Bad Data Causes More AI Failures Than Models Do with Richard Valentine

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
SPEAKER_00 (00:01):
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_02 (00:17):
Howdy, welcome back to another fun filled episode of
The Unknown Secrets of InternetMarketing.
I am your host, Matt Bertram.
I just changed my uh input forthe mic, so hopefully you can
still hear me.
Um I am excited about wherewe're taking this conversation.
I am thinking that thetransition that we really

(00:37):
haven't made to YouTube thisepisode might bring about that
shift.
We are up on YouTube.
We are going to be changing theformat.
Uh, I do have a whiteboardbehind me.
We are going to be doing moreeducation.
I do need to change the intro.
Probably need to change the nameof the podcast.
Um, I'm sorry, guys.
I've been super busy.

(00:58):
Uh we've been doing a lot ofdevelopment work.
Uh, we've we've been launching anumber of things.
I've had a lot going on.
I do have another podcast outthere, maybe launching another
one.
But what I'm really thinking ismaybe rolling them all up into
one because I think it'sbecoming uh like AI all the
time.
Um, and it's just eatingeverything.

(01:18):
AI is just absolutely eatingeverything.
And we were talking a littlebit, and I've hented at uh
persistent memory.
And like as you move the shiftfrom like a chat bot to like a
teammate, um, I I've beenreaching out to some people in
my network, and uh I got one ofthe researchers that's really,
really up there uh with AI andmemory to come on.

(01:41):
He has a couple of provisionalpatents, he's published some
white papers, uh, he's uhdedicated some time teaching me
some things, and so I wanted toshare him with audience.
Ricky Valentine, welcome to theshow.

SPEAKER_01 (01:54):
Hi, thanks for having me.
And you know, AI all the time,that doesn't sound like a bad uh
a bad name to the podcast rightthere, you know?

SPEAKER_02 (02:02):
Well, you know, the conversations that I'm having,
whether it's at kids' softballor you know, at work or
whatever, the conversationalways kind of AI just seeps
back into that conversation,whether it's people concerned
about taking their job, aboutwhat it can do, um, you know,
about like what they're hearinguh in the news, different

(02:25):
industries.
Like I can't get away from it.
And people know that like I, youknow, I'm pretty involved with
it.
So I I get pulled into theseconversations.
And you know, I I spoke at OTCoffshore technology conference,
um, and and we were talkingabout AI on physical devices.

(02:47):
So like that transition out ofIT, which where it's really
lived in a developer community,and and it's being applied to
actual assets and production,and you know, the internet of
things, and like I mean, it wasjust really amazing to uh
moderate a panel of some of thetop experts and like where that

(03:10):
conversation went, not on thepanel necessarily, but in kind
of little corners and throughoutthe conference, people having
different kinds ofconversations.
Um, what was was reallyfascinating.
And you know, as we move from uhyou know using it as a chat bot

(03:32):
to using it not just as athinking partner, but actually
actual you know partner orteammate, memory becomes really,
really important and the datagets polluted and the the calls
of what it remembers and thechallenges versus like when you
start working in an enterpriseenvironment.
So we've been shifting EWRdigital to an AI first company

(03:53):
where we're sharing all the likesame skills, um, where
everybody's connected in on likecertain processes so we can
standardize and scale.
And um, you know, differentpeople have different solutions
for different things, and thenthey they you know, like what is
the right solution?
How do you standardize it?

(04:14):
And you know, you and I havetalked um for hours really uh at
different times, and you know,different people are all gonna
approach a problem differentlyand come up with the right
solution for them, but AI justlooks at that as like noise or
like a bug.
And so how AIs remember stuffand for you to be able to work

(04:39):
with it and train it and youknow set governance to this, I
mean, memory's at the center ofit all.
Um so what yeah, what are youseeing?
Like kind of I'm just teeing itup for you.

SPEAKER_01 (04:52):
So I, you know, I guess kind of where where it all
started for me, uh, and I won'tgo through the the life story,
but it starts at a young age,and the fact of I started out in
the medical field and I alwayshated how much healthcare it,
you know, I I felt like once Igot an understanding for like

(05:14):
technology and stuff, I wouldsee that okay, wait, we're we
have data analysts that are likeprinting out something from a
two million dollar system andthen hand typing it into Excel.
And so like that just made melike really like want to fix
that solution and like have thatdesire at a young age.
But as I as I grew older, Irealized that the advantage was

(05:38):
that when you're in healthcareand in a regulated industry, as
you know, um you kind of have acrystal ball if you can look
outside of your field becauseadaptation just takes so much
longer.
So, you know, if you're lookingat where innovation is happening
in less regulated industries,then you kind of seem like

(05:59):
you're, you know, uh have acrystal ball because because I
can see, okay, hey, well, youknow, they've sped over there.
That how can I incorporate thishere?
And that's just you knowamplified by having AI because
now I don't even have to do allthe footwork of, hey, how do
these two dots connect?
I just have to have thewherewithal to go, hey, can you

(06:20):
tell me how this connects to me?
You know, and so that's reallyhow um how all this came about
for for me was I just saw theway AI was handling different
problems and and the memorything.
And it was, you know, we've beenable to Google something and
find the answer at ourfingertips for a long time

(06:43):
that's really not new.
Uh, you know, if for those ofyou, which I'm sure most of you
are, uh in the day and age ofjust good old Google, you would
type in something and then youfire up a form and you get two
or three different answers.
But sometimes, actually, moretimes than not, those three
answers would all work, but theywouldn't work for you, or one

(07:07):
would work for you.
And so to me, it was like, wait,the great thing about having AI
shouldn't be like I have apretty good memory myself.
I don't needed to remember theanswer.
I needed to remember how I gotto the answer, or what made that
answer the right answer for me.
Um, and so that kind of you knowpulled on this little piece of

(07:29):
thread that uh that got me tocrystallization theory, which is
kind of where I started myjourney.
So I I didn't ever, I nevertouched a line of code until
like 2024.
I'd always been kind of theconduit and mouthpiece between
business and technology, uh, butit was you know a lifetime of

(07:50):
developers telling me, well, youcan't do it that way, or it
doesn't work like that.
And so when this AI kind ofrevolution came about, I was
like, well, you know, now's Inow's as good a time as any to
see like who is wrong, right?
And it turns out we both were.
Um so uh in I mean that in theway that uh it does work like

(08:14):
that.
You can do those things, butalso shipping a product that
lasts 15 or 20 percent, as youknow, and you know the amazing
work you guys do at EWR, um,it's it's a lot harder than than
the internet is making it seemthese days, you know.
Um and so so kind of justlearning from those things and

(08:36):
and and going from not beingable to make a login page to you
know four patents and uh eightbillion tokens on one platform,
um it's been both humbling, butalso just a lot of a lot of uh
you know, hitting every branchon the way down and and learning
from it.

