felt this is ready forhuman consumption.
(26:40):
At that point they had grown toover a trillion or so tokens,
or we could say sentences, rightor phrases, and and so at that
point it kind of had gobbled allof the English language you
could possibly gobble from theweb.
Whether it should have or notis a whole other conversation
(27:01):
copyrights, this and that, butyou know it.
It learned from everything wehumans published and all of a
sudden, people, instead ofpeople who code in python
talking to ai, anyone could talkto ai in english, in their
language actually.
Speaker 2 (27:16):
But in my code.
Speaker 1 (27:18):
All right, well, yes, but you could just talk to it
and it talks back in englishright yeah, you can vibe code
also, but forget vibe coding,you just talk to it.
So all of a sudden people, forthe first time ever, could
really talk to computers.
Now, actually, actually, when Iinvented the virtual assistant
in the late 90s, that was reallythe very first time that people
could truly talk in English tothese things.
But at this point, now youcould talk and ask it anything.
(27:40):
Then you could only ask it5,000 different things.
Now you could ask it millionsof different things, right?
So that's a turning point.
And once you did that,everybody said this is useful.
Now I can ask it anything.
Is there a God, do you love me?
Or who had the hit record in1976, that?
So all of a sudden it cananswer anything because it's
(28:03):
read everything.
It doesn't know the differencebetween fact and fiction,
although a lot of that isgetting sorted out with kind of
supervised learning posts.
So there are thousands ofpeople trying to overseas make
sure that these things areanswering correctly.
And so you know we had a lot ofhallucinations early on.
We have a lot less now.
But, that was the turning point.
Speaker 2 (28:24):
Yeah, they're citing the sources.
It's definitely getting bigger.
What do you think that nextrevolution, if you were to guess
, going to look like?
So, like all the internetbusinesses were created, like it
seems like um, uh, likelearning, like education's going
to be quite affected.
I feel like um, like it atherapist maybe, like talking it
(28:47):
to, like understand what'sgoing on with people, like are
people choosing to talk to bots?
Or like even like a LLM, likelisten to bots talk over humans?
Or like where do you see thatevolution happening?
Where that maybe the, theintersection comes, or the
integration between humans?
(29:07):
All right, let's talk about.
Speaker 1 (29:09):
Let's just pick on education first, cause you
mentioned education.
We talked about psychologists,but let's talk about education.
So I'm on the board of RIT,rochester Institute of
technology.
We've got 20,000 students thisis a very hot topic for the last
two years which is, you know,here's, here's my take on it.
And when I tell the professors,assume that your students are
(29:29):
using this for everything,period, full stop.
If you don't assume that I,what are you doing teaching?
Right, they're, they're usingit for everything.
And so how do we still teachkids, um, critical thinking
skills and and other skills,right?
But uh, in the, in this era?
Well, you know, there was atime that we gave up teaching
(29:50):
kids too much of long division,because once the calculator came
and excel came, nobody's evergoing to ask you to do long
division at work, are they ever?
In fact, if you're doing longdivision with a, a paper and
pencil, I might terminate youthat was my.
Speaker 2 (30:05):
That was my argument, kevin.
In school I was like, if I cando this with a calculator, I'm
never going to do it, not so, sowhy can't?
I use a calculator on this test.
I thought it was ridiculous.
Speaker 1 (30:13):
Right, right, because no one's going to pay you to
not use the calculator.
Speaker 2 (30:16):
All right.
Speaker 1 (30:17):
And this is so.
So now, when we, when we lookat kids in university, right,
and I've got, you know, I've gotprofessors that say, well, I
want them to handwrite the paperand do this and do that, and
what's the purpose?
We're supposed to be preparingkids for the workforce.
You know what?
You know who's going to gethired?
Okay, there's two kids thatcome out of college this year.
One comes out and says yeah,I've heard about ChatGPT, but I
(30:40):
didn't use it.
I was told I shouldn't use it.
I played with it a little bit.
You know, I don't know it well,but I can learn it, okay, nice.
Next kid comes in and says allright, I'll be honest with you,
I used it for everything.
I uploaded documents, Ianalyzed documents, I edited
(31:00):
stuff, I right every, everyassignment.
I did that way and I crushed it.
Not only that, I learned how toedit it afterwards through
other models or use threedifferent models to fool the ai
sensors yeah, yeah, theoriginality.
Speaker 2 (31:09):
I'm about to have someone on that has a plagiarism
ticker on.
Speaker 1 (31:16):
I can tell you it gets very hard to detect, right?
So let me finish.
The point is who gets hired?
Well, the second one, thesecond one, the one who knows
how to use the latest technologygets hired all the time.
It's always been that way.
Period, when the wheel came out, those who put wheels on a cart
won over those who wouldn't usethe wheel.
This is how it works okay.
Speaker 2 (31:38):
So I have a specific question from an enterprise
client that we're actuallydealing with.
Okay, we have a, a company thatgot bought by another company
and there's all these corporatelayers and a lot of, uh, red
tape.
I'm not going to mention thecompany, but they're, you know,
a publicly traded fortune,probably 100 company.
It came down that no contentcan be generated at all with AI,
(32:02):
and they were pushing back onblogs that were actually human
written, but they were saying,oh, this is like 10% AI and and
we were like, no, a human wrotethis and we we had to keep
working it in, like jigsawingthe words to try to get um less
content.
And like I'm I'm actuallydealing with this client still
(32:24):
today and they've eased off alittle bit, but it's.
It's this corporate policy fromlike governance coming down,
saying we want it all humanwritten.
This is all AI.
Like if you were talking to amajor corporation that was
presenting this as policy, whatwould you say to them?
Kevin, you're dead.
That's like yeah, dead.
Speaker 1 (32:46):
Dead, that's like saying we are in 1995, we will
not use the internet forevaluating our competition, what
?
Or for marketing what I mean.
And, by the way, some companiesprobably did that For a while.
Some, you know, a lot ofcompanies said you can't bring
your own device, you can't, youknow I mean all these rules.
Well, in the end, the trainleft the station, right, you
(33:10):
have to assume that yourcompetitor is using the heck out
of it.
And I just said that you canmultiply your brain by 100x by
leveraging these tools the rightway, by 100x.
Okay, so if your competitor ismultiplying what they're doing
by 100x, and in fact, this largecorporation, I'll tell you it's
going to be some startup thatcomes up and goes, this old
kludgy thing.
(33:31):
We're going to crush them andthey will fly by them.
And we've seen this happen overand over again.
We saw it happen in the earlydays of the internet and then we
saw it happen with mobilephones that that was kind of my,
my take on.
Speaker 2 (33:43):
I was like I need to talk to whoever's pushing this
policy to either get thischanged or we might need to like
look at who their competitor isgoing to be, like one of the
challengers, because it justdoesn't even make any sense.
Like these policies are notmaking any sense, and I would
think with like co-pilot and anda lot of people on like
Microsoft software, they wouldbe a lot more in tune with the
(34:07):
direction of the market.
Speaker 1 (34:10):
Well, you know, in every new technology there are
those who crossed the chasmearlier rather than later.
There's the early adopters.
There are and they're going tobe out front, and then there's
real laggards.
That just will not change.
But I can tell you, we've seenthis story before and the
companies that did not embracethe internet died.
(34:33):
Okay, sears, yeah, sears wasthe number one mail order
catalog in the world since 1900until the internet came.
Amazon happened.
Now, sears had the mailinginfrastructure, had the shipping
infrastructure, had thewarehouses, didn't embrace the
(34:53):
internet till very late.
They eventually sort of got onfor a little bit.
By then Amazon won.
It was over, over.
It was over walmart, almostlost, but they turned it around
and got really heavy intowalmartcom, put a ton of money,
came back gonna do fine, right.
But people did ignore theinternet and every, every
(35:13):
commerce company, every retailerwho ignored the internet died
kmart, sears, others.
Right, dad, jc penny, I meandad.
So so we've seen this storydied Kmart, sears, others, right
, dad, jcpenney, I mean dad.
So we've seen this story beforeyour company.
Who is doing this now?
And I don't know who they are,right, they're just excuse me
because they're probablywatching it.
They're just stupid, like it'sjust stupid, right?
Your competitor listen to me,look at me your competitor is
(35:38):
using LLMs and they are going toblow you away in everything
they write.
It's going to be better writingfaster, writing more poignant.
I mean, you could take thisstuff and say I want you to
design a blog post thatconvinces anyone that comes here
that they've got to look at thefollowing three things to
consider us.
And that blog post is going tobe written and you read it and
you go.
I could not have written itthis well, couldn't have done it
.
Speaker 2 (36:03):
See, kevin said it, I didn't say it.
There's a short right there.
We're going to reuse that.
I think that a lot of peopleneed to hear that.
I think a lot of people arefighting the change.
I even had a content managerthat worked for me that, like I
don't want to go this way, Iwanted to and I go.
Well, we're on two differenttrains and going in two
different directions, you know,and um it, it was unfortunate.
(36:24):
So what?
