March 3, 2026 • 39 mins
AI is getting bigger.
Bigger models. Bigger data centers. Bigger energy demands.
But what if that’s the wrong direction?
In this episode, Dustin Hedrick and Brandon Billings challenge the trillion-parameter arms race and explore a different future — one built on Small Language Models (SLMs), decentralized storage, distributed compute, and blockchain coordination.
They break down:
• The hidden environmental cost of hyperscale AI• Why specialized smaller models can outperform massive generalists• How edge-native AI reduces energy, latency, and centralization• Why the next phase of AI may be horizontal — not vertical
Instead of industrial-scale intelligence factories controlled by a handful of corporations, what if millions of participants powered a global AI mesh?
The future of AI isn’t about building bigger buildings.
It’s about building smarter networks.
Build smaller.Build larger.
Weekly ROAR Podcast with Dustin Hedrick & Brandon Billings
Sponsored by https://www.r0ar.io/
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
SPEAKER_01 (00:01):
Welcome back, my friends, for another episode of
your weekly roar.
Folks, I don't know if you knowthis, but you've been with us
closing in on our first year.
So I want to say this is intoour we're getting closer to our
30th episode real soon.
And with a couple weeks off hereand there, this your weekly roar
has been growing.
Also, hope you've seen thegrowth in our topics.
(00:24):
Um, from Roar and Fierce Labs,we've been really going after
some key, key, key conversationsfor the future.
Future blockchain, web3, ai,decentralized storage, and more.
And so um we're rolling outwhite papers one right after the
other.
And with that said, on theheels, or actually out in front
(00:45):
of a white paper that's comingyour way on the AI space, one of
the first of probably four,we're doing this podcast today.
So um, guys, it's a littledifferent again.
Um, I'll say this that uh thisis definitely a stretch for us
to get out in front of somethingand go the opposite direction of
(01:07):
where most development andmarket is going.
Uh, but we feel like we need toweigh in at this point.
We've been in this work for manyyears.
I've been doing it since theearly days, you know, just
toying around with Google stuffwhen it was just TensorFlow and
some Taxanet and all thesedifferent AI startups back then.
I want to say that um IBM threwout Watson.
(01:29):
I mean, all those early days, Iwas jumping in and seeing what I
could get into in Google Labs orwherever else just to learn and
just to grow.
And so as I started learningabout LLMs and more, I fell in
love, saw where it was going,threw some stuff that was code
in the safe, saved it for later.
And of course, as it synergizedwith what we're doing for the
(01:52):
last four and five years inRoar, recognized over the last
two, no, it's been three years,Brandon.
Three years that we had to bringin what we're doing.
So we're we've been a long timein this, a lot longer than this
conversation seems.
It's gonna pop out of nowhere.
Um, we've only been talkingabout AI in relation to
(02:13):
blockchain in the last year anda half, two years, and how we're
programming AI, training it,giving it rails with blockchain
technology and NFTs and thatkind of stuff, um, which is
really cool.
You're able to give apersonality or whatever, but
this is something deeper.
We're digging into thephilosophies behind where AI is
going, the compute andeverything else.
(02:34):
And so, with that said, youknow, this podcast is a little
out of nowhere, but it's not outof nowhere.
It's come from a lot of researchand thought, whatever.
And Brainan and I are going togo through this.
Of course, I'm your other host,Dustin.
Um, and this topic is AI isbuilding power plants, and we're
building a network with FierceLabs.
(02:55):
And so um we are right in thecenter of AI development with
our own LLMs, and you'll hearvery soon SLMs.
And we want to dig into aphilosophy that we feel like
everyone could actually, youknow, gain value from.
And so, with that said, I wantto pass off to one of my best
(03:17):
friends in the world, BrandonBillings.
And so, Brandon, jump in, let'sdo this thing.
SPEAKER_00 (03:22):
All right, Dustin.
So I'm just gonna say it.
Every AI company on earth isracing to build bigger models,
bigger data centers, biggereverything.
And you're over here saying thatthat's wrong.
That's not just wrong, friend.
That is dangerouslyshort-sighted.
(03:44):
Uh, yeah.
So we're coming in hot today,guys.
Uh good.
We should be.
SPEAKER_01 (03:49):
Here's the thing: what's happening right now isn't
just AI innovation, it'sinfrastructure escalation.
And you guys, it's starting tohit my backyard.
I don't know if you're havingone of these big old data
centers built on your lovelywaterways where you've got
protected animals living, butit's hitting our areas in the
worst ways possible.
(04:11):
So we're literally buildingindustrial scale intelligence
factories, which is ridiculous,especially when you think about
folks like Harver Mead andothers that say at some point
we're gonna top out what computecan even do when it comes to the
current physical setup of a chipthat we have or GPUs or CPUs.
So we're we're inching this outto get something that's
(04:33):
basically um only one-third ofthe compute power of a human
brain.
SPEAKER_00 (04:40):
Whole factories.
That sounds dramatic.
Well, it should.
SPEAKER_01 (04:49):
Oh my gosh, it's so fun.
Here's why trillion parametermodels don't just run in the
cloud, they've got to run ingigawatt data centers.
They consume massive watersupplies, folks.
They depend on rare earthextraction, they reshape energy
grids.
(05:09):
This is not software scaling,this is ecological scaling.
Someone has to come up with abetter idea.
So it's gonna keep growing untilsomeone says, stop, time out.
Does this make sense?
We're getting less return forall of this build, and we've
exploited the environment.
At the end of the day, guys,let's just put it this way it's
(05:30):
gonna hit your energy bill, it'sgonna hit your water bill, it's
gonna hit your environment.
Your backyard will be eat up bywhat's coming.
We've been doing the math onwhere we have to go based on
current systems, solutions,software, builds,
infrastructure, and more, and itain't good, friends.
It ain't good.
SPEAKER_00 (05:49):
All right, Dustin.
So let's break this down.
What's the real problem?
SPEAKER_01 (05:54):
We're in an AI arms race, and the AI arms race is
simple.
The idea is more parameters,more GPUs, more power, and if
your model isn't bigger thanmine, you lose.
SPEAKER_00 (06:11):
So it's like nuclear escalation, but for compute.
Yeah, man.
SPEAKER_01 (06:16):
I feel like we're like drawn from the Reagan era
and the 80s.
But it is exactly, exceptinstead of uranium, it's GPUs
and water.
And let's be clear.
Also, rare earths are involvedin this because that's what
these semiconductors andeverything are built out of.
You know, you got gold andeverything else in there.
(06:38):
So, hello, friends.
This is not just aninfrastructure situation, but in
America in the long run, you'reactually talking about something
that affects the very safety andsecurity of our nation.
We now have data centers thatconsume as much electricity as
small cities.
There are places on the planetthat still don't have
electricity.
I go to those places.
(06:58):
I've stayed in mud huts, I'veworked with people that don't
have electricity, I've workedwith people that are using even
hybrid solutions to get somekind of grid set up.
I've been in the far reaches ofAfrica, Asia, I've been in
jungles, I've been in SouthAmerica.
I have been these places, placeswhere people don't have much.
And while they're still not evenin the now world, we're using up
(07:19):
so much electricity with a datacenter that it compares with a
small city in America.
This is ridiculous.
Cooling systems, they're burningthrough billions of gallons of
water annually.
And we're pretending this isjust software progress.
You know, Congress is worriedmore about the idea that AI will
take over and create a Skynetand we're gonna have the
(07:39):
terminator.
And really, the big problemisn't even all that.
We're missing it.
It's the impact if we don't stopon our environment and more.
And I'm not an environmentalist,I'm a steward.
I care about my kids and theirkids and on in generations, and
this is not sustainable.
If the trajectory with thesystems and the solutions we
(07:59):
have, it's not sustainable.
However, there is a way to reseethat.
SPEAKER_00 (08:05):
Uh yeah, so AI isn't just reshaping the internet,
it's reshaping physicalinfrastructure.
Brother, it is.
SPEAKER_01 (08:12):
I mean, every time someone generates something on,
generates an image that's inthat cloud, the compute
resources they're using are likehundreds of times more than a
human being would use to createa graphic.
Great.