SPEAKER_02 (08:55):
Yeah.
So, you know, one of the thingsthat I think is important for
people, like there's a lot ofweb developers that listen to
the show, there's a lot of SEOpeople.
Like when we start talking aboutagents, they're really like
that's short for coding agent,right?
It's a coding agent thatdevelopers use.
So, like if you're not familiarwith a lot of this stuff and

(09:17):
it's really new to you, it'sbecause you haven't been in the
developer world, but thedeveloper uptake on a coding
assistant um is like over 90%,right?
So, like dev heavy environmentsum are utilizing this technology
at a rapid pace.
And you know, for someone to gofrom like you know,

(09:40):
understanding like analytics andbusiness to to be able to use an
agent and orchestrate like asystem and all that, it takes
some system thinking, it takesunderstanding how applications
are built, like going to getsome basic Python knowledge,
like not saying you have to codeit, but saying how it works,

(10:00):
like what you're saying of howit fits together is really
important.
But the caveat to that isanything you can think of, this
is what I believe currently,anything you can think of in
your mind is now possible.
So if you can map out aworkflow, if you can map out

(10:21):
guardrails, if you can um thinkthrough the problem and the
second order, well the firstorder, second order, third order
effects of like what you'regonna do and how that's gonna
impact it.
Like for developers, forexample, you got to think about
throughput, right?
Like you're building, you'revibe coding something, right?
And yeah, there's a bunch ofpeople say it's just super easy

(10:42):
to do it, and you vibe code thisthing, but you can't necessarily
go to market with it.
And like once you start gettinglike a lot of users, all of the
actions that are happening aregetting processed one user at a
time.
That was like even when we weretalking about big Bitcoin back
in the day, like it can'tprocess at the throughput of uh

(11:03):
like a credit card, like right,like you can't do billions of
transactions, so you got tothink through this of what the
end state looks like goingforward.
And there's been a lot of debateon can you vibe code something
or do you have to do it thetraditional way?
And you know, I I was kind ofwatching that debate, and you
know, again, I'm not adeveloper, but I've project

(11:25):
managed some things, and I'mgoing, well, if you train your
system to code it with thetraditional like subject matter
expert knowledge of how you'regonna build it, it's gonna
assist you to build it faster.
So it's not this or that, it'sit's the same thing.

SPEAKER_01 (11:44):
Yeah, and you try and explain it to one of the two
camps, and they don't becausefor the people that want to
spend three months planning,like I'm like, no, like we got
to iterate quick.
And then the people that want toiterate too quick, I'm like, no,
but you gotta have a plan.
And so, like, I soundcontradictory to both, but
you're exactly right.

(12:05):
And that's where I landed wasyou know, there's a famous quote
it's attributed to AbrahamLincoln, but they say it's it's
not really, we don't know who itsaid, but it rings true
nonetheless, and that's uh giveme six hours to chop down a
tree, and I'll spend four hourssharpening my axe.
And you know, it's it really isthat way in this day and age is
that yes, we can iteratequickly, and so we should, but

(12:28):
at the same token, you can saveyourself a lot of heartache.
No, no pun intended on the tokencard.
Uh, you can save yourself a lotof heartache if you, you know,
really do think it out.
And so that was one thing that Inever like having worked with
developers and stuff, I didn'tcome into this as somebody that
like was like, oh, I'm gonnaspend a couple hours a week and

(12:50):
I'm gonna get good at this andI'm never gonna have to learn
anything about code, uh, becauseI just simply knew that that
wasn't true.
I knew that I could, I knew thatI could use it as a crutch, but
I knew that, you know, from fromstarting out where it was just
pattern recognition of, hey, Inoticed every time I'm doing
authentication, like I getmessed up here.

(13:11):
Okay, we need to tell it towatch out for that.
Like, not even knowing what I'mreally trying to watch out for.
I just know that, like, hey,when we get to this point.
Um, and this was back when wehad context windows of just like
16,000 tokens, 32,000 tokens.
So, you know, it does one file,and it's like, where am I?
Like, are we building somethingtoday?
It's like, yeah, we just built200 files.

(13:33):
What do you mean?
Um, so and then also with theregulated industry thing,
because of the fact I had apassion for healthcare, luckily
I was able to get in the uhAzure uh Microsoft Founders Hub
Program and have credits and beable to utilize a lot of their
railing um, you know, a pipelinefor I feel like that's one thing

(13:54):
that a lot of people, um the thehybrids of of business and
development, they're reallyfinding the sweet spot.
And I think it's because ofthings like because I didn't
have the skill, I had to find,you know, not necessarily the
way around it, but the way thatI could get that, you know, uh

(14:16):
HIPAA compliance without meknowing, you know, hey, like,
okay, well, what's a pen test?
You know, um, and so it wasutilizing things like uh, you
know, I now I still had to havemy code be be solid.
And so I kind of used myself asthe uh the scout, if you will.
So I would kind of like moveforward and and make prototypes

(14:38):
and things, and then I'd have,you know, my development team
come from behind, and and itallowed us to kind of they could
work at the pace they wanted toand really have things mapped
out, but it it didn't constraintthe freedom of of being able to
try new things out.
So um, and now it's totallydifferent these days in the fact
that like you really can um do alot of things if you're willing

(15:00):
to just spend some timeplanning.
Uh and and so, but we still havea lot of things, I think, in the
way of governance.
And I know that this issomething that even though me
and you have known each otherfor a while, kind of uh where we
where we started having a lot ofreally long conversations was on
governance.
Um, and you know, that's adividend that's not not gonna

(15:24):
really be uh be paid or or a uha debt that's not gonna be
collected for for probablyanother year for a lot of people
because they don't realize whatbecause things do appear so easy
and and they're not, becauselike you said, with the
throughput and everything, youknow, it's really easy to make a
a superbase backend and and youknow uh mismanage API calls.