Where do you think, like if youwere to break it down, if you
were to like get down into,maybe the metadata and like
cluster it into the things thatare going to take off the
fastest?
I?
I'm seeing it in the graphicside and the video side.
I'm seeing it on the contentside.
I'm seeing people build it allthe way into their workflows
(36:44):
agents replacing certain likeautomations, like where, where
do people need to be preparing,in your eyes, from a marketing
standpoint, like what's going tochange the fastest?
What are like?
I think, like you said, thebrain, like, like your thought,
like you, knowing what goodlooks like is like one of the
(37:06):
most important things, but thenyou can amplify it in every
which way from from what?
The machine learning and theLLMs can do Um and so you should
be leveraging it every waypossible.
So, like, what is that going tolook like?
Or like, are there certaincertifications?
Like I've asked chat, what arethe certifications that I need
(37:27):
to go get?
Like, what would you recommend?
Like, where would you need tolearn?
But inside this knowledge basethere, there are still so many
different sections of thingsthat you can learn, like where
should people focus?
Speaker 1 (37:40):
or how should people look at it.
So the models started to be atthe beginning.
We're good at generating textand have gotten only better.
So we're two and a half yearsin, if you're not generating all
your text or at least having itrewritten by an LLM.
You're crazy.
So often I will write becauseit's just sort of in my mind.
(38:00):
I'll write out 600's, just sortof in my mind.
You know, I'll write out600,000 words or something and
then I'll submit it to say GBD4.0 and say rewrite this in a
more powerful way that attractscustomers better, something like
that.
And, frankly, what I get back isphenomenal and I'll still may
do some edits, but it's reallyreally good.
It's just better prose than I'mcapable of.
(38:22):
Now see, a lot of marketing.
People are not capable ofsaying that sentence.
It's better prose than I couldpossibly write.
They can't say it because theydon't feel that.
Right, this is true withillustrators also.
Okay, so we are all using Gen AIto generate illustrations,
(38:42):
using Gen AI to generateillustrations, photos, et cetera
.
Today, I use it for everything.
My keynote presentations thatare, you know, easily have
hundreds and hundreds of imagesand photos and things.
They're all AI generated.
Today, you would not know thedifference.
Even my headshots people sayI'd like a headshot.
I said, well, I have 200.
How do you have 200?
Well, ai generated 200 of themand they all look exactly like
(39:02):
me and they were generated justfrom a handful of photographs,
but I'm in different poses andI'm sideways and it's
unbelievable, right, it'sunbelievable.
I didn't need a photographerfor that.
And they're all outstanding.
Outstanding Out of the 200, Ithink there were five.
I said, eh, don't like them,but it was amazing, right?
So now you look atillustrations and I talk to
illustrators all the time andsay, well, it just can't do as
(39:22):
good as I can, okay, but in 30seconds or a minute I can have
six different illustrations.
That would have taken anillustrator days or weeks to do
the same thing and we would havehad an early sketch.
It would have gone back andforth and it would have gone
back and forth and it would havegone back color in it and blah,
blah, blah.
Whether they're using adobe,illustrator or photoshop or
(39:44):
whatever they were using to doit by hand or whatever the case
is, I don't, I don't need to dothat.
So even if an actualillustrator could somehow do a
more interesting or better job,they can't do it in 30 seconds
right.
Or for the cause, and so thatdoesn't say there's not a job
for an illustrator.
(40:05):
But if I'm an illustrator I amgoing to use the latest tools,
and when Adobe Illustratorshowed up, there were people who
said I'm still going to use apencil and crayons and other
things.
But eventually most people goton Illustrator and said I guess
that's the tool we're going touse, so we're not really going
(40:25):
to be doing much by hand anymore.
We're going to do it all on thecomputer.
Now we've got the nextgeneration that, based on your
knowledge and experience, willgenerate it for you.
Great illustrators have an eyethat I don't have.
See, I can get stuff out of GBD4.0, and I think it's great,
and someone with the rightmarketing eye or the right
illustration eye will look at itand go the colors are all wrong
for what you're trying toachieve.
(40:45):
I don't know.
I shouldn't design curtains,right.
I don't do that Right.
So that's now the illustratorbecomes very strategic and a
strategic partner to you.
It isn't about the drawing.
They're going to go andgenerate 300 of these things and
then they're going to pick thebest ones and they're going to
come back and you're going to go.
That's better than anything Icould have done so, kevin, what
I'm hearing is two things.
Speaker 2 (41:05):
Um, one is your ego is getting in the way,
potentially absolutely like.
So there's a lot of people thatare very good and they're like
I'm this good, but you know, ifyou look at human intelligence,
these ais are just way over herhead, and getting better is one.
And then the second thing isyou just said that something
very subtly that I've started tosee A strategic partner.
(41:28):
You're shifting into a role ofa partner, or an advisor, or
someone that has this knowledgeon how to best apply it right.
Because everybody has the sametoolkit now, but the craftsman
can utilize it in different waysand articulate it and knows
what they're looking for Ofcourse, because they have an eye
(41:49):
for the art that I don't have,or else I would have gone into
that field.
Speaker 1 (41:52):
So, if you're an illustrator that used to get
$1,500 in illustration, you maystill get $1,500, but you are
going to generate probably allof them via AI, but very, very
complex prompting that would bemore complex than I would know
how to do right, and leveragingthe knowledge you have in the
field, the knowledge you have ofcolors and layout and what
(42:13):
you're trying to achieve, andall of that.
I will pay for that.
But what I won't pay for is anillustrator who illustrates it
in the old way and comes backand gives me two choices and
they may or may not be good, butwhy would I pay for that?
I can go to someone who can bemuch more broad and much faster.
Now maybe the price forillustration drops from $1,500
(42:35):
to $400 or $300, but here's thething, and obviously it's free
online, but here's the thingthat allows the illustrator also
put out three or four or fivetimes more content than they
ever did.
Yes, they have to find morecustomers and yes, they have to
do more, but there's stillbusiness there, right, there'll
be money spent.
It's not about the money.
(42:56):
For me, it's about the speedand the opportunity.
Speaker 2 (43:01):
So the relationship is changing into more of a
partner of either.
I can do it Do you think it'sgoing to be more people building
internal teams with advisors,or do you think people are going
to outsource it or the split'sgoing to just be the same?
Speaker 1 (43:17):
Look, the split will be what the split will be, but
the people that I want to hirehave to be robot overlords, not
doers.
I don't want doers, I can'tafford it, it doesn't make any
sense and it takes too long.
So look at it this way I canuse Illustrator, but I'm not
really an illustrator.
So the stuff that I create inthere you can tell it's by an
amateur, so I shouldn't do itright.
(43:37):
It's not that I can't do it, sookay, I can also use a
multimodal and generate allkinds of things.
You will be able to tell notthat the quality of the
illustration or the photo is bad.
You'll be able to tell thatwhat I generated probably
doesn't fit really well with thedocument, that it's in right.
Why I'm not actually anillustrator, I'm not an artist,
(43:59):
it's not what I do.
So this is true with blog posts.
A great marketeer is going touse Gen AI and generate all
these blog posts, but thenthey're going to go in and edit
and tweak and fix and addressand you're going to tell a
professional worked with this.
Right, it's just a differentlevel than someone pushed a
(44:19):
button and stuck it out there.
Okay, podcasts you can tellwhen it's a real professional
and I've been on a lot ofpodcasts right, and there's
great podcast hosts.
There's good podcast hosts.
There's some that are horrific.
I don't know how I got on thedarn thing right, but I'm good
at you know I'm good about it Idon't and and but, but.
I mean, it's the same thing,right?
(44:41):
There's really great ones.
They're good at it, they'regreat communicators.
There's others, and you've seenthem.
You know who put that person ona podcast?
They shouldn't be there.
Speaker 2 (44:50):
Yeah, so so I have all these questions here and
we're, we're, we're runningclose to lightning round.
Yeah, um, and I I actuallybefore we do that, because it
might even be better served foranyone listening to just open it
up for you to share anythingthat we didn't cover that you
think would be valuable to talkabout, and maybe even talk about
(45:11):
some of the things you're doingand how you see those
businesses applying AI to levelup.
Speaker 1 (45:18):
One of the things I've done is I trained my own
clone and it's a video, two-wayvideo clone.
So if you go to my website,kevinsracecom, you go to the
bottom.
It says talk to Kevin's clone.
You can do a chat, you can doan audio or you can do literally
a video.
It looks like it is me, it's meand it's a clone of me, and I
trained it on all of my writings, right, and all my podcasts.
(45:42):
So it's actually pretty darngood.
So you can just ask itquestions and it's like it's
virtual Kevin, which is prettyinteresting, and the answers are
quite aligned with what I wouldactually say.
Speaker 2 (45:51):
So, kevin, I'm going to do a follow-up to this
podcast, where I'm going to goask your, your, your bot, these
questions, and then I'll add itto this.
I think it'd be hilarious, so,and I'm sure it'll be accurate
as well.
So that's really cool.
What, what are yeah, what aresome other stuff that you're
doing?