You save time and you cut agraphic designer out of your
(08:33):
work.
Good for you, but at what cost?
We gotta ask that.
And if this continues for 30 to50 years on this trajectory, you
don't get smarter AI, you get AIoligopolies that are controlling
energy, compute, and access.
And believe me, it's alreadygoing on because we got guys
trying to get to Congress sayingwe need to stop everyone else
(08:56):
small out there with AIs becausethey're gonna create a bad one.
It's gonna have agency and it'sgonna terminate us.
And that is bull crap.
I've said over and over that AIis not smart enough to be
creative on its own.
And I mean it.
I mean it.
I challenge someone to show mewhere it has had any kind of
cognizance, it is conscious ofitself, it's self-aware or
(09:20):
creative.
It has only been able to do whatcreators, oracles that are
humans, are impacting in it.
And so, no, worrying less aboutTerminator agency where it's
showing some level of awareness.
I I call uh I won't even say aname.
I want to say, well, I'll callAnthropic and Cloud on the
carpet, open AI, whoever, thedoomsayers.
(09:41):
I call on the carpet because allthey're trying to do is cut out
the smaller players.
And if you do that, you do thatin 30 to 50 years, you won't
have that smarter AI.
You're gonna have dumber AI andless environment, and you're
gonna regret it.
SPEAKER_00 (09:57):
Wow.
Uh, so what's the alternative?
You can't just slow down AI andits development.
No, we do not slow it down.
SPEAKER_01 (10:06):
So people in their brains, they think just because
we start asking, should we dosomething, it means we're gonna
run a lot slower because we'renot doing what we can do.
So we're not just asking, can wedo it and sprinting?
What we're doing is asking,should we do it, and then
sprinting.
So the sprint is after the rightdirection.
(10:29):
Let me tell you something.
One degree, one degree off intrajectory on a ship in the
ocean is a massive problemwhenever they're traveling from
the far reaches of one continentto another.
The difference can be massive.
Leaving Asia, coming to the US,thinking that you're going to
(10:52):
Tijuana, Mexico or whatever, orsome port off the edge of
Mexico, and ending up inWashington State because you're
one degree off when you started.
See, here's the problem.
We are what we're not one degreeoff.
We're a hundred degrees off inAI.
Slow down and rethink to changethe trajectory because we're not
(11:14):
going to end up where we want.
We're not using the mindset thateven folks like Stephen Covey
have called out for years.
Carnegie, Covey, and others,some of the greatest thinkers,
have said that you start withthe end in mind.
So we need to start with the endin mind and go backwards from
that and get that trajectoryright in the first place.
And so I'm just saying timeout,pause in the conversation.
(11:35):
Let's reframe the trajectory andget it right right here.
My kids are gonna pay thisprice.
And so I get tired of owners ofbusinesses and government that
have no kids because they don'thave a stake in this.
I've got kids, man, and my kids,grandkids, and on are gonna have
to pay the price for the lack ofstewardship right here.
(11:56):
So we don't slow it down.
We just we just we flip it andthen we run.
Instead of building one giantbrain in a warehouse, I mean,
seriously, the warehouses havewhat, 33% of the compute of a
human brain?
Great job.
You got one third of a human,and you're draining our
resources on the planet.
Way to go, big Google.
(12:18):
Way to go, big mega, meta,Facebook, whatever.
Way to go, anthropic and cloud.
Great job.
You got a third of a humanbrain, you're placed a bunch of
workforce, you've screwed aneconomy, and yet you're gonna
deplete all the resources andsay that it's progress?
My gosh, come on, let's get halfa brain.
(12:39):
So let's build millions orbillions of smaller brains
everywhere.
SPEAKER_00 (12:50):
Well, I think that's starting to push past the
philosophical.
Why don't you go tell us whatthis actually means?
It's a quiet revolution in AI.
SPEAKER_01 (13:00):
And I've been talking about it for a while.
I talked at it at uh Cosm thislast fall, just a little bit.
I just put inklings out there.
I'm out under some pseudonyms onX.com.
Yes, I do that.
And I just mess withconversations because it's fun.
I like to figure out what peopleare thinking and doing.
And I've been feeding thisconversation, this idea that
(13:22):
there could be a quietrevolution in AI development.
And it would mean instead ofLLMs, which are large language
models, we start thinking aboutsmall language models, not
trillion parameter monsters.
Models that are like 500 millionto 7 billion parameters, they're
(13:45):
highly specialized, highlyefficient, laser focused.
I mean stuff that doesn't takeup data centers.
It takes up like three to sevengigs of space.
So working in my space, I seethe size of models because I'm
having to work with them in thelocal.
We're building an AI in ourrural world, we're building
(14:06):
localized versions of AI that wecan run on a cell phone.
I've said that for a year now,and I mean it.
We ran AI on bare metal rails onApple literal laptops, the last
versions before Apple said, Hey,we built stuff in there that
allows you to do this.
We were running on it beforethey did it, which is kind of
(14:28):
how we roll.
So we want these models to behighly specialized, highly
efficient, and laser focused,and then they actually become
that one smart, you know, chunkof brain out there that others
can call on.
SPEAKER_00 (14:47):
So you're saying they wouldn't know everything,
but they would know one thingextremely well.
Extremely well.
SPEAKER_01 (14:54):
Stop trying to make AI self-aware and start using AI
where it's very powerful.
AI is the one of the mostpowerful um human replicating
things that can replicate how ahuman thinks, similarly.
It is incredible at learning,it's incredible at storing
information, and the algorithmsallow it to communicate the way
(15:18):
the data store works, it canpull information and
reconstitute it.
Can it create?
No.
So why are we trying for that?
Why is that what we're after?
Why is the goal to createsomething that's self-aware and
has its own human-like agency?
Why aren't we utilizing thesethings for what they're really
great at?
(15:39):
And those SLMs, they they don'tneed a trillion-parameter model
to audit a smart contract.
You don't need it to analyzeDeFi risk.
You don't need it to managestorage allocation.
You need precision.
I'm often having to dumb downmodels that I'm using from other
LLMs.
(16:00):
I have to dumb them down andfocus them in applications I
build.
There's a lot of applications wehave not released yet, and we
are in the process releasingthis year that take not just our
model, but other models that areout there that are LLMs or SLMs
or hybrids, and we focus them inon just one thing.
And this is where it's becomemagic because we can size those
(16:21):
things down and get them reallysmart.
These these things can createprecision, and that's important
when you're dealing with modelsout there that are honestly
biased.
You can cut out bias with someof the things we've built and
how we're doing it.
And here's the kicker SLMs,small language models, can use
(16:41):
10 to 100 times less energy inthis process.
Think about that.
And it just makes sense insteadof running through every
eventuality to come back to thatone precise thing you want, it's
going already scaled down andit's getting even more precise.
SPEAKER_00 (17:01):
So, I mean, there you go.
That's amazing, Dustin.
I I would say that's not even anincremental change, but that's
literally a disruptive way ofthinking about it.
That's it.
That's it.
SPEAKER_01 (17:13):
And we've always thought this way with stuff
we've built.
Um, because when you build on ashoestring and you don't have
big VC investment, and you buildwith limited teams and you build
where people have to be reallycritical, and they know that
this is make it or break it, youget really smart at how to make
things work better for you.
And that's what we do.
(17:33):
Like we really do that well withwith Fears Labs and Roar, and it
is disruptive.
And not only that, SLMs can runon edge hardware, on consumer
GPUs, on distributed nodes, on aon a Google cell phone.
Don't tell me it can't becauseI've done it.
I'm telling you, it's amazing.
That means we stop building AIslike skyscrapers and start
(17:57):
building them likeneighborhoods.
SPEAKER_00 (18:01):
Well, let's talk about what nobody wants to talk
about, and that's theenvironmental cost.
SPEAKER_01 (18:07):
Well, I mean, if we continue this trajectory,
exponential GPU production, rareearth mining expansion, water
depletion and drought zones,grid instability already being
proven.
Let's not even get in the factthat everyone has a data center
nearby, is talking about thefact that their water bill and
their utility bill go up, up,up.
(18:29):
So grid instability, landconversion, which really bothers
me having grown up on a farm andunderstanding food supply.