(15:48):
And oh, I just had you know,there's one person on my
platform and I had 15,000 APIcalls in a day, you know, like
so you amplify that intosomething that's just uh you
either have you know Richie Richas a a brother or or you know um
you're you're just SOL.

SPEAKER_02 (16:07):
So um yeah, I I I mean I I think that um that's
what I saw first, right?
And and I think that what whereyou came from is really that
sweet spot.
Like you're you're interactingwith developers, so you
understand some of thevocabulary, and then you
understand the business piece,and you're like sitting in the
middle of it.
Um, and so if people are on oneor the other, they need to go

(16:31):
try to build that vocabulary andunderstanding from that other
area because that's that'sreally where we're sitting as
like a human in the loop, as abridge.
We're we're a bridge between thethe coding agent uh and what
what the final output is andwhat the business or uh the goal

(16:52):
is that we're trying to achieve.
And you're you're the one thatis gonna facilitate that.
So if you're a solo operator orif you're working at a company,
you're still building anoperating system, right?
That you're operating in oryou're operating in together and

(17:14):
you're coagulating likedifferent um systems, and and
then really it's aboutstandardizing those systems.
It was just like in in digitalmarketing, the biggest issue
that we had previously was datafragmentation, like data was all

(17:35):
over the place, and also thebiggest issue right now in
moving into like OT is um thedata that is being produced is
unclean.
So even the div or or whatyou're talking about is is about
the penance, or that's not theright word, uh of just not

(17:56):
utilizing your data properly.
Like anybody that's using a CRM,there's a lot of people here
that are using CRM.
Is your data pristine?
Right?
Would you trust your life onthat data?
All right.
Um, how up to date is it?
How current it is, and you know,like I know a lot of people that
are using Salesforce or Hubstaffor whatever, but they're not

(18:20):
using it the way that developersintended it to be used.
And so the data is unclean andit can't be processed.
Now you amplify that to usingsomething on a physical device
of like uh you know, a pump on awell, um, that data better be
right, like because you'reyou're running all kind of math
and uh heretic like that thingcould blow up, right?

(18:43):
Or something bad could happen.
And so you gotta trust the data,and you gotta also the latency
is a really big issue there.
But like all of these, these uhlike sins, or I don't know what
the right like frame is to thinkabout this, but they're all
gonna come due.
And and so from an operationsbackground, the thing I zoned in

(19:06):
on was like, oh my gosh, you'vegot to have the governance set
up right, and everybody's got toagree if you're working in an
enterprise environment, andokay, why why are these
hallucinizations I can't eventalk uh happening?
Well, it's because of bad dataor you didn't give it enough
information, like it's not theAI, it's the human user.

(19:27):
Um, and so I don't know, Idoubled down on that, and then
there's now regulatory laws thatare coming down the pipe that
people are not expecting.
I'm going, okay, like why don'tlike there's a period, you know,
says it's September, but theymight have pushed it out a
little bit.
Um on Moto Point, I set up likea countdown timer because it's

(19:49):
like it's coming, right?
And and and so it's like getyourself ready, um, audit your
tech stack, understand what'sgoing on, map out your API
calls, like bring your tokenusage down.
Like it's it's gonna be awin-win all around if you set up
the system properly from thebeginning and get everybody

(20:10):
onboard on that system.
Because if not, the problemsyou're having, let's say just
use the CRM as something thateverybody can follow, is gonna
get amplified.
The problems are gonna just getamplified with AI and the noise
and degradation, and like you'rejust trying to fill up the
context window of all thisinformation.
And you know, you know, whenyou're talking to AI, after a

(20:34):
certain amount of time, it likewill compress the conversation,
but you lose context and youwere having this great
conversation, and then oh, likenow it doesn't even remember
like what you said.
And when I was using AI as athinking partner a couple years
ago, that was like the mostfrustrating thing ever.
I was like, Oh, I was in likethis really good place, and you

(20:55):
like wanted to keep that chatthat way, and you know, and then
like I remember when uh uh uhchat GBT like 4.0, like they
upgraded it and they took awaythe persistent memory, and that
was when I switched to Claude.

SPEAKER_01 (21:08):
Like, I was just like, I just lost my thinking
partner and it can't rememberanything, but it at least with
Claude, like it would tell youhey, start a new conversation
with Chat GBT.
You know, you'd be thinkingnothing was wrong, and then
you'd be like, wait, you justgot a lot dumber, like you know,
uh and uh to your point.

(21:29):
So I was actually uh with a goodfriend of mine, Hank, uh in and
he's a uh a vice president oftreasury at a at a one of the
bigger, bigger bank bank brands.
And uh we were out one night andI was talking about you know how
look, we have to, you know, I'vehad been so focused on the
viewpoint of like not making a aa world where where there

(21:54):
wouldn't be hallucinationsbecause I thought that that was
the crux of enterprise, becauseyou know.
So remember that I've just gonethrough my like odyssey of like
coding the past year.
So I kind of lost touch withlike a little bit of my business
activate while I was in thehole.
And uh and he goes, Ricky,they're not worried about the

(22:14):
models hallucinating, they'rescared of their own data.
And just there was, you know,that's in the top five most
profound moments that I've hadin the past year.
And then to him, he's like,Ricky, like what?
Like, do you are you okay?
Like, do you know, like, do yousee something?
And I'm just like, you know,molding this over.
And I'm like, dude, I neverthought of that.

(22:37):
And so as I'm working throughthese problems, I'm thinking
about, okay, now we have to goback and we have to look at
ingestion, you know, and and anduh and backtracking that uh to
the point where at one point Iwas like just convinced that,
hey, okay, we're just gonna havelike shadow hospitals and like

(22:58):
shadow billing companies, andlike all they're gonna do is
like, you know, just uhbasically sit there and be the
like non-biased, like, okay,we're gonna decide what like has
right authority and like whatgets like saved is memory.
Um, because you know, when youthink about it, uh you know,
there might have been thatperiod of time where like uh

(23:19):
Susan, you know, was out onmaternity leave, and and so your
books aren't, you know, like uhyou, you know, you had a temp
that was, you know, doingbookkeeping.
And so, you know, those thingsthat you don't really feel so
much at the time when you havesomething that can remember all
of this and it's using thosedata points to the T, you know,

(23:39):
you you know, back then your CPAjust came in at the end of the
year, he said, Oh, I'm gonna fixthese things, you know, did some
you know magic with the numbersor got you some you know
electric vehicle credits orsomething, and you were on your
way, you know.
But when when each data point,you know, and that's where
really I got frustrated too, wasyou it was all or none.