This is great.
Speaker 1 (46:06):
Well, yeah, I'm, I'm, I'm, I'm, I'm working on a
cybersecurity company calledtoken ring.
Um, it really addresses 90% ofthe ransomware problem by tying,
uh, uh, everything you log intoto your fingerprint and it's in
a ring form or in a few otherform factors, and it's proximity
, so the fingerprint has to bein proximity to the laptop that
(46:28):
you're on logging in.
It's really fun.
Another AI company where we useAI to find all of the bugs in
software.
Basically, it's very, very cool.
And it's at the UX levelsend-to-end testing, what's
called regression testing.
So there's that.
And obviously's at the UXlevels end to end testing,
what's called regression testing.
So there there, so there's that, and obviously I'm speaking in.
The one thing about speakingI'll say is when you say I've
got this big client and they'rea fortune, you know, whatever.
(46:49):
That's who I speak for, right,that's who I speak.
When I'm at there, I just just,you know, got just got done
speaking to a multi-billiondollar corporation with like 40
divisions and they had all theirleadership there and no one was
using anything in AI.
Few people were playing with it, right, and my message was okay
, guys, this train left thestation.
(47:10):
You're, you're now in thecaboose, but don't be run over
by the darn train Right.
Speaker 2 (47:14):
What do you think that those companies are looking
at implementing where like inwhat's you know what divisions
of the company like where?
Speaker 1 (47:21):
do you?
Speaker 2 (47:21):
think the applications happening the
fastest.
Speaker 1 (47:28):
First thing that gets in large companies that gets
rolled out is is usually aco-pilot by Microsoft, because
it's safe, it's efficient, it'seasy, you know what the price
per user is, and and and.
Then they, and then they starttraining people to say, any
document that you're doingshould be assisted with Copilot.
Now, obviously that's not themost advanced thing in the world
, but it's a great place tostart.
When you've got thousands ofemployees right, you're trying
(47:50):
to regulate this and be safe andet cetera.
And you've got Copilot forcoding.
You've got Copilot for Word,excel and PowerPoint.
Now you can't do the more fancythings like a GPT-4-0 can do.
So I do a lot of documentanalysis, like.
I uploaded an annual report thatwas 150 pages.
I said tell me what's workingwell at the company and what
(48:12):
isn't working well at thecompany.
Boom and boom.
I would have had to read 150pages to find that and figure it
out and look at the numbers.
I just have the answer.
I'm going into a meeting.
That's amazing, right, and save, save me three or four hours of
trying to get through thatdocument.
I no longer have to get throughit.
Read it for me and gave me thecliff notes.
Okay, that's brilliant.
Speaker 2 (48:33):
Yeah, investing I'm seeing investing in like trading
and stuff like that.
Like the analysis that thesethings can do are pretty
incredible.
It's just interesting, too,because I've asked it about some
crypto based things and it willlike give me different answers
and I'm like, well, why did yougive me that and then why did it
change?
And so, as I continue to learnabout how they're structured, it
(48:54):
helps me understand what theoutput's gonna be.
You know, let's like wrap uphere and, and you know what
would be one, if there was onething that I think you kind of
mentioned it, and if you want torephrase it, that that that's
fine too.
But, like, what is one thingthat you feel is a unknown
(49:14):
secret of internet marketingwhen, when utilizing these LLMs,
or what do people need to bethinking about it doing most?
Speaker 1 (49:22):
well, you know, the overarching thing is something I
I kind of quote yoda here and Isay play, not do, stop playing,
because I talked to moremarketing people.
Well, I played with a littlebit and I did one blog post
partially, and I said you,you've missed the entire whole
thing here.
Everything you do must beinformed by okay, and so, even
(49:47):
if you're not ready to have it,generate everything, just use it
as an ideator, right?
Um, how might I approach this?
Boom, boom, boom, boom, boom.
What are 10 things I shouldthink about when boom, boom,
boom, boom, boom.
Right, and, yes, you'reprofessional, you probably knew
six of them.
The other four you didn't thinkof.
You couldn't have right.
And and how do I be?
What are some ways I can beatthis competitor at boom, boom,
(50:10):
boom, boom, boom?
What are some ways I can havemy company show up in your llm
versus these other companies?
It will tell you.
It's not a secret.
It will tell you so.
When you not a secret, it willtell you so when you think of it
that way you go.
Why would I do anything withjust my brain?
I've got my brain plus thisthing.
It's my brain on steroids,right?
(50:31):
That's the big lesson.
Stop playing around, stop doingoccasional things.
Do everything, every email,every text.
I get an email from a customerI'm really unhappy about.
I feed it into for me GBD 4.0.
I say I really want to savethis customer.
How might I respond in email?
And what it comes back with isbrilliant.
(50:55):
I mean, the pros is brilliant,right, it's like a psychiatrist
that understands my customer,and so I may take some of it or
not take some of it.
By the way, I got this emailfrom my spouse this morning.
She's really mad at how shouldI respond?
Trust me, you are disconnectedfrom your frontal lobe.
It is not.
It is fully connected.
(51:16):
Oh, you know you're angry andgo how dare, no, no, no, no,
no's.
If you now you can tell it.
Also, I'd like a response thatsaves my marriage, or I want a
response that ends my marriage.
Speaker 2 (51:25):
It will write that I think that's a great way to end,
kevin.
Thank you so much.
What's the best way for someoneto find out more about what
you're doing?
Speaker 1 (51:35):
Yeah, kevin's racecom , you can fill out a form there.
Learn more, you can.
My LinkedIn is there.
I do answer my DMS, et cetera.
Speaker 2 (51:44):
All right, I will put all that in the show notes for
you.
That listening.
Last name is S U R a C E Kevin.
Thank you so much for coming onuntil the next time.
My name is Matt Bertram.
Bye-bye for now.
This is the Unknown Secrets of Internet Marketing,
your insider guide to thestrategies top marketers use to
crush the competition.
Ready to unlock your businessfull potential, let's get
started.
Speaker 2 (00:17):
Hi, welcome back to another fun-filled episode of
the Unknown Secrets of InternetMarketing.
I am your host, Matt Bertram.
As everybody's been askingabout AI and LLMs and where the
future's going, I thought it'dbe great to bring on one of the
forefront experts in the fieldon it, I think.
Kevin, how many keynotes do youdo a year?
(00:38):
Something like 50 or 60keynotes.
Speaker 1 (00:39):
Yeah, between 40 and 50 typically.
I just got off a plane fromanother keynote.
Speaker 2 (00:45):
So we are lucky to have Kevin Serais right?
It's Kevin Serais, and he hasbeen in Applied AI.
He's been known as, like, thefather of chatbots.
Is that right, father?
Speaker 1 (01:00):
of the virtual assistant.
I invented the darn thing andit was almost 30 years ago, and
so Siri and Alexa and all ofthat had to license all of that
original technology to continueto survive.
Speaker 2 (01:11):
Well, I just feel honored to have you here and I
have some questions.
Like I said that, I had someLLMs help me create because I
thought it was appropriate, butI just want to open it up to you
of what's most topical rightnow.
What are you looking at?
What are you talking about?
Speaker 1 (01:28):
to just kind of set the table for everyone Well,
look in the AI space, andespecially if you're in
marketing and maybe you're inmarketing if you're listening to
this podcast.
When I talk to marketing people, you have everyone from I don't
really use it, I don't believein it all the way to.
I'm doing everything with it,okay, and it broadly meaning Gen
(01:51):
AI, and that could bemultimodal.
It could be analysis of papersor documents, analysis of
spreadsheets, analysis ofperformance, analysis of ROI and
CAC and LTV and things likethat Certainly, creation of
content.
But I would open with thisright, if you are not the expert
(02:15):
, the expert in doing everythingyou can in your job not playing
, but doing well then someoneelse will be the expert and you
will be gone, because the peopleusing this technology,
including me.
I use Gen AI for every singlething, right.
So I ideate, I create content,I modify the content, I analyze
(02:39):
all kinds of documents, right,and I'm easily 10 times more
productive than I was 10 yearsago and actually critically
thinking more.
Speaker 2 (02:49):
I mean, the multimodal component is
fascinating because you canupload documents, you can take
pictures of stuff, it can helpanalyze that and now it
remembers, like your previousconversation, so it's building
that kind of body of knowledge.
We actually have an AI thatanswers our phone.
So we did end up kind ofswitching over to that.
(03:10):
I kind of felt like if the bigagencies or the big marketing
companies like, let's say,google and Meta and stuff like
that, you can't get anybody onthe phone unless you like know
somebody, I was like, well, whycan't everybody else do it too?
Right, and so so we implementedthat probably a year and a half
ago, and it was really quiteeffective of capturing the right
(03:30):
information, qualifying thepeople, passing it off to the
right department, and we had to,we had to work on a little bit
to kind of polish it for peoplethat, and we had to.
Hey, we're an AI, right, like,and and that's changing Some
people are wanting to talk toAIs more than humans.
So I don't know what.
What are you?