We could literally see AIinfrastructure rival aviation
emissions long term.
And I'm saying long-term here,but the truth is that's
conservative.
I'm talking about it's reallyshort term, less than 30 years.
I'm just saying from the mathwe're seeing and and the the
(18:52):
sheer push for big, big, bigcentralized, that's what we're
seeing.
That's amazing.
It's unnecessary becausedistributed SLM architecture
changes the equation.
Lower power density, smallerclusters, passive cooling, idle
hardware use.
(19:13):
There are some amazing Web3 guysout there already sharing their
processing power.
I've been a part of it.
Sharing processing power touniversities, GPUs and CPUs on
unused hardware that they justhave plugged in that's asleep
for the most part, running inthe background so that the
universities can use their LLMsand algorithms and more to come
(19:35):
up with things, for instance,like solving AIDS and cancer
research.
So using idle hardware across alot of folks that are
participants, shifting workloadsto renewable, heavy, heavy
regions instead of megawattcampuses.
We get we're gonna get insteadof those big massive megawatt
campuses, we're gonna getmicrowatt participation.
SPEAKER_00 (20:02):
So this is where people get skeptical.
SPEAKER_01 (20:11):
Yeah, now I'm gonna scare everybody off real quick.
Um blockchain web three crypto.
There it is!
SPEAKER_00 (20:17):
Run for your life!
SPEAKER_01 (20:19):
No.
Um actually, our belief ofFierce Labs and Roar is that um
blockchain crypto web 3 has beenlong um something that people
have pumped and dumped orwhatever.
We built in here forever.
We've invested, we've been inother projects, and uh, and
let's just be really honest.
A lot of it has been solutionslooking for problems.
I've said that for years,plural.
(20:41):
However, if we stop, just justjust put all that on the shelf
and let's not call it blockchaincrypto, whatever, and let's call
it web3.
And just from our last podcast,we talked about the fact that
web two could really benefitfrom web 3's Rails.
So if we built this in a railworld, um, in whether it's
(21:02):
blockchain or whatever you wantto call it, hybrid lime wire,
napster, whatever, P2P,something similar, but that pays
people back for beingparticipants, which is the Web3
idea.
Decentralized storage,decentralized compute with a
benefit to the participants forusing micro amounts of their
(21:25):
machines, their storage, theirGPU, their CPU.
If you see it like that insteadof blockchain, crypto, whatever,
if you see it as decentralizedcompute and storage, which is
what we're talking about, with areturn for the participants, you
can see that this could actuallycoordinate it.
There's a difference.
If millions of people arecontributing storage and compute
(21:48):
through a Web3 model that AI isbuilt into, well, you already
need verifiable accounting,performance tracking, and
trustless reward systems.
And that's what What Web3 isgood at.
You get it?
So that is really some of thecore logic of Web3.
SPEAKER_00 (22:08):
So this goes beyond speculation.
It's really about coordination.
Yeah, exactly.
SPEAKER_01 (22:14):
It's it's where systems and ecosystems like Roar
come into play.
And we're we're actually goingto prove it because we're we're
already building it.
I hate putting stuff out thereand giving people an edge on us,
but I just felt like theconversation's gone so fast and
so far that if we didn't wedidn't stake uh a marker in the
dirt and put up a flag somewhereand call it out before we kept
(22:37):
going, that either legislatureor some global elite body or
oligarchs or something are goingto control the narrative and
we're gonna be on a collisioncourse with an eventuality that
we believe is absolutelychangeable right here, right
now.
SPEAKER_00 (22:54):
All right, Dustin.
So how does Roar fit into this?
Well, Roar isn't about chasingtrillion parameter vanity
metrics.
SPEAKER_01 (23:02):
Yeah, I called that out.
So yeah, I'm being ugly on this.
I I can be.
I don't have a pony in theparade that you can see, so
who's gonna pick on me, right?
I don't matter.
But what we have under hooddoes.
We're building distributedinfrastructure, edge native
intelligence, decentralizedcoordination layers, tokenized
incentive systems.
(23:23):
And we've been here for fiveyears doing it.
We kind of got a good idea whatwe're doing, let alone all the
years prior from my early daysin Bitcoin going back to 2010, I
think it was, and then beyondthat, and social media and
platforms that go all the wayback to Zanga.
And yes, that's at the turn ofthe last century, my friends.
I've been here a minute.
(23:43):
I've been in web one, web two,and now web three.
And so we have a touch on allthis.
It's a little bit different fromothers.
I don't know how many others,when they went to college, got
to mess with the ARPANET, butI'm just saying we have a
different view on things becausewe've been here a minute.
So we get some of thisinfrastructure, things that can
change, coordination layers,tokenized incentive systems,
(24:04):
distributed infrastructure, edgenative intelligence.
Our vision of decentralizedstorage, compute participation,
and SLM orchestration is nottheoretical.
I got like four white papers tochase this in a minute after
this one.
We've been building towards thislogic for a long time.
We've been moving here the wholetime.
It's been in our roadmap for our10-year plan, and we just ain't
(24:27):
told nobody about it until now.
So that's the direction theindustry has to move to if it
wants to survive long term.
I'm just saying, if we don'twant to become the Matrix or
Skynet or others, then weprobably want to move this way.
And I still don't believe inthat kind of agency and
creativity at that level.
I'm just making fun of everyoneout there because we may not be
in a place where AI takes overand puts humans as servants, but
(24:50):
we definitely are going to be ina place where we've used up all
of our ecological resources, ourbeautiful beaches, waterways,
our animals, our plants, ourcommunities, our freshwater, our
open land, our agrarian society,and the fact that America
represents 25% of the productionof all the food on the planet,
(25:10):
and we're buying up farms toturn into data centers, which
pisses me off.
SPEAKER_00 (25:17):
That's what we have to do to survive long term.
I would say, how do you reallyfeel about it?
But I think we know.
Um, so Roar's not just buildinganother hyperscale facility,
right?
We're building a mesh.
Exactly.
SPEAKER_01 (25:32):
So horizontal intelligence layer, not vertical
concentration.
I've been against centralizationsince the beginning.
It's in our white paper we justlaunched even a few weeks ago,
but it predates that becauseit's core for Web3 values from
the beginning, from the earlydays of Bitcoin's white paper,
(25:52):
that has been the bad guy,centralization.
And so we're just not about thatvertical concentration.
We are about a horizontalintelligence layer.
I think everyone should playnice with each other in smaller
chunks that are well positioned.
We've already got leaders inspecific kinds of AI.
Why don't you stop trying tofight with each other and
(26:14):
collaborate and then allow otherparticipation through meshes?
Why not have your idiot savantsall over the place that only
know one thing and then offloadprocessing between like a mesh?
We already think about mesh withIoT connecting services like
Amazon Sidewalk.
(26:34):
Nobody wants to talk about thefact that that works.
But why are we not doing in thesame kind of thinking this with
the processing for AI or more?
Um, we got to end the verticalconcentration, we got to end the
centralization, we've got to endthe obliteration of our natural
resources, our communities, ourutilities, our grids, and get
(26:55):
back to the idea that therecould be a democratization of
all of it and recognize as well,even in that, Web3 has a better
solution for making sure thingsare not biased, they're they are
censorship resistant, and theyhave the right rails and
firewalls.
We can build those into ablockchain, and they are
ledgerized.
(27:16):
And I've taught everyone, I'mthe first one I know of that's
done it, that you can take an AImodel and train it to
communicate with an individualholding an NFT on a wallet as if
it's that personality and dothings specific to it.
I have done that and made it soit can learn back with itself
with some of the biggest modelsout there.
And you guys, if you're at oneof those biggest models, you've
(27:38):
probably seen some of my work.
I'm just saying that my friends,we've been doing this whether
you knew who we were or not.
We've been on the front end ofit, and I'm telling you, it
works.
We're not talking about theory,we're talking about function.
SPEAKER_00 (27:53):
All right, Yoda, paint the big picture, show us
the way of the Jedi.
Jedi ways is what we're nothere.
SPEAKER_01 (28:01):
Okay, let's paint the big picture.