(24:02):
And that's where like my workhas kind of been in memory is
you know, you it's eithershaving everything down to fit
one answer or it's rememberingeverything, and then what's the
point, you know?
Um, and it's the same thing withcontext, you know, it's either
taking everything too literal orlike not remembering.
I'm like, didn't we just, youknow, uh, and so kind of

(24:26):
threading that needle.
Um, and that's where looking atthat, and back to my analogy of
the forms uh and makingcrystallization theory is
basically as you compound thoseanswers and as more people in
your network um, you know,validate those answers, and it

(24:48):
has to have human right, like ahuman is the only one with right
authority.
You know, that's another thingthat kind of drives me crazy is
I'm such an efficiency person,but even I can't get through my
head like, wait, you know, we'resaving all this time, so we
can't use some of that save timeto like double check the work
and like you know, like we wanteverything to compound, learn,

(25:11):
and just get smarter.
I I like to say it gets smarter,but it doesn't get wiser, you
know.
Like, so we're just we're we'resaying more, more, more, get it
smarter, get it smarter.
But it's I'll take you know, itbeing wise and being able to
produce something, you know,that that has some merit that I
didn't think of, or you know,that uh and so that's kind of uh

(25:34):
kind of where where I I leftwith that.
And I'm kind of losing my yeah,no, no.

SPEAKER_02 (25:38):
So there's two things that you said that that I
want to uh highlight.
We've started to change ourworkflows, okay?
We've we've started to changeour workflows where we have some
power users that can produce amassive amount of output, but
QAQC becomes criticallyimportant because as of today,

(26:00):
like I can't trust the output'sgonna be perfect every time.
And like, unless you even setlike uh like hooks, for example,
if you're on COD, sometimes youset a rule, it forgets the rule
because it gets lost in thecontext window because it
doesn't reread everything allthe time.
Well, I think it it does, andthat's why you use so many

(26:23):
tokens, but it it doesn't weightthem, like right to what you're
saying, and what you keed in on,and and I believe this is
happening to kind of bring someof this back to digital
marketing uh for everyonelistening, because this is a
mindset shift.
All of these platforms arealready working on math, okay?

(26:44):
They're already runningformulas, like all the different
social media platforms andGoogle uh are all you know
already running these things inthe background.
And one of the things you'retalking about is like the value
of authorship, right?
Like who's saying what, thedegradation rate of when did
they say it, like the importanceof how current it is.

(27:07):
You're seeing all these thingskind of come to light uh when
when you're using agents andwhen you're looking at memory.
But like Google, for example,you know, uh a review on Google
uh or a comment on a socialmedia post, who the person is in
association with the post, andtypically in association with

(27:29):
you, because sometimes they showit to people that like your uh
content or they show it to yournetwork, like so there's
different weighting factors, butthen what it is they're saying,
right?
Short, long, is it related tothe post?
Um and then when did they sayit?
You'll actually see on posts uhlike there's a rating system,

(27:52):
right?
And so you'll see it on the postthat the most recent post
typically stays at the top, itgets uh auditioned, and then it
falls to kind of where it sits.
And so think about that withmemory.
And I think you you and I weretalking about this uh before,
like when someone goes to sleep,I think your brain is like

(28:14):
compressing all that knowledge,and like what does it choose to
remember and not remember inshort-term and long-term memory?
And even a lot of where all thismemory stuff is going is the
most elegant piece of equipmentis the brain, right?
And we're starting to replicatethe brain in a lot of ways, uh,

(28:34):
is what it like.
I just see a lot of parallels.
I don't know.

SPEAKER_01 (28:38):
Yeah, so you know, I uh I've lived a lot of lives,
and in one of those, uh, I wasworking in brain mapping for
personal injury cases.
Um, and we used uh a reallygreat machine that was uh
quantitative EEG EKG and evokepotentials test uh that was able
to see TBI but also be able tosee, and I'm gonna circle this

(29:00):
back to yeah, I don't know whatyou're saying, but uh anyways,
so it was it was a great scanand it it allowed me to kind of
like learn a lot about thebrain, that you know, get just
be very interested in it.
And so one of the things that Iput in my my memory program is
is the ability to forget, youknow, like that is the decay.

(29:24):
So I use a weight formula umthat has a decay, and then based
on the importance, kind ofexactly like you were saying,
based on the importance and theweight of that is at what rate
it decays.
Uh and you know, when it comesto like the other day, I was
thinking of things in terms ofuh an agency, you know, for for
digital marketing, and like, youknow, people don't realize like

(29:48):
how much governance is like itaffects every industry.
So, like, for instance, youknow, you don't know if uh, you
know, somebody might not know iftheir employee, you know, did as
much work or that that uh thatsubcontractor did as much work
as they say they did, or or whathave you.
Well, that's drift, you know,like that's uh you know, the uh

(30:11):
is is the the digital truth.

SPEAKER_02 (30:14):
What is that what is that?
What is this?
Like bar what where does thetruth start?
Like, what are you using as thereference?
And then what's that deviationfrom that or that drift of of
where it's going?
Because you're like, okay, andthat's like scope creep, guys.
Like, that's basically whathappens is you're like quote
this and you're like, oh,everything's gonna go perfect,

(30:35):
and I can charge this, and thefeedback from the client's gonna
be immediate.
Like, you have to start buildingin, um, knowing that that there
is gonna be that that scopecreep, but how do you limit how
much of that there is?
And I I look hand hand hats offto all the other digital market.
I'm going to a conference uhnext week, actually, that's just

(30:58):
for agency owners, right?
Because I really want to see, Iwant to kind of benchmark too,
kind of where we're at versusall the other agencies, what
they're talking about, howpeople are dealing with the
injection uh of of AI into theirworkflows.
But to there's not a lot ofreally big agencies, uh Ricky.
There's like you gotta reallyscale your business to get those

(31:22):
efficiencies of scale.
There's a lot of like a couplepeople and a dog, I say, like
you know, there's a couple guysand a dog in a house, like
agencies, um, uh, or there's alot of freelancers.
Um, and hopefully, if you're oneof those people that has your
dog right next to you, pleasedon't be offended because
sometimes my dog uh will be hereas well.