What are you most concernedabout?
(03:53):
Like, I know, as we weretalking in the pre-interview,
there are certain industriesthat are a little slower to
adopt.
There's people that I'm talkingto in other industries that are
using it as a play thing, like,hey, I made a little video with
my kids, I made some images,pictures.
But, like you said, marketingand there's another industry
that I'm not going to name thatis really risk-taking, on the
(04:16):
cutting edge and seems to driveeverything forward, and so I'm
seeing, from an AI governanceeven standpoint from
corporations, marketing'sleading the charge.
I mean, is that somethingyou're seeing?
Speaker 1 (04:29):
Yeah, well, here's the thing.
Look, when you look at this eraof Gen AI, a lot of what it
does is generate content.
It does a lot of other thingsfinance and other things but it
easily generates content.
So an example I'm in the techindustry and if I wanted someone
(04:49):
outside to write a blog postfor me, I would hire that person
and it might be $500 to $1,500,and it might take a week or so
and they'd write my next blogpost and I go, that's it for
that week.
I'd do it again next week, anddo it again next week, right,
and we'd go back and forth, etcetera.
(05:10):
Well, today, if I really wantto write a blog post, I start
thinking about writing all 52blog posts for the year today.
Yeah.
And I can ask a modern model.
First of all, go review myentire website.
From that entire website,generate 52 possible topics for
my blog post.
Then I start going to each oneand say now, generate a thousand
(05:35):
words for each one that alignwith other things we're already
saying on the website, butpromote what we're trying to say
without being too commercial,et cetera, et cetera.
So I can set those bounds.
So, basically, at the end ofthat I've got 52 blog posts and
I have that in under an hour,maybe 30 minutes, and then after
that I'm going to spend therest of the day editing those,
cleaning them up, throwing a fewout, replacing them, et cetera.
(05:55):
By the end of the day I've doneblog posts for a year.
I didn't hire anyone.
I'm done in a day.
That is at least a thousand Ximprovement over any other way
we've ever done this.
A thousand X, that is marketchanging.
That is a complete change andanyone can do what I said.
You know, with just a few hoursof experience of understanding
(06:18):
how to ask for what you want andget it right.
We've never seen that kind ofdemocratization of content
generation, good or bad.
You actually don't need 10,000hours of marketing knowledge to
actually execute that plan, editthose things and have them
ready to go out.
You know, probably a juniorperson could do that.
(06:40):
They could do a pretty good jobat it.
Speaker 2 (06:42):
So that was, I think, one of the reasons why Google
added to the kind of frameworkof EAT expertise, authority and
trust.
They've added from theexpertise to experience, like
having experienced that layer,and they've kind of brought back
authorship of like who's sayingit right as a lawyer, a doctor,
or is it me or is it you?
Like it's going to weight thosethings differently.
(07:04):
I also think like the nextlayer is what I'm more concerned
about and where I'm focusing mytime is Now the ideation and
the cleanup phase.
Like I've seen some workflowswhere it can continue to ideate
that you can, you know, uploaddifferent documents.
It can pull in Right, I do itall the time.
Speaker 1 (07:22):
Here's five documents I've already written.
Combine them, clean it up andcut it to a thousand words.
Boom.
You know not that I can't do it, but why would I?
I wouldn't do a better job thanit can.
And, by the way, it's anenglish major phd, right, I'm
not right.
It learned literally a trillionenglish phrases.
Speaker 2 (07:41):
I think it's pretty good at it yeah, well, I'm
looking at like so the agent,like so, with these questions
that I got and we'll have to getto some of these because I
think they're pretty good I putin your website and I said go
call the website, summarize thewebsite, give me some questions,
that would be helpful.
And it broke it apart into somegreat categories and it
(08:01):
referenced what we said.
Said that agent um, you can nowuse automation attachments to
attach it to wordpress and say,okay, those blogs that you wrote
, go, ideate it.
Maybe there's a human, uh,person, checking it in the
middle, like a subject matterexpert checking it.
Okay, okay it.
And then that agent continuesto work autonomously to go post
it on the website based upon aworkflow for the next 52 weeks
(08:25):
for the next 52 weeks.
Like, or you're like, um, youknow, and now I'm seeing you
know the spam get out of controlof people putting these things
on autopilot and going.
I want to blog a day about allthese topics, on all these like
it, like what?
What are the?
The rate of content creation?
Uh, google can't even keep upwith.
It is like unindexing stuff andand we need so much compute
(08:47):
powder to do it.
Like, are there bigger issueswhen you look at it at a higher
level, of things that you'reseeing from a trend?
Speaker 1 (08:54):
well?
Well, certainly so.
So, so let's just, first of all, we've been through this
already and I'm going to use thepodcast as an example.
Okay, it used to be 20 yearsago.
There were, I don't know, 20 or30 radio broadcasts that
mattered on a national basis.
Maybe there were 10, thereweren't many, right?
But if you wanted to broadcastand be heard nationally, you had
(09:16):
to get on a national radionetwork or build a radio network
, right, an independent radionetwork, and have a syndicated,
and people did that.
So, like I said, maybe there'sa dozen, maybe there's two dozen
, it wasn't many.
In the last decade, aspodcasting became practical and
very inexpensive like for acouple hundred bucks, I've got
everything I want and it soundsgood, and it looks good and
(09:36):
everything else.
We have approximately 45million podcasters.
We've democratized the abilityto broadcast.
Now, there's a long tail onthat.
The vast majority of them haveprobably less than 100, you know
watchers or listeners a week.
And then there's the Mr Beast.
You know that's making 65million a year, right?
So we democratized access tocreate this content and push it
(10:01):
out.
Now, that didn't destroyanything.
Didn't destroy anything.
In fact, it allowed us tocreate these niches that
literally only a hundred peoplelisten to a week, but it's your
most.
You think it's very importantwhat that person is saying, or
their guests or or whatever itis.
It's fascinating.
So we have been through this.
We've like seen this happen,you could argue.
(10:21):
But we did that with video andYouTube.
Before YouTube, you know, therewas no ability to share your
videos basically anything withanyone.
Now there are millions ofvideos uploaded every day.
Every single day Doesn'toverload anything.
Google's built the servers tohandle it, right, and they're
(10:41):
indexed and they're, uh, closedcaptioned and everything else.
It's fascinating, right?
So we've seen this happenbefore.
So now we're generating animmense amount of marketing
content.
That's mostly ai generated,right, because it's just so good
and so dead internet theory is?
Speaker 2 (10:57):
I've kind of gone back to that.
Speaker 1 (10:58):
I don't know I know, you know that's dead because all
of it is just ai talking to AIand no humans care.
But?
But that would be like sayingthe dead podcast theory.
But it isn't really right.
There are 45 million podcastsand someone's listening to
virtually all of them.
Some at the very far tail it'sjust their family listens, right
, but nevertheless, that's whatit is.
(11:20):
And the same with YouTube.
I mean you could say it's thedead video theory.
There's a ton of YouTube videosthat have only been watched two
times I mean probably millions,maybe hundreds of millions and
then there's some that have beenwatched hundreds of millions of
times.
So you democratize the abilityto create this content.
It creates a long tail, butnothing seems to blow up.
(11:40):
We get used to it as humans andwe go interesting.
Now youtube is starting to seeand we're going to see a lot
more over the next two years asthis technology advances lots
and lots of ai generated videos.
Oh yeah, well, what I like aboutthat is it's still, in general,
the thought of the human.
That is, I came up with a topic, a subject, maybe a script,
(12:02):
maybe an idea, maybe whatever,and I generated a short or long
or whatever long form video orwhatever it is.
I came up with a topic asubject, maybe a script, maybe
an idea, maybe whatever and Igenerated a short or long or
whatever long form video, orwhatever it is, and I'm now
using tools that allow me tolook as good as Hollywood.
I couldn't do that before.
Right, I could look as good asan iPhone, which is incredibly
good, by the way but I couldn'tlook like Hollywood.
I couldn't do those effects.
Well, now I can.
(12:23):
I can have a rhinoceros run onthe moon if I want.
There's no problem to do that.
Hollywood's been doing that for25 years with CGI, but now
everyone can do it with AI orsoon, and so we'll have a
proliferation of that.
But still, you know, if youthink about it, there's a human
behind it somewhere.
So, yes, we're generating ablog post today.
So, yes, we're generating ablog post today.
(12:43):
And people used to generate ablog post today.
It was really hard.
You had to go around thecompany and find the next person
to write something and line itall up and manage it, and now I
can really just have AI do it,and I have to come up with some
ideas, but they've come off myown website, so it's fine.
Speaker 2 (12:58):
So, kevin, on the early adopter curve, which I
think we're kind of like in thatlike general adoption is kind
of starting to happen, thepeople that are pushing the edge
like so, like music, likeyou're heavily involved in music
AI in music.
I remember there was a guy thatI don't know if it was art or
music or something, but hebasically won the award and then
came out and said it was all AIgenerated Right, and even in
(13:21):
podcasts you got the notebook LMRight.