Imagine this millions ofparticipants worldwide.
Some provide storage, someprovide inference, some host
specialized SLMs, some verifyresults.
All of them are willingparticipants.
(28:22):
That's key.
Blockchain tracks performanceand rewards fairly.
We can call it something else ifyou don't like the name.
A ledger that's time-stampingactivity that keeps everybody
honest.
It's kind of a proof of what isverifiably true, right?
Compute routes dynamically torenewable heavy grids, idle GPUs
(28:47):
become more productive, andolder hardware gets extended
life.
SPEAKER_00 (28:57):
I'm just processing what you just said.
That's um it's pretty brilliant.
Uh what that means is instead ofAmazon building a hundred new
mega facilities, 10 millionpeople participate in AI
infrastructure.
You know, that's the shift.
SPEAKER_01 (29:15):
It's from centralized dominance to
distributed intelligence.
Now, the problem here is, nooffense to you AI developers out
there, um I'd say that there'sum there's probably um well,
I'll just say it straight up.
I think there's a real lack ofhumility.
(29:36):
So I don't see a lot of themsaying to their boards and their
teams, hey, how about we share?
You know, in kindergarten welearn that sharing is caring,
and we learn to keep our handsto ourselves.
And yet these guys are grabbingeach other's technologies and
lying about it, and thenfighting over who said what and
did what first, and then runningto Congress and going, stop
(29:57):
them.
They're trying to do somethingbad.
And you guys, you sound like youneed to go back to kindergarten.
And then if you really want toget in there, you're there's a
lot of you that say you'renonprofity, but you're really
profity-y because you gotshareholders of VCs and
everything else, and yourhumility is at an all-time low,
and you think that you're theonly one that can get us there.
Good for you.
(30:17):
My belief is that there are amillion geniuses out there, and
that we need to evendecentralize the land claim to
genius status and AI and alloweveryone to be a part of
distributed intelligence.
It's safer, it's better, it'smore humble, and in the long
run, I want to just put outthere it can use older
(30:39):
technology and more.
Think about what this could dofor economies.
So you're over in Africasomewhere, or let's say
Greenland, it's cold there.
That's good for compute, right?
So you don't even need to havecoolers, you've got an Iceland
mass.
And you're over there, and we'rehearing about how they've been
challenged with finance becauseDenmark ain't doing right and
(31:00):
the US is involved.
I got nothing to say about that,except wouldn't it be great if
they could just run some oldcomputer hardware in a super
cold environment and getresources because they're paid
to share their GPU and CPU?
People in mud huts with you knowa solar panel and cellular
signal can actually be aparticipant in the stuff we're
(31:22):
building and be a part ofcompute, sharing the load across
everywhere, but alsodemocratizing the control and
the status, as well as financingeconomies that would have never
gotten a touch any part of biginfrastructure or money in this
space.
(31:42):
Guys, this is a different kindof thinking.
SPEAKER_00 (31:46):
Yeah, this makes me think of that quote, a rising
tide raises all ships, right?
When everybody participates,it's beneficial to everyone.
It's amazing.
So that's let's let's take alook.
We're at that dramatic fork inthe road, then aren't we?
SPEAKER_01 (32:04):
Yeah.
Um, I didn't want to do thispodcast.
I didn't want to do this call.
I'm way ahead of where I want tobe with the other white papers,
so I took them, ripped themapart, made another white paper
on this, and then and none ofthem are out, and then we we've
turned it into this conversationpodcast, and I really didn't
want to do it.
(32:25):
But I'm running out of time.
Um, I've been in this longenough to realize that there's
some deadlines on stewardship.
Someone has to get out here andsay this.
Somebody has to have theconversation or at least start
the narrative or at least openup for a forum of discussion.
And if that means somebody beatsus to the punch and they build
better than us or whatever,great, then I've just saved my
(32:46):
kids' problems.
I would love to be thebeneficiary and my community be
the beneficiary of all of whatwe're talking about here as well
as all the build we're doing.
But if somebody beats me thereand it saves my backyard so that
I'm not overtaken by a datacenter, God bless them, go for
it, you know?
So that's why we're doing this.
And it's because we see thesetwo futures.
(33:07):
There's a fork in the road, andthere's future A, and it's
megadata centers dominate, AImonopolies control access,
governments can turn things onand off of censorship,
environmental pressuresescalate, energy grids will be
strained, and then corporateconcentration will intensify.
(33:28):
With that comes all kinds ofthings.
Like I said, censorship, bias,control, and worse.
Then there's future B.
Specialized SLM mesh networks,which in a sense can be
firewalled, controlled, and madewith breakers they can get
(33:50):
turned off.
Community nodes, blockchaincoordination of some sort,
storage, decentralized storagecoordination, whatever.
Lower environmental footprint,distributed revenue, and global
participation.
SPEAKER_00 (34:09):
Wow.
So one future centralizes power,the other decentralizes
intelligence.
Exactly.
SPEAKER_01 (34:17):
And over 50 years, what's what's the result?
The environmental differencecould mean gigatons of carbon
dioxide.
For folks that care about that,that's massive.
The economic difference couldmean trillions redistributed
instead of consolidated.
So trillions redistributed.
Oh, I can't talk.
(34:39):
Redistributed, there comes mysouthern, instead of
consolidated.
And so, I mean, that economicimpact for anyone participating
in countries that are open toit, it'd be massive.
Folks that have minimalhardware, a cell phone or more,
a Raspberry Pi or more, canparticipate at some level in the
(35:02):
layers of these systems.
Um, that's what we're building,and that's where we're going.
SPEAKER_00 (35:09):
Uh that's amazing, Dustin.
Um, so the current AI industryis building bigger buildings.
And you're saying the future isactually to build a bigger
network.
Yeah.
We don't need bigger models andbigger warehouses.
SPEAKER_01 (35:25):
We need smaller intelligence everywhere.
SPEAKER_00 (35:28):
So AI doesn't scale by becoming gigantic, it scales
by becoming distributed.
The more and more distributed,the better.
And that's it.
SPEAKER_01 (35:37):
That really is the core.
We build larger impact bybuilding smaller systems
together.
We always say we're bettertogether.
Both sides of the aisle inCongress will say we're better
together, but how do they act?
Here's the truth.
In this, we are better together.
Somebody's got to say it outloud.
Somebody needs to encourageeveryone to at least, at least
(35:58):
take a look at it.
You know, Elon, at least thinkabout the fact that you could
offload, and I've said this overand over for the last year, you
could offload process power fromyour centralized data center for
your XAI.
You could offload it to thecyber trucks that aren't being
used.
You got a whole fleet out theresomewhere, or even run in the
background.
If you've already got the AI inthem, why isn't it running in
(36:20):
the background utilizing some ofthe compute power back to the
the grid of all of them?
I mean, it could be a part ofthe agreement when you get one.
Why not?
I'm telling you, it's a wholedifferent thinking and it would
have a bigger impact and betterresults.
SPEAKER_00 (36:33):
So it's we build bigger by building smaller.
Yeah.
Um, so that's not anti-progress,it's just a better way of
thinking about it.
It's smarter progress.
SPEAKER_01 (36:46):
And if we don't shift, I told you about that
being a degree off on a shipwhen you start that journey
across the ocean, can throw youoff on a whole different place
on the planet than you aremeaning to go.
If we don't shift now, AIbecomes another centralized
industrial complex.
I'm calling it now.
(37:07):
And let's be honest, others havecalled it before me.
Don't read Isaac Asimov, youdon't be scared out of your
wits.
I've read almost everything theman's written on purpose.
AI becomes that centralizedindustrial complex, and if we do
do this, if we do shift, it'sgonna become a global mesh of
intelligence instead.
(37:27):
Horizontal, not vertical, anddecentralized, not concentrated.
SPEAKER_00 (37:33):
So we need to build smaller.
SPEAKER_01 (37:35):
Yep, there it is, to build bigger.
That's it.
That's it.
And brother, thanks for being ontoday.
Folks, thanks for listening.
Um we did not want to be thebearer of this news, but if no
one else is gonna say it, we'regonna roar about it.
Thanks for being on your weeklyroar.
Know that this does mean thereis a white paper and coming.