(31:42):
So uh so I I meant that I've metthe stereotype I'm talking about
myself, but um there's not a lotof agencies that can scale,
right?
And that are in that phase ofscaling because there is so much
governance and there's so muchdomain knowledge, and there's so
much that has to happen.
But the shift with AI is you canstart building teams and agents

(32:07):
that have domain knowledge thatyou can build governance around
that can start to help you fillin those gaps and to get through
that that area of uh like thescaling from the small company
to the large company.
But it's so much thinking thathas to go into it, and you're

(32:28):
not just building one brand,you're building multiple brands
for other people based on youknow the lead gen that if that's
the focus of what you're doingfor those clients.
And uh, and so I I think andyou're starting to hear it to
talk about with AI, there'sgonna be billion-dollar
businesses um built with AI.

(32:49):
I think the job of a digitalmarketing agency is like 10
times harder because you're notjust building your business,
you're building businesses forother people, and all the inputs
and the workflows, and uh howthat client likes to work, and
the communication and thepersonality is constantly in
flux and changing, and you gotto plug in all these other

(33:11):
systems to fit into your system,and and that's where I think you
know, uh, like this agenteconomy is is gonna uh there's
gonna be standardization of howbusinesses interact potentially.
And and uh I don't know what areyou what are your thoughts
around that?

SPEAKER_01 (33:28):
Yeah, yeah, it's I do have to say hats off to the
agency owners because as I'vebeen doing more, you know, uh me
and you have been talking more,so I just you know I'm naturally
curious.
And uh I don't know whatagencies did the AI, but like
they do not see you as a tan,like like a market to go to.
Like you did it, was it soundedpersonal.

(33:50):
Like every time I would questionAI about like, hey, making this
solution or that solution,because because you're right, I
guess there's just you know,there's there there's kind of
like there's big agencies andthen there's you know, kind of
like mom and pop operations.
But one thing that tipped my hatoff in all seriousness, too, is
that you have to standardize theresult, but make it different

(34:14):
for every customer.
You know, talk about a paradoxthere, you know, like I need to
have the consistency in mydeliverability, but no two
answers can technically be thesame, you know.
Um, so I know that just being uhuh WordPress was kind of like my
my first love, so to speak.
Uh and I know that that's reallythe hard time.

SPEAKER_02 (34:36):
Houston boy.

SPEAKER_01 (34:38):
That's one of the hard times that they're having
over with at with WordPress withAI is yeah, sure, you can get it
to do anything, but try and getit to do 200 SKU product product
descriptions in the same way,and you know, like even if and
so if you don't understand, uhyou know, and and even me, and
even then it can still take itsown uh turn on it.

(35:00):
You know, if you're not makingthose outputs in the schema, you
know, outputs that like hey,this has to be like this.
If you're just relying on a on ayou know, a large language model
to just hey, it's gonna come outthis way every time, even with
the prompt, um, there's gonna besome variance.
Um yeah, no, I was just oh yeah,keep going.

SPEAKER_02 (35:23):
I was I was basically I was just saying you
can do that with a harness,right?

SPEAKER_01 (35:28):
Yeah, yeah.
Well, well, and yeah, but eventhen it just uh it's you do that
with with with you know to behonest, I don't know.
I just I went through a lot ofdifferent scenarios that like I
come from regulated industries,and after looking at like what a
like mid-level agency becauselike first off, you're like

(35:48):
you're in a you know, if you'rekind of you know in the in the
middle or or kind of higher up,like if you're higher up, like
somebody's you're making acustom solution and like you
already have had it for a while.
Like if you're in the middle,you're not a big enough market
for somebody really to build foryou, but like you're building

(36:09):
for everybody else.
So then how do you build foryourself?
You know, like uh it's so so Ijust I I I consider myself a
like a man that walks betweensilos and solves problems.
And even I was like, it's Ispent a whole night just like,
well, what if I used Dockerprofiles and and each cut client

(36:29):
had a profile, and then it waslike, Well, yeah, but then you
know, permissions, okay.
Well, then we're gonna make anMCP that then, and then I was
like, what about efferennialcontainers, you know, like like
uh and so making the client cometo you instead of you logging
into the client, and uh it itjust it was a fun thought
experiment, but yeah, I'll giveyou some tools that have gone

(36:51):
down that path, and that hasactually been like the scalable
solutions to basically you knowuh install something uh or they
have to plug into it and they'reworking within your system, but
then you get into datagovernance issues of like
they're collecting everybody'sdata and process, and so there's
a there's this is a tough thisis a tough industry, and it's

(37:15):
it's certainly tough if you'recoming in to crack it.

SPEAKER_02 (37:18):
How I have perceived it today, and I've been bringing
on different agency owners thathave moved through this path.
Um, and where I see them goingis and this is how I view myself
as well.

SPEAKER_01 (37:33):
I'm trying to build an elite team that thinks about
AI because I saw how cool thisstuff, and I won't I won't let
the cat out of the bag, but Isaw the cool stuff.
Like you were kind of like thetrajectory you're taking with
EWR, and I was like, that'scool.
I want to think about that for alittle bit, and then I was like,

(37:55):
No, I don't, that's hard.
Like, I don't no wonder, youknow, like uh that's that's a
difficult, like I don't know howyou put together so much you put
together because uh yeah, I didthat for one night, and I was
like, Yeah, I'm going back toregulated industries like you
can have that.

SPEAKER_02 (38:10):
Yeah, no, and AI what will like say, oh, this is
not where you should be spendingyour time.
Well, here's the thing I thinkwith that knowledge and all the
domain knowledge that I pickedup from different industries,
building an AI first team andbuilding a leap like team of
project managers, right?

(38:31):
You can now take that and applyit to anything.
Okay.
And I think the whole game nowis to take your business and add
that AI layer and make as manypeople at the possible as
possible AI first.
And so that's why I thinksmaller teams where everybody
can talk the same vocabulary andlanguage and wear those glasses

(38:54):
and look at problems their ownway, and then also interact and
understand governance of like,we're gonna, I'm gonna have like
my enterprise, I'm gonna have mylike individual, then I'm gonna
have my agents, and then I haveagents that have like different
skills, like you need tounderstand that hierarchy or
that structure and be able tohave that in your mind's eye of
how those operations work.