You can upload a document andhave two ais talk and eventually
, that's right, that's I do it.
Speaker 1 (13:28):
By the way, I use it all the time.
By the way, that is, use ofnotebook lm real quickly to
generate a podcast is amazingfor companies.
Let me tell you why.
I can take our latest pressrelease that no one's going to
read and create a 15 minutepodcast that everyone can listen
to on the way to work, and theywill.
I know I use this as an exampleall the time, so it looks like a
(13:48):
toy.
And then you go actually thisis quite valuable and in fact I
can put out other content.
That way, I can take our ownpress releases from any company
and say you might want to listento it as a podcast instead, and
these two people talk about itand they interact and the whole
thing's written, you know,scripted by AI, off of only the
data that was in that, in thatpress release.
Speaker 2 (14:09):
So, kevin, where do marketing agencies and marketers
?
Because that's mainly wholistens, and I know you, you
have a company out there thatdoes like uh it, it checks code
and and you know you're you're.
I want you to speak to some ofthat, cause we have a lot of web
developers that listen to thiscall.
A lot of agencies listen tothis call.
What do they need to be doingto prepare to continue to keep
(14:31):
pace with the changes that arehappening?
Because if it's democratizedand everybody's more productive,
it just raises the bar right,it does.
Speaker 1 (14:41):
That's right.
We talk about this a little bitin that when you use Gen AI,
you use it not to play and notto toy, but you use it as part
of your workflow all the time,and it could be with agents or
non-agent, it doesn't reallymatter.
What happens is we'remultiplying your brain by 10 or
(15:01):
20 or 100x.
So all of a sudden, we'veamplified your brain.
It's your brain, you kicked itoff.
It doesn't actually do anythingby itself.
Even agentic doesn't do it byitself.
You've set it up.
You set it up, you program it.
You say go execute these thingsand it's on my command and I
can turn you off tomorrow, right, so we've multiplied your brain
, we've amplified your brain,and and and so this is really
(15:25):
critical, because if we'veamplified Matt's brain and I'm
competing with Matt and I refuseto use this stuff Matt is now a
hundred times more productivethan I am, and the clients will
begin to expect that.
Because that's what happens theclients, the clients begin to
expect that the agencies use thestate of the art, whatever it
(15:46):
is right, whatever the state ofthe art is at that time, and and
so, look, there was a time whenthe internet didn't exist, but
marketing agencies did exist.
We go back to mad men.
We can go before mad men, right, I see the mad men post.
Yeah, so the fact of the matteris we've been marketing for a
long time.
We've been marketing beforethere were newspapers, right, I
mean this is incredible.
(16:07):
And as new technology come,look when the internet hit, if
you were a market agency thatsaid, well, we don't do that
internet stuff, you know, westill just do print ads.
Well, eventually your businessdried up, just went away, right,
so, this isn't new.
So if you're not, you know,Jenny, I's been with us a couple
of years, two and a half years.
If you are not at the forefrontof that, it's okay.
(16:29):
Your clients will find someonewho is.
Speaker 2 (16:33):
Yes, that's true, like I I think I I just got back
from a future of SEO conferenceand the stuff that I'm like no,
that's possible.
You know there's a lot of kindof roadblocks and how to figure
it out and what's the right.
You know what's the right LLMto do it.
But I mean they're they'reabout three months, four months
ahead of where I want to be andI look at that going.
(16:56):
I got to catch up Like it'sonly three or four months, but
that's.
I mean that that's a that's somuch time, like that's so much
time in this world of howeverything's progressing.
I mean Google just rolled outall this stuff with youtube and
(17:17):
I mean they can make, like youwere talking about, the videos
that hollywood can make you canmake that for 40 bucks, like you
can make scenes for 40 bucks,like you can make, uh,
commercials, that are 15 000commercials for you know, 100
bucks and I I've, I've, uh I.
Speaker 1 (17:26):
I demonstrate some of the some commercials that have
already been made fully with aiand these are for very large
companies who easily would havespent a million or more on that
commercial.
Not 15K, they would have spenta million on the commercial.
Speaker 2 (17:38):
Yeah, yeah, yeah.
Speaker 1 (17:39):
Probably 20 million placements right, or 50 million
or 100 million and they wereable to make it for
substantially less than that.
Now, they could have made itfor $40, but they hired a team
anyone.
They probably spent a hundredgrand on it instead of a million
.
But the point is no question,you are going to see Hollywood
grade commercials shot with nocameras, no sound, no lighting,
(18:03):
no actors, no actors.
I thought it was interestingwhen SAG was on strike last year
and they demanded that theyhave AI in their contract.
So if someone uses theirlikeness, they get paid.
And I'm going, guys, you'vealready missed the boat.
We're not going to use any ofyour likeness.
With all due respect, therewill be films in the future and
(18:24):
commercials and everything elsethat are made strictly with AI
characters.
And then people say to me no,no, me, no, no, we won't want to
watch those.
Really, you want to talk aboutShrek they're already.
Speaker 2 (18:35):
Well, they're already watching them, like the
influencers online, and and someof these yes, people, like some
of them, claim that, oh, I'm ahuman, but half of them are not
human and you can't tell thedifference no like you, can not
tell the difference cannot tellthe difference that the, the
technology has come along so farand so fast.
Speaker 1 (18:52):
You cannot tell the difference and it's game
changing and I think it's great.
I mean, look, it's here rightTrain left the station.
Now you brought up one thing isthe future of SEO.
Let's talk about that, becauseI think the market is in quite a
turmoil for the moment.
The turmoil is this Every day,day, more people are using ai to
(19:14):
make the recommendations forthem, whether it's a local
restaurant, a local dry cleaneror a car to buy or whatever.
It is right.
We have spent 30 yearsunderstanding seo and and yes, I
know that the google algorithmschange and we reverse engineer
them we try to figure it out.
You know we've all been throughthis right, but we understood
(19:35):
about backlinks.
We understood about how you'rechatted up on other websites.
We understood about buildingall of these other sites so that
you're referenced on the sitesLike this is the tradition of
what we do.
All of a sudden, you know thechat GPT isn't recommending you
as the best dry cleaner in townand there is no algorithm that
changes that, either quickly oreasily.
(19:56):
That's true.
What do you do with that?
So we're going to figure it out, but but the whole idea of
jumping to websites may be gone.
People may not do much atwebsites.
They may take therecommendations of the AI and
just execute on that.
Speaker 2 (20:13):
So Kev, that's actually Google like two weeks
ago just launched like all thisnew stuff and it's pushing it
towards what you talked aboutYouTube and like agentic AI and
and really the the the customerjourney is is like I don't want
to go do all the research, Idon't want to click on all the
websites, like just do it allfor on all the websites.
Just do it all for me, aggregateit, tell me what I should do,
give me okay ChatGPT like sixlinks, and it's going to be
(20:37):
different based upon the wholeconversation.
What that knows about you,that's right.
It's very, very personalized.
Speaker 1 (20:45):
Matt, it's already part of Google search now,
because they have let AI answerthe search for you the overview.
Search the traditional searchand more and more people are
choosing the AI search because,did you remember?
Well, you remember it's stilltoday.
I've got to be on the firstpage of the Google search, right
?
I got to be on Google results.
I got to be on the first page.
Got to be on the first page.
Now it doesn't matter if you'retop on the first page.
(21:05):
There is no page.
Speaker 2 (21:12):
Yeah, ai answered the question.
Well, you know, on the buyingcycle, they're going to go head
to head against amazon nowbecause, well, right,
everything's happening onyoutube or everything's.
And I think the ai overviewsare like slowing down the llms
or trying to slow down the lmsto go to a new platform.
But I'm seeing a lot of peopleswitch over that don't even go
back to google unless they'regoing to buy.
Now, right, they, they're doingall the research and ideation,
(21:33):
uh, in the LLM.
And then then they're like okay, let me go verify some stuff,
let me just look at the websiteand make the purchase.
Traffic's been cut in half,actually by 60% on average
across the board.
And the buy cycle uh, google'sjust saying just buy it now with
us, like you don't even have toleave Google, right, you don't
(21:53):
have to leave YouTube.
Like it's very disruptive to alot of people.
Speaker 1 (21:59):
Especially to you who's in this marketing game,
and you go.
Everything I've learned over 30years doesn't work anymore.
Speaker 2 (22:06):
And that's why I'm bringing people like you on to
help make sense of all this andpeople that are listening.
I mean, we've been doing thispodcast for 12 years and the
rate of change, like I have thisgraph I'm about to do a webinar
to talk about some of thisstuff but there's a standard
deviation change to 2024 overthe last 15 years, of all the
(22:26):
changes that were kind of likeone a year, like that pace sped
up.
All the changes that were kindof like one a year, like like
that that pace sped up.
And then from 2024 to 2025,it's like three standard
deviations higher of all thechanges in their testing stuff,
like daily and like everything'sgoing crazy, and even like
traffic is a big metric.
That's not a metric anymore.
It's like and almost when youget into the top positions or
(22:48):
you get into the AI overviews,people also ask your traffic
disappears because people aregetting their answer on Google.