(37:58):
And so um over the next weekyou'll see it publicized or
whatever through our channels.
And as well, we're also givingyou a sneak peek into what's
being developed and alreadyeither at market that we're in
alpha testing on with ourinitial alpha partners, or we're
leaning into bringing into betafor you to use.
(38:21):
Um, it's time we're here, andyou're gonna hear more and more
about it.
So thanks for being on.
Be sure to follow us in all theplaces you find us, whether it's
X, YouTube, Rumble, or others.
And then stay connected, guys.
There's so much news coming out.
March is based, April's gotanother one.
April's gonna be hot.
May's got a we've got slogansfor each one.
(38:42):
You're gonna want to be a partof it all the way through the
rest of this year.
It's gonna get funner.
Brayden, thanks for being on.
Everybody, thanks for joiningwith us.
Y'all have a great one.
Thanks, everybody.
Take care.
Welcome back, my friends, for another episode of
your weekly roar.
Folks, I don't know if you knowthis, but you've been with us
closing in on our first year.
So I want to say this is intoour we're getting closer to our
30th episode real soon.
And with a couple weeks off hereand there, this your weekly roar
has been growing.
Also, hope you've seen thegrowth in our topics.
(00:24):
Um, from Roar and Fierce Labs,we've been really going after
some key, key, key conversationsfor the future.
Future blockchain, web3, ai,decentralized storage, and more.
And so um we're rolling outwhite papers one right after the
other.
And with that said, on theheels, or actually out in front
(00:45):
of a white paper that's comingyour way on the AI space, one of
the first of probably four,we're doing this podcast today.
So um, guys, it's a littledifferent again.
Um, I'll say this that uh thisis definitely a stretch for us
to get out in front of somethingand go the opposite direction of
(01:07):
where most development andmarket is going.
Uh, but we feel like we need toweigh in at this point.
We've been in this work for manyyears.
I've been doing it since theearly days, you know, just
toying around with Google stuffwhen it was just TensorFlow and
some Taxanet and all thesedifferent AI startups back then.
I want to say that um IBM threwout Watson.
(01:29):
I mean, all those early days, Iwas jumping in and seeing what I
could get into in Google Labs orwherever else just to learn and
just to grow.
And so as I started learningabout LLMs and more, I fell in
love, saw where it was going,threw some stuff that was code
in the safe, saved it for later.
And of course, as it synergizedwith what we're doing for the
(01:52):
last four and five years inRoar, recognized over the last
two, no, it's been three years,Brandon.
Three years that we had to bringin what we're doing.
So we're we've been a long timein this, a lot longer than this
conversation seems.
It's gonna pop out of nowhere.
Um, we've only been talkingabout AI in relation to
(02:13):
blockchain in the last year anda half, two years, and how we're
programming AI, training it,giving it rails with blockchain
technology and NFTs and thatkind of stuff, um, which is
really cool.
You're able to give apersonality or whatever, but
this is something deeper.
We're digging into thephilosophies behind where AI is
going, the compute andeverything else.
(02:34):
And so, with that said, youknow, this podcast is a little
out of nowhere, but it's not outof nowhere.
It's come from a lot of researchand thought, whatever.
And Brainan and I are going togo through this.
Of course, I'm your other host,Dustin.
Um, and this topic is AI isbuilding power plants, and we're
building a network with FierceLabs.
(02:55):
And so um we are right in thecenter of AI development with
our own LLMs, and you'll hearvery soon SLMs.
And we want to dig into aphilosophy that we feel like
everyone could actually, youknow, gain value from.
And so, with that said, I wantto pass off to one of my best
(03:17):
friends in the world, BrandonBillings.
And so, Brandon, jump in, let'sdo this thing.
SPEAKER_00 (03:22):
All right, Dustin.
So I'm just gonna say it.
Every AI company on earth isracing to build bigger models,
bigger data centers, biggereverything.
And you're over here saying thatthat's wrong.
That's not just wrong, friend.
That is dangerouslyshort-sighted.
(03:44):
Uh, yeah.
So we're coming in hot today,guys.
Uh good.
We should be.
SPEAKER_01 (03:49):
Here's the thing: what's happening right now isn't
just AI innovation, it'sinfrastructure escalation.
And you guys, it's starting tohit my backyard.
I don't know if you're havingone of these big old data
centers built on your lovelywaterways where you've got
protected animals living, butit's hitting our areas in the
worst ways possible.
(04:11):
So we're literally buildingindustrial scale intelligence
factories, which is ridiculous,especially when you think about
folks like Harver Mead andothers that say at some point
we're gonna top out what computecan even do when it comes to the
current physical setup of a chipthat we have or GPUs or CPUs.
So we're we're inching this outto get something that's
(04:33):
basically um only one-third ofthe compute power of a human
brain.
SPEAKER_00 (04:40):
Whole factories.
That sounds dramatic.
Well, it should.
SPEAKER_01 (04:49):
Oh my gosh, it's so fun.
Here's why trillion parametermodels don't just run in the
cloud, they've got to run ingigawatt data centers.
They consume massive watersupplies, folks.
They depend on rare earthextraction, they reshape energy
grids.
(05:09):
This is not software scaling,this is ecological scaling.
Someone has to come up with abetter idea.
So it's gonna keep growing untilsomeone says, stop, time out.
Does this make sense?
We're getting less return forall of this build, and we've
exploited the environment.
At the end of the day, guys,let's just put it this way it's
(05:30):
gonna hit your energy bill, it'sgonna hit your water bill, it's
gonna hit your environment.
Your backyard will be eat up bywhat's coming.
We've been doing the math onwhere we have to go based on
current systems, solutions,software, builds,
infrastructure, and more, and itain't good, friends.
It ain't good.
SPEAKER_00 (05:49):
All right, Dustin.
So let's break this down.
What's the real problem?
SPEAKER_01 (05:54):
We're in an AI arms race, and the AI arms race is
simple.
The idea is more parameters,more GPUs, more power, and if
your model isn't bigger thanmine, you lose.
SPEAKER_00 (06:11):
So it's like nuclear escalation, but for compute.
Yeah, man.
SPEAKER_01 (06:16):
I feel like we're like drawn from the Reagan era
and the 80s.
But it is exactly, exceptinstead of uranium, it's GPUs
and water.
And let's be clear.
Also, rare earths are involvedin this because that's what
these semiconductors andeverything are built out of.
You know, you got gold andeverything else in there.
(06:38):
So, hello, friends.
This is not just aninfrastructure situation, but in
America in the long run, you'reactually talking about something
that affects the very safety andsecurity of our nation.
We now have data centers thatconsume as much electricity as
small cities.
There are places on the planetthat still don't have
electricity.
I go to those places.
(06:58):
I've stayed in mud huts, I'veworked with people that don't
have electricity, I've workedwith people that are using even
hybrid solutions to get somekind of grid set up.
I've been in the far reaches ofAfrica, Asia, I've been in
jungles, I've been in SouthAmerica.
I have been these places, placeswhere people don't have much.
And while they're still not evenin the now world, we're using up
(07:19):
so much electricity with a datacenter that it compares with a
small city in America.
This is ridiculous.
Cooling systems, they're burningthrough billions of gallons of
water annually.
And we're pretending this isjust software progress.
You know, Congress is worriedmore about the idea that AI will
take over and create a Skynetand we're gonna have the
(07:39):
terminator.
And really, the big problemisn't even all that.
We're missing it.
It's the impact if we don't stopon our environment and more.
And I'm not an environmentalist,I'm a steward.
I care about my kids and theirkids and on in generations, and
this is not sustainable.
If the trajectory with thesystems and the solutions we
(07:59):
have, it's not sustainable.
However, there is a way to reseethat.
SPEAKER_00 (08:05):
Uh yeah, so AI isn't just reshaping the internet,
it's reshaping physicalinfrastructure.
Brother, it is.
SPEAKER_01 (08:12):
I mean, every time someone generates something on,
generates an image that's inthat cloud, the compute
resources they're using are likehundreds of times more than a
human being would use to createa graphic.
Great.
You save time and you cut agraphic designer out of your
(08:33):
work.