(39:15):
And then you can just take AIand you can start applying it to
different industries.
And so one of the things thatI'm actually working on, we were
we were kind of talking aboutbefore this call, is I mean, I I
just went through this GoldmanSachs 10,000 small business
program.
If you're uh a business owner, II would encourage you to go look
at it.
It's a completely free program.

(39:37):
Um, it's it's basically the samehours as NBA, but it's applied
to your business.
And um, I was like the only realtrue agency owner in my cohort,
but I was also the there wasmaybe one or two other people
that were using automation.
They weren't maybe using like agentic AI um or autonomous AI or

(39:57):
whatever word your industry isis using.
Uh it's a similar vocabulary,but but they all are asking me
to do a talk about about this,and like they all understand the
uh potential of what it could betheir business, but they don't

(40:18):
know where to start or where toapply it.
And I was like, okay, let me dolike an exercise to get your
mindset right so you have themindset shift.
And I think we're kind of doingthat with this call to to
hopefully to agree you're you'regrabbing on what we're saying,
and you you can kind of uh seesee the trajectory now and kind

(40:40):
of see a better picture of wherethings are going.
And and we can apply this.
Like I have workflows to applyto social media and I have it to
apply to paid ads, like Mantis,for example, uh has been
ingesting all the data of Meta,right?
And so, like, if you want to doum if you understand the basics,

(41:01):
the fluency of AI, and then youwant to apply it to to Mantis uh
for paid ads, because really ifyou're gonna be running like
Facebook or Instagram ads, likeit has so much of the data set
that maybe the others don't havethat layer on top of the model,
right?
And so you can get morecustomized output.

(41:22):
But if you understand this, youcan pivot, you can change, and
it's just what are you gonna putyour focus on?
And so what I'm trying to buildis is an elite team that all
speak the language and worktogether, and we can apply it to
any business problem.
And those are the businessestoday that are winning, the
businesses that have alreadygone through not just digital

(41:44):
transformation, right?
Everybody's like, oh, I justmade it through digital
transformation, sort of.
Uh, and now you're like, nowit's the next mountain or the
next peak, is like you need tostart climbing AI because the
people that are already upthere, they're gonna be they're
gonna be so far ahead of youthat it's gonna be really,
really hard to catch up.
And these moats are starting toshrink of these big companies

(42:07):
that are utilizing their brandand their trust that they've had
for 40 years.
I mean, that's gonna go a longway, but it's getting eaten away
at.
And so you got a elite team ofpeople that say we're gonna
tackle this business problem andwe're gonna extract value from
this, that, this, this kind ofarbitrage.

(42:28):
Like it's just AI's eating theworld, like what I was saying at
the beginning of this call.
AI is eating the world.

SPEAKER_01 (42:34):
And I think what's interesting and different about
AI is you know, usually you haveto you have to drag people into
new technology because it's it'smore tactical to practical.
You know, it's more the more theyou know, the the innovators or
the really high-tech people thatget to it first.
But because we have thisconsumer component of it that
has grown so rapidly, now it'slike businesses and like you

(42:57):
know, everyday people that maybearen't so technically inclined,
they know that they need it, butthey don't know, you know, and
the people that they're callingon, like not you, but like you
know, they're calling on peoplethat they think know AI, and
it's like they don't in there.
So, you know, one of the biggestthings that I like to tell
people, I kind of like had madethis like almost like a daily
devotional um on my website thatlike uh and I was just posting

(43:21):
it because my like family, itreally started with me like
trying to teach my family how toutilize AI.
Yes, it's like look, it's not inmodules or like learning how to
do it, it's in learning likelook, just talk to the agent,
like gone are the days of like Ineed a prompt, like you know,
that's like people when they youwould carry their uh like floppy

(43:43):
discs, you know, and you had a aprogram on like you know, uh 200
floppy discs.
It's like look, just talk to itand experiment and just talk to
it like you would a person,like, hey, can I do this?
Can I do that?
Um, and and by doing that andjust you know, getting
comfortable with it because Isee so many people it never
fails.

(44:03):
You you're you're walking themthrough utilizing AI and they
want it so badly like they wantevery word like because the one
of the first things I do isinstall like Whisper Flow or
some sort of dictation on there.
I'm like, look, you don't evengot time.
But they they focus so much onthose first three words and they
can barely get them out.
I say look relax.

(44:25):
Just talk to it like just talk.
And then that flows into youknow now I have my mom like
she's running her whole businesslike from a folder on her
desktop and Cloud Code and I'mlike you're doing better she's
like oh you know I'm sorry RickyI'm like no I'm just glad
finally somebody's like you knowlike all my tech nerd friends

(44:46):
are like there you know or likepeople that uh that they think
that they want my uh my advicethat they they're more hard
headed than more setting theirways but and that's where I
think people that you might feellike you're late to the party
you might feel like you'regetting left behind but truly
like you can be a thought leaderin this place if you dedicate 12

(45:07):
eight to 12 months just you knowa little bit of time each day
and just you know compoundinterest um it's it's not pretty
it's not vibe coding an app in aweekend and it's making money
next week um it's not pretendinglike that and selling a course
about it it's getting in thetrenches getting like you've
done you know all the educationand uh you know I'm lucky that I

(45:31):
just get to kind of like heyMatt like yeah talk to me about
that of course you you know anduh so I between that and and you
know get up I get I get but uhbut yeah people don't realize
how there's no one's we're allfiguring this out you know it's
kind of like when you're comingof age and you realize that like
your parents are just trying youknow they're doing what they can

(45:52):
they don't know you know that'sthe same way with AI like
there's nobody that we don'teven know how these things and
this is a conversation foranother day we don't even know
how they actually work like noone does so you can't uh you you
know I wouldn't consider anybodyan expert at this what if we
don't know how they're made wellokay so like here's what I'm

(46:14):
struggling with right now okayso I have like a like I I didn't
talk about AI I was certainlyexperimenting we were using it a
lot but as as kind of you know aa thought leader or you know
somebody that's out thereinfluenced whatever you want to
call it I didn't want to talkabout it until I was super
educated and to your point thethe compound interest of the

(46:34):
foundational knowledge andbuilding and building and
building like I think certainthings are just well understood
but when you talk to people ifthey don't have that depth of
knowledge there's some gaps thatyou have to traz traverse to get
to the next level right and umwhat I want to be able to do is

(46:55):
help get those people into thatmindset so they don't have to
fall in that ditch and gothrough all that learning.