They don't have to go to yourwebsite for it.
So it actually is like inverse.
It's really interesting.
I just did a post about that onLinkedIn is revenues up,
impressions are up, traffic'sdown, like it's.
It's wild and, and so I'mtrying to make sense of all
(23:10):
where, where this is going froma futurist standpoint, like
where's the puck going to?
Like what are the things thatwe need to be doing?
I know you have somecertifications on your site Like
what do people need to bethinking about to prepare for
this?
Because marketing is going tobe is being affected first.
Right.
Speaker 1 (23:30):
Again, we went through this change when the
internet came and all of asudden it was brand new.
There was this entire new realmthat everyone in marketing had
to learn, that knew nothingabout it the year before, yeah,
and and and.
We went from yellow pages andphone books and libraries and
and print ads and all of that tojust doing searches on Google
(23:52):
or Ping or Yahoo at the time anda few others right,
unbelievable, unbelievablechange.
We did get through that change,but from that change created
entire new capabilities inmarketing agencies.
Yep.
And you've honed it and honed it, and honed it, and now we're
going through as big of a change, probably faster.
Faster because this technologyLLMs is built already upon
(24:17):
smartphones and internet, whichalready exist.
So we're riding this third wave, but it's riding on top of the
other waves, right, and that'swhy you went from zero to 500
million users or so on ChatGPT,in, or two Um, and in fact, llms
is the fastest growing newcategory in the history of
(24:38):
humans.
So what?
Speaker 2 (24:40):
how do you like?
This is something I know thatyou talk about in your talks
like what was that confluencepoint of things coming together
in kind of the AI space for thisto just go vertical?
Speaker 1 (24:52):
right, oh sure, easy, easy to talk about that.
So, so, first of all, deeplearning.
We solved deep learning in 2012.
We could do relatively deep, afew layers deep before that, but
in 2012, because of the, thealgorithms and the math and
cloud, google published aseminal paper that said actually
(25:14):
, we can go as deep as we wantnow unlimited depths.
So that means, instead oflearning 10 words, 100 words or
phrases or 1,000 phrases, youcould learn unlimited phrases
and you could build a neural netout of that.
Right, that was seminal,seminal paper and it was put out
there for the world tounderstand.
Well, by 2017, google again hada great idea.
(25:37):
They said you know, translatingEnglish to French sucks when
you do it word by word, becausethe French sentence is mangled
right because of the pronounsand the nouns and the verbs and
they're all out of order forwhat a French sentence should be
structured like versus anEnglish sentence.
So what would happen if welearned the phrase and then we
translated phrase by phrase?
That's a transformer model.
(25:58):
They published that document in2017.
All right, that's eight yearsago.
And out of that document,people said I wonder if, instead
of translating and, by the way,that's when OpenAI was founded.
I wonder if, instead oftranslating and, by the way,
that's when OpenAI was founded,based on the transformer I
wonder if, instead oftranslating, we just learned
basically the whole Englishlanguage and then you could
(26:19):
basically ask a question.
Instead of translating, we'dcome back in English, but we
would pose the answer alignedwith your query, right?
And so OpenAI had ChatGPT 1,1.0, 1.5.
I was using it like on 2.0 orsomething, and finally, at 3.5,
they felt this is ready forhuman consumption.
(26:40):
At that point they had grown toover a trillion or so tokens,
or we could say sentences, rightor phrases, and and so at that
point it kind of had gobbled allof the English language you
could possibly gobble from theweb.
Whether it should have or notis a whole other conversation
(27:01):
copyrights, this and that, butyou know it.
It learned from everything wehumans published and all of a
sudden, people, instead ofpeople who code in python
talking to ai, anyone could talkto ai in english, in their
language actually.
Speaker 2 (27:16):
But in my code.
Speaker 1 (27:18):
All right, well, yes, but you could just talk to it
and it talks back in englishright yeah, you can vibe code
also, but forget vibe coding,you just talk to it.
So all of a sudden people, forthe first time ever, could
really talk to computers.
Now, actually, actually, when Iinvented the virtual assistant
in the late 90s, that was reallythe very first time that people
could truly talk in English tothese things.
But at this point, now youcould talk and ask it anything.
(27:40):
Then you could only ask it5,000 different things.
Now you could ask it millionsof different things, right?
So that's a turning point.
And once you did that,everybody said this is useful.
Now I can ask it anything.
Is there a God, do you love me?
Or who had the hit record in1976, that?
So all of a sudden it cananswer anything because it's
(28:03):
read everything.
It doesn't know the differencebetween fact and fiction,
although a lot of that isgetting sorted out with kind of
supervised learning posts.
So there are thousands ofpeople trying to overseas make
sure that these things areanswering correctly.
And so you know we had a lot ofhallucinations early on.
We have a lot less now.
But, that was the turning point.
Speaker 2 (28:24):
Yeah, they're citing the sources.
It's definitely getting bigger.
What do you think that nextrevolution, if you were to guess
, going to look like?
So, like all the internetbusinesses were created, like it
seems like um, uh, likelearning, like education's going
to be quite affected.
I feel like um, like it atherapist maybe, like talking it
(28:47):
to, like understand what'sgoing on with people, like are
people choosing to talk to bots?
Or like even like a LLM, likelisten to bots talk over humans?
Or like where do you see thatevolution happening?
Where that maybe the, theintersection comes, or the
integration between humans?
(29:07):
All right, let's talk about.
Speaker 1 (29:09):
Let's just pick on education first, cause you
mentioned education.
We talked about psychologists,but let's talk about education.
So I'm on the board of RIT,rochester Institute of
technology.
We've got 20,000 students thisis a very hot topic for the last
two years which is, you know,here's, here's my take on it.
And when I tell the professors,assume that your students are
(29:29):
using this for everything,period, full stop.
If you don't assume that I,what are you doing teaching?
Right, they're, they're usingit for everything.
And so how do we still teachkids, um, critical thinking
skills and and other skills,right?
But uh, in the, in this era?
Well, you know, there was atime that we gave up teaching
(29:50):
kids too much of long division,because once the calculator came
and excel came, nobody's evergoing to ask you to do long
division at work, are they ever?
In fact, if you're doing longdivision with a, a paper and
pencil, I might terminate youthat was my.
Speaker 2 (30:05):
That was my argument, kevin.
In school I was like, if I cando this with a calculator, I'm
never going to do it, not so, sowhy can't?
I use a calculator on this test.
I thought it was ridiculous.
Speaker 1 (30:13):
Right, right, because no one's going to pay you to
not use the calculator.
Speaker 2 (30:16):
All right.
Speaker 1 (30:17):
And this is so.
So now, when we, when we lookat kids in university, right,
and I've got, you know, I've gotprofessors that say, well, I
want them to handwrite the paperand do this and do that, and
what's the purpose?
We're supposed to be preparingkids for the workforce.
You know what?
You know who's going to gethired?
Okay, there's two kids thatcome out of college this year.
One comes out and says yeah,I've heard about ChatGPT, but I
(30:40):
didn't use it.
I was told I shouldn't use it.
I played with it a little bit.
You know, I don't know it well,but I can learn it, okay, nice.
Next kid comes in and says allright, I'll be honest with you,
I used it for everything.
I uploaded documents, Ianalyzed documents, I edited
(31:00):
stuff, I right every, everyassignment.
I did that way and I crushed it.
Not only that, I learned how toedit it afterwards through
other models or use threedifferent models to fool the ai
sensors yeah, yeah, theoriginality.
Speaker 2 (31:09):
I'm about to have someone on that has a plagiarism
ticker on.
Speaker 1 (31:16):
I can tell you it gets very hard to detect, right?
So let me finish.
The point is who gets hired?
Well, the second one, thesecond one, the one who knows
how to use the latest technologygets hired all the time.
It's always been that way.
Period, when the wheel came out, those who put wheels on a cart
won over those who wouldn't usethe wheel.
This is how it works okay.
Speaker 2 (31:38):
So I have a specific question from an enterprise
client that we're actuallydealing with.
Okay, we have a, a company thatgot bought by another company
and there's all these corporatelayers and a lot of, uh, red
tape.
I'm not going to mention thecompany, but they're, you know,
a publicly traded fortune,probably 100 company.
It came down that no contentcan be generated at all with AI,
(32:02):
and they were pushing back onblogs that were actually human
written, but they were saying,oh, this is like 10% AI and and
we were like, no, a human wrotethis and we we had to keep
working it in, like jigsawingthe words to try to get um less
content.
And like I'm I'm actuallydealing with this client still
(32:24):
today and they've eased off alittle bit, but it's.
It's this corporate policy fromlike governance coming down,
saying we want it all humanwritten.
This is all AI.
Like if you were talking to amajor corporation that was
presenting this as policy, whatwould you say to them?
Kevin, you're dead.
That's like yeah, dead.
Speaker 1 (32:46):
Dead, that's like saying we are in 1995, we will
not use the internet forevaluating our competition, what
?