Good for you, but at what cost?
We gotta ask that.
And if this continues for 30 to50 years on this trajectory, you
don't get smarter AI, you get AIoligopolies that are controlling
energy, compute, and access.
And believe me, it's alreadygoing on because we got guys
trying to get to Congress sayingwe need to stop everyone else
(08:56):
small out there with AIs becausethey're gonna create a bad one.
It's gonna have agency and it'sgonna terminate us.
And that is bull crap.
I've said over and over that AIis not smart enough to be
creative on its own.
And I mean it.
I mean it.
I challenge someone to show mewhere it has had any kind of
cognizance, it is conscious ofitself, it's self-aware or
(09:20):
creative.
It has only been able to do whatcreators, oracles that are
humans, are impacting in it.
And so, no, worrying less aboutTerminator agency where it's
showing some level of awareness.
I I call uh I won't even say aname.
I want to say, well, I'll callAnthropic and Cloud on the
carpet, open AI, whoever, thedoomsayers.
(09:41):
I call on the carpet because allthey're trying to do is cut out
the smaller players.
And if you do that, you do thatin 30 to 50 years, you won't
have that smarter AI.
You're gonna have dumber AI andless environment, and you're
gonna regret it.
SPEAKER_00 (09:57):
Wow.
Uh, so what's the alternative?
You can't just slow down AI andits development.
No, we do not slow it down.
SPEAKER_01 (10:06):
So people in their brains, they think just because
we start asking, should we dosomething, it means we're gonna
run a lot slower because we'renot doing what we can do.
So we're not just asking, can wedo it and sprinting?
What we're doing is asking,should we do it, and then
sprinting.
So the sprint is after the rightdirection.
(10:29):
Let me tell you something.
One degree, one degree off intrajectory on a ship in the
ocean is a massive problemwhenever they're traveling from
the far reaches of one continentto another.
The difference can be massive.
Leaving Asia, coming to the US,thinking that you're going to
(10:52):
Tijuana, Mexico or whatever, orsome port off the edge of
Mexico, and ending up inWashington State because you're
one degree off when you started.
See, here's the problem.
We are what we're not one degreeoff.
We're a hundred degrees off inAI.
Slow down and rethink to changethe trajectory because we're not
(11:14):
going to end up where we want.
We're not using the mindset thateven folks like Stephen Covey
have called out for years.
Carnegie, Covey, and others,some of the greatest thinkers,
have said that you start withthe end in mind.
So we need to start with the endin mind and go backwards from
that and get that trajectoryright in the first place.
And so I'm just saying timeout,pause in the conversation.
(11:35):
Let's reframe the trajectory andget it right right here.
My kids are gonna pay thisprice.
And so I get tired of owners ofbusinesses and government that
have no kids because they don'thave a stake in this.
I've got kids, man, and my kids,grandkids, and on are gonna have
to pay the price for the lack ofstewardship right here.
(11:56):
So we don't slow it down.
We just we just we flip it andthen we run.
Instead of building one giantbrain in a warehouse, I mean,
seriously, the warehouses havewhat, 33% of the compute of a
human brain?
Great job.
You got one third of a human,and you're draining our
resources on the planet.
Way to go, big Google.
(12:18):
Way to go, big mega, meta,Facebook, whatever.
Way to go, anthropic and cloud.
Great job.
You got a third of a humanbrain, you're placed a bunch of
workforce, you've screwed aneconomy, and yet you're gonna
deplete all the resources andsay that it's progress?
My gosh, come on, let's get halfa brain.
(12:39):
So let's build millions orbillions of smaller brains
everywhere.
SPEAKER_00 (12:50):
Well, I think that's starting to push past the
philosophical.
Why don't you go tell us whatthis actually means?
It's a quiet revolution in AI.
SPEAKER_01 (13:00):
And I've been talking about it for a while.
I talked at it at uh Cosm thislast fall, just a little bit.
I just put inklings out there.
I'm out under some pseudonyms onX.com.
Yes, I do that.
And I just mess withconversations because it's fun.
I like to figure out what peopleare thinking and doing.
And I've been feeding thisconversation, this idea that
(13:22):
there could be a quietrevolution in AI development.
And it would mean instead ofLLMs, which are large language
models, we start thinking aboutsmall language models, not
trillion parameter monsters.
Models that are like 500 millionto 7 billion parameters, they're
(13:45):
highly specialized, highlyefficient, laser focused.
I mean stuff that doesn't takeup data centers.
It takes up like three to sevengigs of space.
So working in my space, I seethe size of models because I'm
having to work with them in thelocal.
We're building an AI in ourrural world, we're building
(14:06):
localized versions of AI that wecan run on a cell phone.
I've said that for a year now,and I mean it.
We ran AI on bare metal rails onApple literal laptops, the last
versions before Apple said, Hey,we built stuff in there that
allows you to do this.
We were running on it beforethey did it, which is kind of
(14:28):
how we roll.
So we want these models to behighly specialized, highly
efficient, and laser focused,and then they actually become
that one smart, you know, chunkof brain out there that others
can call on.
SPEAKER_00 (14:47):
So you're saying they wouldn't know everything,
but they would know one thingextremely well.
Extremely well.
SPEAKER_01 (14:54):
Stop trying to make AI self-aware and start using AI
where it's very powerful.
AI is the one of the mostpowerful um human replicating
things that can replicate how ahuman thinks, similarly.
It is incredible at learning,it's incredible at storing
information, and the algorithmsallow it to communicate the way
(15:18):
the data store works, it canpull information and
reconstitute it.
Can it create?
No.
So why are we trying for that?
Why is that what we're after?
Why is the goal to createsomething that's self-aware and
has its own human-like agency?
Why aren't we utilizing thesethings for what they're really
great at?
(15:39):
And those SLMs, they they don'tneed a trillion-parameter model
to audit a smart contract.
You don't need it to analyzeDeFi risk.
You don't need it to managestorage allocation.
You need precision.
I'm often having to dumb downmodels that I'm using from other
LLMs.
(16:00):
I have to dumb them down andfocus them in applications I
build.
There's a lot of applications wehave not released yet, and we
are in the process releasingthis year that take not just our
model, but other models that areout there that are LLMs or SLMs
or hybrids, and we focus them inon just one thing.
And this is where it's becomemagic because we can size those
(16:21):
things down and get them reallysmart.
These these things can createprecision, and that's important
when you're dealing with modelsout there that are honestly
biased.
You can cut out bias with someof the things we've built and
how we're doing it.
And here's the kicker SLMs,small language models, can use
(16:41):
10 to 100 times less energy inthis process.
Think about that.
And it just makes sense insteadof running through every
eventuality to come back to thatone precise thing you want, it's
going already scaled down andit's getting even more precise.
SPEAKER_00 (17:01):
So, I mean, there you go.
That's amazing, Dustin.
I I would say that's not even anincremental change, but that's
literally a disruptive way ofthinking about it.
That's it.
That's it.
SPEAKER_01 (17:13):
And we've always thought this way with stuff
we've built.
Um, because when you build on ashoestring and you don't have
big VC investment, and you buildwith limited teams and you build
where people have to be reallycritical, and they know that
this is make it or break it, youget really smart at how to make
things work better for you.
And that's what we do.
(17:33):
Like we really do that well withwith Fears Labs and Roar, and it
is disruptive.
And not only that, SLMs can runon edge hardware, on consumer
GPUs, on distributed nodes, on aon a Google cell phone.
Don't tell me it can't becauseI've done it.
I'm telling you, it's amazing.
That means we stop building AIslike skyscrapers and start
(17:57):
building them likeneighborhoods.
SPEAKER_00 (18:01):
Well, let's talk about what nobody wants to talk
about, and that's theenvironmental cost.
SPEAKER_01 (18:07):
Well, I mean, if we continue this trajectory,
exponential GPU production, rareearth mining expansion, water
depletion and drought zones,grid instability already being
proven.
Let's not even get in the factthat everyone has a data center
nearby, is talking about thefact that their water bill and
their utility bill go up, up,up.
(18:29):
So grid instability, landconversion, which really bothers
me having grown up on a farm andunderstanding food supply.