SPEAKER_02 (47:01):
They can expedite it and and I people have been
asking me to start a coachingprogram for a long time.
And like I'm like like it's alot to teach somebody how to do
SEO or you know digitalmarketing.
Like there's I mean I have likethree well the first book I
wrote I read 300 books before Iwrote my first book okay like
and I was like highlighting itup and I tried to cram as much

(47:24):
in there as I could and I'm I'mworking on another book by the
way I that's how I processstuff.
Okay like I talk about it youknow I I I listen I read I write
like like I just I I'm just likea like I it flows and and I'm
ready now to to to teach peoplewhere they need to go like and I

(47:47):
feel comfortable enough that I Ido believe I'm somewhat of an
expert on some of this stuff.
I mean AI just told me I'm liketop one percent I just did a
post on Facebook about it butI'm like but then I also
discounted why it told me that Iknew and I was like it's like
sycophantic like the these theseLLMs are gonna be a mirror and
tell you what you want to hearbased upon the output you get it

(48:09):
because it wants to give you theright answer.
Right.
And I'm like you got to knowthis stuff.
And and I think that whereeverything is going and this
goes back for me.
So people that have beenlistening to this podcast know
my mom was one of the firstemployees of Microsoft.
I was there I watched it she wasthere for 30 something years and

(48:29):
she was one of the firstemployees I was playing video
games when they like before theXbox when they were just coming
out with online PC games I wasplaying multiplayer games with
my mom's like teammates andshe's like you can't talk to
them like that like that's notsomebody you should talk to and
I didn't know like I was youknow I was just like this is who
it was and she said you knowyeah and and I look I was gonna

(48:53):
be a computer science major likelike I I I went the business
route but but I can tell you itit was a love for me that
brought me back to it and I justlove technology and I saw the
writing on the wall I saw whathappened right you put on your
resume worldwide web I haveexperience with the world like I

(49:14):
can read also when people putlike AI experience on their
resume I I know about like wheretheir experience level is based
on uh how they talk about it umuh and it's the it it it's gonna
eat everything it's gonna AI iseating everything so if you have

(49:36):
domain expertise in what youhave like just apply AI to it be
expert in AI like that's what Iwant my kids like understand how
to use AI and then apply it toreal world knowledge and your
domain expertise and the moredomain expertise you have and
real world knowledge as soon asyou add that AI layer to it it

(49:57):
makes you really really powerfuland if you work with a bunch of
people and like I gravitate likeright where we great you and I
are like gravitating to eachother with different kinds of
experience and there's otherpeople I'm gravitating to those
people and those people startworking together to achieve
goals like you know it's it'sgonna be amazing.
It's absolutely going to beamazing.

SPEAKER_01 (50:18):
And and I do want to say for for you know I know a
lot of business owners uh listenand and tune in and uh you know
10 000 uh 10 000 people a dayare are of retirement age
retiring and what you know someof those people might be feeling
at this point in time that thatthat kind of they're oh yeah

(50:39):
they're but they're up but theydon't realize that like they
have their value is so so likeif you own a business today if
you're looking at selling yourbusiness today look at the value
of the data so we've the thesewithout getting too technical
about it you know these LLMsthese these AIs they've learned

(51:02):
everything they can from theinternet the the the private
data of you know let's say youowned a logistics company an SEO
company or this was any forthose of you that have been you
know meticulous in recordkeeping it is your time they
said they said it never younever need that well guess what
they were wrong you can fileeach other to them uh and I saw

(51:25):
a company actually just got likea$300 million dollar raise the
other day talking about justthis and it's that look like the
domain in a world where the thattechnology gap is taken it's the
domain experts and understandingthe way the processes work you
know the only the the realunicorns I think are going to be
the people that are come from adevelopment background that

(51:49):
throw themselves into businessinto real world processes
because then they'll have bothand that and that'll just be an
unfair advantage but it's it'sit is you know and I hate when
they talk about you know like uhjunior devs are going to be gone
but I think where it wherethey're seeing that is that the
it until a certain level domainexperience beats technical

(52:16):
knowledge right now and thenwhat you know once you get in
the upper echelon to tech so Ithink that's where people may be
feeling that disconnect orthinking that's gonna happen
it's it's more so that look likeyou know right now you know but
if a technical dev can just getsome of that system thinking I
think it's kind of it's kind ofa freaky Friday situation where

(52:37):
if if this side could do thisand that side could do that you
know that we're all we can allbe successful.
I don't know my true definitionand I'll kind of like end on
this my uh my thing about AGI isI don't see it I'm sure we'll
get there but my definition ofAGI is AI bringing the best out

(52:59):
of people and being your crutchin what you're bad at and
helping you identify that.
And so I think that's becausethat's the way I use it.
I mean kind of I know me and youshare this kind of like ethos of
it that uh you know use for goodum like it really can be a
powerful tool not on a like heyit told me to do this so I'm

(53:22):
gonna go do it but on a justthought experiment of like hey
we could you know just takingsome inventory and and seeing
hey like am I good at this can Iget better at it like so so
Ricky I want to give give you anopportunity uh to to kind of
share some of your papers and umyou know what you have out there
but I want to end on this guyslike Sam Altman started white

(53:46):
well I don't know if he startedwhite combinator but he ran
white combinator for forever andhe started a lot of businesses
or got involved in a lot ofbusinesses like Airbnb like you
should go look at hisbusinesses.

SPEAKER_02 (53:57):
He left to go run open AI if you look at his
businesses all of them can useAI like all businesses can use
AI but it's like a hub and smokemodel.
He's taking open AI and he'sgonna apply AI to all of his
businesses to get that unfairadvantage because of how they're
thinking and where they're atlike that's that's what I'm

(54:20):
seeing on a smaller level rightI'm going I'm going there's a
big transition of businessesgetting handed off uh as as or
passed down from newergeneration but there's a lot of
businesses that are going tochange hands.
You got to capture that domainknowledge but you can apply AI
to these different businesses umand then apply capital to these

(54:42):
different businesses and andbuild that kind of operational
workflow between humans andagents and scale these
businesses quite quickly.
I mean I recently uh bought outmy three partners at the agency
because I wanted to take EWR ina little bit different direction
and I already got uh a number ofuh investors and people that

(55:07):
want to uh do do different kindsof opportunities together that
that know what we're capable ofthe team they've worked with us
for a long time and I'm kind ofgoing I'm I'm not kind of but I
am going through that processI'm saying kind of because it's
all new to me right I don't havethe the the JV capital raise