Or for marketing what I mean.
And, by the way, some companiesprobably did that For a while.
Some, you know, a lot ofcompanies said you can't bring
your own device, you can't, youknow I mean all these rules.
Well, in the end, the trainleft the station, right, you
(33:10):
have to assume that yourcompetitor is using the heck out
of it.
And I just said that you canmultiply your brain by 100x by
leveraging these tools the rightway, by 100x.
Okay, so if your competitor ismultiplying what they're doing
by 100x, and in fact, this largecorporation, I'll tell you it's
going to be some startup thatcomes up and goes, this old
kludgy thing.
(33:31):
We're going to crush them andthey will fly by them.
And we've seen this happen overand over again.
We saw it happen in the earlydays of the internet and then we
saw it happen with mobilephones that that was kind of my,
my take on.
Speaker 2 (33:43):
I was like I need to talk to whoever's pushing this
policy to either get thischanged or we might need to like
look at who their competitor isgoing to be, like one of the
challengers, because it justdoesn't even make any sense.
Like these policies are notmaking any sense, and I would
think with like co-pilot and anda lot of people on like
Microsoft software, they wouldbe a lot more in tune with the
(34:07):
direction of the market.
Speaker 1 (34:10):
Well, you know, in every new technology there are
those who crossed the chasmearlier rather than later.
There's the early adopters.
There are and they're going tobe out front, and then there's
real laggards.
That just will not change.
But I can tell you, we've seenthis story before and the
companies that did not embracethe internet died.
(34:33):
Okay, sears, yeah, sears wasthe number one mail order
catalog in the world since 1900until the internet came.
Amazon happened.
Now, sears had the mailinginfrastructure, had the shipping
infrastructure, had thewarehouses, didn't embrace the
(34:53):
internet till very late.
They eventually sort of got onfor a little bit.
By then Amazon won.
It was over, over.
It was over walmart, almostlost, but they turned it around
and got really heavy intowalmartcom, put a ton of money,
came back gonna do fine, right.
But people did ignore theinternet and every, every
(35:13):
commerce company, every retailerwho ignored the internet died
kmart, sears, others.
Right, dad, jc penny, I meandad.
So so we've seen this storydied Kmart, sears, others, right
, dad, jcpenney, I mean dad.
So we've seen this story beforeyour company.
Who is doing this now?
And I don't know who they are,right, they're just excuse me
because they're probablywatching it.
They're just stupid, like it'sjust stupid, right?
Your competitor listen to me,look at me your competitor is
(35:38):
using LLMs and they are going toblow you away in everything
they write.
It's going to be better writingfaster, writing more poignant.
I mean, you could take thisstuff and say I want you to
design a blog post thatconvinces anyone that comes here
that they've got to look at thefollowing three things to
consider us.
And that blog post is going tobe written and you read it and
you go.
I could not have written itthis well, couldn't have done it
.
Speaker 2 (36:03):
See, kevin said it, I didn't say it.
There's a short right there.
We're going to reuse that.
I think that a lot of peopleneed to hear that.
I think a lot of people arefighting the change.
I even had a content managerthat worked for me that, like I
don't want to go this way, Iwanted to and I go.
Well, we're on two differenttrains and going in two
different directions, you know,and um it, it was unfortunate.
(36:24):
So what?
Where do you think, like if youwere to break it down, if you
were to like get down into,maybe the metadata and like
cluster it into the things thatare going to take off the
fastest?
I?
I'm seeing it in the graphicside and the video side.
I'm seeing it on the contentside.
I'm seeing people build it allthe way into their workflows
(36:44):
agents replacing certain likeautomations, like where, where
do people need to be preparing,in your eyes, from a marketing
standpoint, like what's going tochange the fastest?
What are like?
I think, like you said, thebrain, like, like your thought,
like you, knowing what goodlooks like is like one of the
(37:06):
most important things, but thenyou can amplify it in every
which way from from what?
The machine learning and theLLMs can do Um and so you should
be leveraging it every waypossible.
So, like, what is that going tolook like?
Or like, are there certaincertifications?
Like I've asked chat, what arethe certifications that I need
(37:27):
to go get?
Like, what would you recommend?
Like, where would you need tolearn?
But inside this knowledge basethere, there are still so many
different sections of thingsthat you can learn, like where
should people focus?
Speaker 1 (37:40):
or how should people look at it.
So the models started to be atthe beginning.
We're good at generating textand have gotten only better.
So we're two and a half yearsin, if you're not generating all
your text or at least having itrewritten by an LLM.
You're crazy.
So often I will write becauseit's just sort of in my mind.
(38:00):
I'll write out 600's, just sortof in my mind.
You know, I'll write out600,000 words or something and
then I'll submit it to say GBD4.0 and say rewrite this in a
more powerful way that attractscustomers better, something like
that.
And, frankly, what I get back isphenomenal and I'll still may
do some edits, but it's reallyreally good.
It's just better prose than I'mcapable of.
(38:22):
Now see, a lot of marketing.
People are not capable ofsaying that sentence.
It's better prose than I couldpossibly write.
They can't say it because theydon't feel that.
Right, this is true withillustrators also.
Okay, so we are all using Gen AIto generate illustrations,
(38:42):
using Gen AI to generateillustrations, photos, et cetera
.
Today, I use it for everything.
My keynote presentations thatare, you know, easily have
hundreds and hundreds of imagesand photos and things.
They're all AI generated.
Today, you would not know thedifference.
Even my headshots people sayI'd like a headshot.
I said, well, I have 200.
How do you have 200?
Well, ai generated 200 of themand they all look exactly like
(39:02):
me and they were generated justfrom a handful of photographs,
but I'm in different poses andI'm sideways and it's
unbelievable, right, it'sunbelievable.
I didn't need a photographerfor that.
And they're all outstanding.
Outstanding Out of the 200, Ithink there were five.
I said, eh, don't like them,but it was amazing, right?
So now you look atillustrations and I talk to
illustrators all the time andsay, well, it just can't do as
(39:22):
good as I can, okay, but in 30seconds or a minute I can have
six different illustrations.
That would have taken anillustrator days or weeks to do
the same thing and we would havehad an early sketch.
It would have gone back andforth and it would have gone
back and forth and it would havegone back color in it and blah,
blah, blah.
Whether they're using adobe,illustrator or photoshop or
(39:44):
whatever they were using to doit by hand or whatever the case
is, I don't, I don't need to dothat.
So even if an actualillustrator could somehow do a
more interesting or better job,they can't do it in 30 seconds
right.
Or for the cause, and so thatdoesn't say there's not a job
for an illustrator.
(40:05):
But if I'm an illustrator I amgoing to use the latest tools,
and when Adobe Illustratorshowed up, there were people who
said I'm still going to use apencil and crayons and other
things.
But eventually most people goton Illustrator and said I guess
that's the tool we're going touse, so we're not really going
(40:25):
to be doing much by hand anymore.
We're going to do it all on thecomputer.
Now we've got the nextgeneration that, based on your
knowledge and experience, willgenerate it for you.
Great illustrators have an eyethat I don't have.
See, I can get stuff out of GBD4.0, and I think it's great,
and someone with the rightmarketing eye or the right
illustration eye will look at itand go the colors are all wrong
for what you're trying toachieve.
(40:45):
I don't know.
I shouldn't design curtains,right.
I don't do that Right.
So that's now the illustratorbecomes very strategic and a
strategic partner to you.
It isn't about the drawing.
They're going to go andgenerate 300 of these things and
then they're going to pick thebest ones and they're going to
come back and you're going to go.
That's better than anything Icould have done so, kevin, what
I'm hearing is two things.
Speaker 2 (41:05):
Um, one is your ego is getting in the way,
potentially absolutely like.
So there's a lot of people thatare very good and they're like
I'm this good, but you know, ifyou look at human intelligence,
these ais are just way over herhead, and getting better is one.
And then the second thing isyou just said that something
very subtly that I've started tosee A strategic partner.
(41:28):
You're shifting into a role ofa partner, or an advisor, or
someone that has this knowledgeon how to best apply it right.
Because everybody has the sametoolkit now, but the craftsman
can utilize it in different waysand articulate it and knows
what they're looking for Ofcourse, because they have an eye
(41:49):
for the art that I don't have,or else I would have gone into
that field.
Speaker 1 (41:52):
So, if you're an illustrator that used to get
$1,500 in illustration, you maystill get $1,500, but you are
going to generate probably allof them via AI, but very, very
complex prompting that would bemore complex than I would know
how to do right, and leveragingthe knowledge you have in the
field, the knowledge you have ofcolors and layout and what
(42:13):
you're trying to achieve, andall of that.
I will pay for that.
But what I won't pay for is anillustrator who illustrates it
in the old way and comes backand gives me two choices and
they may or may not be good, butwhy would I pay for that?
I can go to someone who can bemuch more broad and much faster.
Now maybe the price forillustration drops from $1,500
(42:35):
to $400 or $300, but here's thething, and obviously it's free
online, but here's the thingthat allows the illustrator also
put out three or four or fivetimes more content than they
ever did.