We could literally see AIinfrastructure rival aviation
emissions long term.
And I'm saying long-term here,but the truth is that's
conservative.
I'm talking about it's reallyshort term, less than 30 years.
I'm just saying from the mathwe're seeing and and the the
(18:52):
sheer push for big, big, bigcentralized, that's what we're
seeing.
That's amazing.
It's unnecessary becausedistributed SLM architecture
changes the equation.
Lower power density, smallerclusters, passive cooling, idle
hardware use.
(19:13):
There are some amazing Web3 guysout there already sharing their
processing power.
I've been a part of it.
Sharing processing power touniversities, GPUs and CPUs on
unused hardware that they justhave plugged in that's asleep
for the most part, running inthe background so that the
universities can use their LLMsand algorithms and more to come
(19:35):
up with things, for instance,like solving AIDS and cancer
research.
So using idle hardware across alot of folks that are
participants, shifting workloadsto renewable, heavy, heavy
regions instead of megawattcampuses.
We get we're gonna get insteadof those big massive megawatt
campuses, we're gonna getmicrowatt participation.
SPEAKER_00 (20:02):
So this is where people get skeptical.
SPEAKER_01 (20:11):
Yeah, now I'm gonna scare everybody off real quick.
Um blockchain web three crypto.
There it is!
SPEAKER_00 (20:17):
Run for your life!
SPEAKER_01 (20:19):
No.
Um actually, our belief ofFierce Labs and Roar is that um
blockchain crypto web 3 has beenlong um something that people
have pumped and dumped orwhatever.
We built in here forever.
We've invested, we've been inother projects, and uh, and
let's just be really honest.
A lot of it has been solutionslooking for problems.
I've said that for years,plural.
(20:41):
However, if we stop, just justjust put all that on the shelf
and let's not call it blockchaincrypto, whatever, and let's call
it web3.
And just from our last podcast,we talked about the fact that
web two could really benefitfrom web 3's Rails.
So if we built this in a railworld, um, in whether it's
(21:02):
blockchain or whatever you wantto call it, hybrid lime wire,
napster, whatever, P2P,something similar, but that pays
people back for beingparticipants, which is the Web3
idea.
Decentralized storage,decentralized compute with a
benefit to the participants forusing micro amounts of their
(21:25):
machines, their storage, theirGPU, their CPU.
If you see it like that insteadof blockchain, crypto, whatever,
if you see it as decentralizedcompute and storage, which is
what we're talking about, with areturn for the participants, you
can see that this could actuallycoordinate it.
There's a difference.
If millions of people arecontributing storage and compute
(21:48):
through a Web3 model that AI isbuilt into, well, you already
need verifiable accounting,performance tracking, and
trustless reward systems.
And that's what What Web3 isgood at.
You get it?
So that is really some of thecore logic of Web3.
SPEAKER_00 (22:08):
So this goes beyond speculation.
It's really about coordination.
Yeah, exactly.
SPEAKER_01 (22:14):
It's it's where systems and ecosystems like Roar
come into play.
And we're we're actually goingto prove it because we're we're
already building it.
I hate putting stuff out thereand giving people an edge on us,
but I just felt like theconversation's gone so fast and
so far that if we didn't wedidn't stake uh a marker in the
dirt and put up a flag somewhereand call it out before we kept
(22:37):
going, that either legislatureor some global elite body or
oligarchs or something are goingto control the narrative and
we're gonna be on a collisioncourse with an eventuality that
we believe is absolutelychangeable right here, right
now.
SPEAKER_00 (22:54):
All right, Dustin.
So how does Roar fit into this?
Well, Roar isn't about chasingtrillion parameter vanity
metrics.
SPEAKER_01 (23:02):
Yeah, I called that out.
So yeah, I'm being ugly on this.
I I can be.
I don't have a pony in theparade that you can see, so
who's gonna pick on me, right?
I don't matter.
But what we have under hooddoes.
We're building distributedinfrastructure, edge native
intelligence, decentralizedcoordination layers, tokenized
incentive systems.
(23:23):
And we've been here for fiveyears doing it.
We kind of got a good idea whatwe're doing, let alone all the
years prior from my early daysin Bitcoin going back to 2010, I
think it was, and then beyondthat, and social media and
platforms that go all the wayback to Zanga.
And yes, that's at the turn ofthe last century, my friends.
I've been here a minute.
(23:43):
I've been in web one, web two,and now web three.
And so we have a touch on allthis.
It's a little bit different fromothers.
I don't know how many others,when they went to college, got
to mess with the ARPANET, butI'm just saying we have a
different view on things becausewe've been here a minute.
So we get some of thisinfrastructure, things that can
change, coordination layers,tokenized incentive systems,
(24:04):
distributed infrastructure, edgenative intelligence.
Our vision of decentralizedstorage, compute participation,
and SLM orchestration is nottheoretical.
I got like four white papers tochase this in a minute after
this one.
We've been building towards thislogic for a long time.
We've been moving here the wholetime.
It's been in our roadmap for our10-year plan, and we just ain't
(24:27):
told nobody about it until now.
So that's the direction theindustry has to move to if it
wants to survive long term.
I'm just saying, if we don'twant to become the Matrix or
Skynet or others, then weprobably want to move this way.
And I still don't believe inthat kind of agency and
creativity at that level.
I'm just making fun of everyoneout there because we may not be
in a place where AI takes overand puts humans as servants, but
(24:50):
we definitely are going to be ina place where we've used up all
of our ecological resources, ourbeautiful beaches, waterways,
our animals, our plants, ourcommunities, our freshwater, our
open land, our agrarian society,and the fact that America
represents 25% of the productionof all the food on the planet,
(25:10):
and we're buying up farms toturn into data centers, which
pisses me off.
SPEAKER_00 (25:17):
That's what we have to do to survive long term.
I would say, how do you reallyfeel about it?
But I think we know.
Um, so Roar's not just buildinganother hyperscale facility,
right?
We're building a mesh.
Exactly.
SPEAKER_01 (25:32):
So horizontal intelligence layer, not vertical
concentration.
I've been against centralizationsince the beginning.
It's in our white paper we justlaunched even a few weeks ago,
but it predates that becauseit's core for Web3 values from
the beginning, from the earlydays of Bitcoin's white paper,
(25:52):
that has been the bad guy,centralization.
And so we're just not about thatvertical concentration.
We are about a horizontalintelligence layer.
I think everyone should playnice with each other in smaller
chunks that are well positioned.
We've already got leaders inspecific kinds of AI.
Why don't you stop trying tofight with each other and
(26:14):
collaborate and then allow otherparticipation through meshes?
Why not have your idiot savantsall over the place that only
know one thing and then offloadprocessing between like a mesh?
We already think about mesh withIoT connecting services like
Amazon Sidewalk.
(26:34):
Nobody wants to talk about thefact that that works.
But why are we not doing in thesame kind of thinking this with
the processing for AI or more?
Um, we got to end the verticalconcentration, we got to end the
centralization, we've got to endthe obliteration of our natural
resources, our communities, ourutilities, our grids, and get
(26:55):
back to the idea that therecould be a democratization of
all of it and recognize as well,even in that, Web3 has a better
solution for making sure thingsare not biased, they're they are
censorship resistant, and theyhave the right rails and
firewalls.
We can build those into ablockchain, and they are
ledgerized.
(27:16):
And I've taught everyone, I'mthe first one I know of that's
done it, that you can take an AImodel and train it to
communicate with an individualholding an NFT on a wallet as if
it's that personality and dothings specific to it.
I have done that and made it soit can learn back with itself
with some of the biggest modelsout there.
And you guys, if you're at oneof those biggest models, you've
(27:38):
probably seen some of my work.
I'm just saying that my friends,we've been doing this whether
you knew who we were or not.
We've been on the front end ofit, and I'm telling you, it
works.
We're not talking about theory,we're talking about function.
SPEAKER_00 (27:53):
All right, Yoda, paint the big picture, show us
the way of the Jedi.
Jedi ways is what we're nothere.
SPEAKER_01 (28:01):
Okay, let's paint the big picture.
Imagine this millions ofparticipants worldwide.