(55:28):
like I'm getting those domainknowledge people around me to
okay take something public or todo something like that.
And I'm assessing all thebusinesses I've seen over you
know a decade plus and goingwhat are the businesses that we
could apply AI to and whatindustries that could move the
needle the most and so I'm nowviewing everything from that

(55:50):
skill set and and we're we'rewe're starting to move towards
that and we're I mean we'rewe're attracting companies like
Circle like circle.com found usthey hired us to to do some of
this stuff and and like theworld is just absolutely
shifting and and I think yourpoint on if you have that domain
knowledge or you have that datayou are in the cat bird seat uh

(56:15):
if that I don't even know whatthat really means to be honest
but um like if you're in theright spot and if you've now
left mainstream you've beenworking for somebody for a long
time whatever and you'restepping into consultancy like
those are the people that aregoing to really move the needle

(56:36):
if they get the basic knowledgein AI.
And that's really where I thinkthat opportunity is is that
domain knowledge AI knowledgeand apply those together.
And hopefully this podcast hasencouraged you to say hey I can
do this I understand this thisis new but I've learned a lot of
new things and this is going toamplify whatever whatever I do.

(56:58):
And I'm just encouraging you totake action now if you want to
like or follow or reach out tous comment if you like kind of
where I'm taking this podcast umplease let me know.
We are going to move more tomaybe YouTube and some private
trainings which I've done in thepast but but really focus on AI.

(57:20):
And I I just don't see any otherfuture without it on the bad
side though right cPanel whichis based here in Houston it's
got millions of websites takendown I think that that was
probably one of the newer modelsthat people are using it for
nefarious things.
There's gonna be a lot of thatthere's there's always going to
be good versus evil sort ofthing yin and yang and um you

(57:43):
know like I I'm surprised morebad stuff hasn't happened yet
but but I I'm I'm I believehumans are naturally good and
are focused on good.

SPEAKER_01 (57:52):
And if we have more humans on the good side making
the world a better place uh likethe world's gonna look amazing
uh in a couple years and I'mexcited to see that so Ricky how
do people get in touch with youhow do they follow your work how
they find more out about memoryand kind of some of the other
things that you're doing yeah souh I have a website

(58:13):
Richardvalentine.dev umdrvalentine on pretty much every
social uh so dr valentine's feltlike the holiday uh and so yeah
you're gonna be able to some ofit's not gonna make any sense
but then I also have likethere's a whole page uh I think
I'm on day 20 now of uh smalllike 20 minute daily devotionals

(58:37):
to AI that like it starts youfrom like making a folder so
we're gonna make you your ownlike brain like on your computer
and walk you through the AI touh and it's it's gonna be more
of a teacher man to fishmentality because that's you
know uh I to your point aboutthe consultancy and stuff I know

(58:58):
it might seem like a crowd spacebut you have to understand that
all these people that that youthat that the listeners might be
seeing out there that areselling courses they're not
really doing that consultancythere is so many small
businesses that could use yourhelp your expertise and um you

(59:19):
know a lot of these things ifyou look at it from the the
workflow level and not the inand solving a problem but but
okay what is that problemactually made of uh you'll
notice that all of a sudden thatsolution you made for one it
dominoes and it can be used inso many other places um so yeah
I'm LinkedIn uh Richard AlValentine uh but yeah I'm sure

(59:44):
you'll have the the links andalways looking to uh provide
value where I can uh I don'thave a course so you don't gotta
worry about that uh he's notselling anything I just asked
him to come on because I wanteduh I wanted to share some of our
conversations with you becauseI've I'm like after we get off a
like a big session I'm like ohwe should have recorded that you

(01:00:06):
know and maybe we will goingforward right and uh you know I
I think uh a lot of people umare gonna need help and I think
a lot of people want help rightand I I think it comes down to
like trust and relationshipsright because everybody can say
whatever they want to say that'swhy I eat uh expertise authority

(01:00:26):
trust Google added experienceand they want to verify that in
some way and authorship ways andlike they're trying to say well
all the information you said itat the beginning all the
information in the world is outthere but how do you use it what
to do with it what's moreimportant what's valuable to you
these are all the things you gotto get figured out and and I

(01:00:48):
think if you had a guide to helpyou do that um you could get so
much more done and I I mean Iwanted to start a nonprofit
honestly that could help peoplewith data hygiene uh issues like
we don't get taught that inschool uh but data hygiene I
think is where it starts andthen viewing how to view

(01:01:10):
everything from an AI firstmentality um is is like an
indoctrination to a certainextent and um we've had to do
that one by one and I would loveto do it in groups with people
and you you find out withbusiness owners and um different
people in your network there'sso many commonalities uh and

(01:01:31):
there's so many ways to worktogether um and and you know the
the the amount of people thatare using AI really using AI
versus the amount of people thataren't like it's it's a pyramid
inverted right or if youunderstand that analogy if
you're listening um you'reyou're at the tip of the pyramid
uh if you get into it now andit's gonna be the tail on this

(01:01:55):
thing is gonna be who knows howlong uh and also when are people
gonna stop learning and plateauum but it it's gonna be 10 years
like why if you know where it'sgonna be would you not get
started now and and and that's Ithink the biggest thing I want
you to take away from this isand I've been trying to get you
if you've listened to this takeaction now because the people

(01:02:17):
that are using it are gettingbetter and better and more
refined.

SPEAKER_02 (01:02:21):
And at a certain point it's gonna be really
really hard to catch up rightnow I think you can jump in
there but it's gonna be reallyhard to catch up.
So um if you like this episodelike I said please like it share
it follow us let me know I'mdoing a good job tell me not to
talk as much uh tell me thissounds bad like I want to hear
your feedback uh I I'm having AIlook at all my comments comments

(01:02:44):
and roll them up and let me knowabout them so I will see your
comments uh it's hard on all thedifferent platforms that people
comment on a review would besuper helpful um it it it's
called Shaiko share like ball usif you've been listening for 16
years um but until the next timeuh Ricky thanks for being on my
name is Matt Bertram bye bye fornow bye guys

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Crime Junkie

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Bad Data Causes More AI Failures Than Models Do with Richard Valentine

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Crime Junkie

Dateline NBC

All Episodes

Bad Data Causes More AI Failures Than Models Do with Richard Valentine

Stuff You Should Know