Yes, they have to find morecustomers and yes, they have to
do more, but there's stillbusiness there, right, there'll
be money spent.
It's not about the money.
(42:56):
For me, it's about the speedand the opportunity.
Speaker 2 (43:01):
So the relationship is changing into more of a
partner of either.
I can do it Do you think it'sgoing to be more people building
internal teams with advisors,or do you think people are going
to outsource it or the split'sgoing to just be the same?
Speaker 1 (43:17):
Look, the split will be what the split will be, but
the people that I want to hirehave to be robot overlords, not
doers.
I don't want doers, I can'tafford it, it doesn't make any
sense and it takes too long.
So look at it this way I canuse Illustrator, but I'm not
really an illustrator.
So the stuff that I create inthere you can tell it's by an
amateur, so I shouldn't do itright.
(43:37):
It's not that I can't do it, sookay, I can also use a
multimodal and generate allkinds of things.
You will be able to tell notthat the quality of the
illustration or the photo is bad.
You'll be able to tell thatwhat I generated probably
doesn't fit really well with thedocument, that it's in right.
Why I'm not actually anillustrator, I'm not an artist,
(43:59):
it's not what I do.
So this is true with blog posts.
A great marketeer is going touse Gen AI and generate all
these blog posts, but thenthey're going to go in and edit
and tweak and fix and addressand you're going to tell a
professional worked with this.
Right, it's just a differentlevel than someone pushed a
(44:19):
button and stuck it out there.
Okay, podcasts you can tellwhen it's a real professional
and I've been on a lot ofpodcasts right, and there's
great podcast hosts.
There's good podcast hosts.
There's some that are horrific.
I don't know how I got on thedarn thing right, but I'm good
at you know I'm good about it Idon't and and but, but.
I mean, it's the same thing,right?
(44:41):
There's really great ones.
They're good at it, they'regreat communicators.
There's others, and you've seenthem.
You know who put that person ona podcast?
They shouldn't be there.
Speaker 2 (44:50):
Yeah, so so I have all these questions here and
we're, we're, we're runningclose to lightning round.
Yeah, um, and I I actuallybefore we do that, because it
might even be better served foranyone listening to just open it
up for you to share anythingthat we didn't cover that you
think would be valuable to talkabout, and maybe even talk about
(45:11):
some of the things you're doingand how you see those
businesses applying AI to levelup.
Speaker 1 (45:18):
One of the things I've done is I trained my own
clone and it's a video, two-wayvideo clone.
So if you go to my website,kevinsracecom, you go to the
bottom.
It says talk to Kevin's clone.
You can do a chat, you can doan audio or you can do literally
a video.
It looks like it is me, it's meand it's a clone of me, and I
trained it on all of my writings, right, and all my podcasts.
(45:42):
So it's actually pretty darngood.
So you can just ask itquestions and it's like it's
virtual Kevin, which is prettyinteresting, and the answers are
quite aligned with what I wouldactually say.
Speaker 2 (45:51):
So, kevin, I'm going to do a follow-up to this
podcast, where I'm going to goask your, your, your bot, these
questions, and then I'll add itto this.
I think it'd be hilarious, so,and I'm sure it'll be accurate
as well.
So that's really cool.
What, what are yeah, what aresome other stuff that you're
doing?
This is great.
Speaker 1 (46:06):
Well, yeah, I'm, I'm, I'm, I'm, I'm working on a
cybersecurity company calledtoken ring.
Um, it really addresses 90% ofthe ransomware problem by tying,
uh, uh, everything you log intoto your fingerprint and it's in
a ring form or in a few otherform factors, and it's proximity
, so the fingerprint has to bein proximity to the laptop that
(46:28):
you're on logging in.
It's really fun.
Another AI company where we useAI to find all of the bugs in
software.
Basically, it's very, very cool.
And it's at the UX levelsend-to-end testing, what's
called regression testing.
So there's that.
And obviously's at the UXlevels end to end testing,
what's called regression testing.
So there there, so there's that, and obviously I'm speaking in.
The one thing about speakingI'll say is when you say I've
got this big client and they'rea fortune, you know, whatever.
(46:49):
That's who I speak for, right,that's who I speak.
When I'm at there, I just just,you know, got just got done
speaking to a multi-billiondollar corporation with like 40
divisions and they had all theirleadership there and no one was
using anything in AI.
Few people were playing with it, right, and my message was okay
, guys, this train left thestation.
(47:10):
You're, you're now in thecaboose, but don't be run over
by the darn train Right.
Speaker 2 (47:14):
What do you think that those companies are looking
at implementing where like inwhat's you know what divisions
of the company like where?
Speaker 1 (47:21):
do you?
Speaker 2 (47:21):
think the applications happening the
fastest.
Speaker 1 (47:28):
First thing that gets in large companies that gets
rolled out is is usually aco-pilot by Microsoft, because
it's safe, it's efficient, it'seasy, you know what the price
per user is, and and and.
Then they, and then they starttraining people to say, any
document that you're doingshould be assisted with Copilot.
Now, obviously that's not themost advanced thing in the world
, but it's a great place tostart.
When you've got thousands ofemployees right, you're trying
(47:50):
to regulate this and be safe andet cetera.
And you've got Copilot forcoding.
You've got Copilot for Word,excel and PowerPoint.
Now you can't do the more fancythings like a GPT-4-0 can do.
So I do a lot of documentanalysis, like.
I uploaded an annual report thatwas 150 pages.
I said tell me what's workingwell at the company and what
(48:12):
isn't working well at thecompany.
Boom and boom.
I would have had to read 150pages to find that and figure it
out and look at the numbers.
I just have the answer.
I'm going into a meeting.
That's amazing, right, and save, save me three or four hours of
trying to get through thatdocument.
I no longer have to get throughit.
Read it for me and gave me thecliff notes.
Okay, that's brilliant.
Speaker 2 (48:33):
Yeah, investing I'm seeing investing in like trading
and stuff like that.
Like the analysis that thesethings can do are pretty
incredible.
It's just interesting, too,because I've asked it about some
crypto based things and it willlike give me different answers
and I'm like, well, why did yougive me that and then why did it
change?
And so, as I continue to learnabout how they're structured, it
(48:54):
helps me understand what theoutput's gonna be.
You know, let's like wrap uphere and, and you know what
would be one, if there was onething that I think you kind of
mentioned it, and if you want torephrase it, that that that's
fine too.
But, like, what is one thingthat you feel is a unknown
(49:14):
secret of internet marketingwhen, when utilizing these LLMs,
or what do people need to bethinking about it doing most?
Speaker 1 (49:22):
well, you know, the overarching thing is something I
I kind of quote yoda here and Isay play, not do, stop playing,
because I talked to moremarketing people.
Well, I played with a littlebit and I did one blog post
partially, and I said you,you've missed the entire whole
thing here.
Everything you do must beinformed by okay, and so, even
(49:47):
if you're not ready to have it,generate everything, just use it
as an ideator, right?
Um, how might I approach this?
Boom, boom, boom, boom, boom.
What are 10 things I shouldthink about when boom, boom,
boom, boom, boom.
Right, and, yes, you'reprofessional, you probably knew
six of them.
The other four you didn't thinkof.
You couldn't have right.
And and how do I be?
What are some ways I can beatthis competitor at boom, boom,
(50:10):
boom, boom, boom?
What are some ways I can havemy company show up in your llm
versus these other companies?
It will tell you.
It's not a secret.
It will tell you so.
When you not a secret, it willtell you so when you think of it
that way you go.
Why would I do anything withjust my brain?
I've got my brain plus thisthing.
It's my brain on steroids,right?
(50:31):
That's the big lesson.
Stop playing around, stop doingoccasional things.
Do everything, every email,every text.
I get an email from a customerI'm really unhappy about.
I feed it into for me GBD 4.0.
I say I really want to savethis customer.
How might I respond in email?
And what it comes back with isbrilliant.
(50:55):
I mean, the pros is brilliant,right, it's like a psychiatrist
that understands my customer,and so I may take some of it or
not take some of it.
By the way, I got this emailfrom my spouse this morning.
She's really mad at how shouldI respond?
Trust me, you are disconnectedfrom your frontal lobe.
It is not.
It is fully connected.
(51:16):
Oh, you know you're angry andgo how dare, no, no, no, no,
no's.
If you now you can tell it.
Also, I'd like a response thatsaves my marriage, or I want a
response that ends my marriage.
Speaker 2 (51:25):
It will write that I think that's a great way to end,
kevin.
Thank you so much.
What's the best way for someoneto find out more about what
you're doing?
Speaker 1 (51:35):
Yeah, kevin's racecom , you can fill out a form there.
Learn more, you can.
My LinkedIn is there.
I do answer my DMS, et cetera.
Speaker 2 (51:44):
All right, I will put all that in the show notes for
you.
That listening.
Last name is S U R a C E Kevin.
Thank you so much for coming onuntil the next time.
My name is Matt Bertram.
Bye-bye for now.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.