Some provide storage, someprovide inference, some host
specialized SLMs, some verifyresults.
All of them are willingparticipants.
(28:22):
That's key.
Blockchain tracks performanceand rewards fairly.
We can call it something else ifyou don't like the name.
A ledger that's time-stampingactivity that keeps everybody
honest.
It's kind of a proof of what isverifiably true, right?
Compute routes dynamically torenewable heavy grids, idle GPUs
(28:47):
become more productive, andolder hardware gets extended
life.
SPEAKER_00 (28:57):
I'm just processing what you just said.
That's um it's pretty brilliant.
Uh what that means is instead ofAmazon building a hundred new
mega facilities, 10 millionpeople participate in AI
infrastructure.
You know, that's the shift.
SPEAKER_01 (29:15):
It's from centralized dominance to
distributed intelligence.
Now, the problem here is, nooffense to you AI developers out
there, um I'd say that there'sum there's probably um well,
I'll just say it straight up.
I think there's a real lack ofhumility.
(29:36):
So I don't see a lot of themsaying to their boards and their
teams, hey, how about we share?
You know, in kindergarten welearn that sharing is caring,
and we learn to keep our handsto ourselves.
And yet these guys are grabbingeach other's technologies and
lying about it, and thenfighting over who said what and
did what first, and then runningto Congress and going, stop
(29:57):
them.
They're trying to do somethingbad.
And you guys, you sound like youneed to go back to kindergarten.
And then if you really want toget in there, you're there's a
lot of you that say you'renonprofity, but you're really
profity-y because you gotshareholders of VCs and
everything else, and yourhumility is at an all-time low,
and you think that you're theonly one that can get us there.
Good for you.
(30:17):
My belief is that there are amillion geniuses out there, and
that we need to evendecentralize the land claim to
genius status and AI and alloweveryone to be a part of
distributed intelligence.
It's safer, it's better, it'smore humble, and in the long
run, I want to just put outthere it can use older
(30:39):
technology and more.
Think about what this could dofor economies.
So you're over in Africasomewhere, or let's say
Greenland, it's cold there.
That's good for compute, right?
So you don't even need to havecoolers, you've got an Iceland
mass.
And you're over there, and we'rehearing about how they've been
challenged with finance becauseDenmark ain't doing right and
(31:00):
the US is involved.
I got nothing to say about that,except wouldn't it be great if
they could just run some oldcomputer hardware in a super
cold environment and getresources because they're paid
to share their GPU and CPU?
People in mud huts with you knowa solar panel and cellular
signal can actually be aparticipant in the stuff we're
(31:22):
building and be a part ofcompute, sharing the load across
everywhere, but alsodemocratizing the control and
the status, as well as financingeconomies that would have never
gotten a touch any part of biginfrastructure or money in this
space.
(31:42):
Guys, this is a different kindof thinking.
SPEAKER_00 (31:46):
Yeah, this makes me think of that quote, a rising
tide raises all ships, right?
When everybody participates,it's beneficial to everyone.
It's amazing.
So that's let's let's take alook.
We're at that dramatic fork inthe road, then aren't we?
SPEAKER_01 (32:04):
Yeah.
Um, I didn't want to do thispodcast.
I didn't want to do this call.
I'm way ahead of where I want tobe with the other white papers,
so I took them, ripped themapart, made another white paper
on this, and then and none ofthem are out, and then we we've
turned it into this conversationpodcast, and I really didn't
want to do it.
(32:25):
But I'm running out of time.
Um, I've been in this longenough to realize that there's
some deadlines on stewardship.
Someone has to get out here andsay this.
Somebody has to have theconversation or at least start
the narrative or at least openup for a forum of discussion.
And if that means somebody beatsus to the punch and they build
better than us or whatever,great, then I've just saved my
(32:46):
kids' problems.
I would love to be thebeneficiary and my community be
the beneficiary of all of whatwe're talking about here as well
as all the build we're doing.
But if somebody beats me thereand it saves my backyard so that
I'm not overtaken by a datacenter, God bless them, go for
it, you know?
So that's why we're doing this.
And it's because we see thesetwo futures.
(33:07):
There's a fork in the road, andthere's future A, and it's
megadata centers dominate, AImonopolies control access,
governments can turn things onand off of censorship,
environmental pressuresescalate, energy grids will be
strained, and then corporateconcentration will intensify.
(33:28):
With that comes all kinds ofthings.
Like I said, censorship, bias,control, and worse.
Then there's future B.
Specialized SLM mesh networks,which in a sense can be
firewalled, controlled, and madewith breakers they can get
(33:50):
turned off.
Community nodes, blockchaincoordination of some sort,
storage, decentralized storagecoordination, whatever.
Lower environmental footprint,distributed revenue, and global
participation.
SPEAKER_00 (34:09):
Wow.
So one future centralizes power,the other decentralizes
intelligence.
Exactly.
SPEAKER_01 (34:17):
And over 50 years, what's what's the result?
The environmental differencecould mean gigatons of carbon
dioxide.
For folks that care about that,that's massive.
The economic difference couldmean trillions redistributed
instead of consolidated.
So trillions redistributed.
Oh, I can't talk.
(34:39):
Redistributed, there comes mysouthern, instead of
consolidated.
And so, I mean, that economicimpact for anyone participating
in countries that are open toit, it'd be massive.
Folks that have minimalhardware, a cell phone or more,
a Raspberry Pi or more, canparticipate at some level in the
(35:02):
layers of these systems.
Um, that's what we're building,and that's where we're going.
SPEAKER_00 (35:09):
Uh that's amazing, Dustin.
Um, so the current AI industryis building bigger buildings.
And you're saying the future isactually to build a bigger
network.
Yeah.
We don't need bigger models andbigger warehouses.
SPEAKER_01 (35:25):
We need smaller intelligence everywhere.
SPEAKER_00 (35:28):
So AI doesn't scale by becoming gigantic, it scales
by becoming distributed.
The more and more distributed,the better.
And that's it.
SPEAKER_01 (35:37):
That really is the core.
We build larger impact bybuilding smaller systems
together.
We always say we're bettertogether.
Both sides of the aisle inCongress will say we're better
together, but how do they act?
Here's the truth.
In this, we are better together.
Somebody's got to say it outloud.
Somebody needs to encourageeveryone to at least, at least
(35:58):
take a look at it.
You know, Elon, at least thinkabout the fact that you could
offload, and I've said this overand over for the last year, you
could offload process power fromyour centralized data center for
your XAI.
You could offload it to thecyber trucks that aren't being
used.
You got a whole fleet out theresomewhere, or even run in the
background.
If you've already got the AI inthem, why isn't it running in
(36:20):
the background utilizing some ofthe compute power back to the
the grid of all of them?
I mean, it could be a part ofthe agreement when you get one.
Why not?
I'm telling you, it's a wholedifferent thinking and it would
have a bigger impact and betterresults.
SPEAKER_00 (36:33):
So it's we build bigger by building smaller.
Yeah.
Um, so that's not anti-progress,it's just a better way of
thinking about it.
It's smarter progress.
SPEAKER_01 (36:46):
And if we don't shift, I told you about that
being a degree off on a shipwhen you start that journey
across the ocean, can throw youoff on a whole different place
on the planet than you aremeaning to go.
If we don't shift now, AIbecomes another centralized
industrial complex.
I'm calling it now.
(37:07):
And let's be honest, others havecalled it before me.
Don't read Isaac Asimov, youdon't be scared out of your
wits.
I've read almost everything theman's written on purpose.
AI becomes that centralizedindustrial complex, and if we do
do this, if we do shift, it'sgonna become a global mesh of
intelligence instead.
(37:27):
Horizontal, not vertical, anddecentralized, not concentrated.
SPEAKER_00 (37:33):
So we need to build smaller.
SPEAKER_01 (37:35):
Yep, there it is, to build bigger.
That's it.
That's it.
And brother, thanks for being ontoday.
Folks, thanks for listening.
Um we did not want to be thebearer of this news, but if no
one else is gonna say it, we'regonna roar about it.
Thanks for being on your weeklyroar.
Know that this does mean thereis a white paper and coming.
(37:58):
And so um over the next weekyou'll see it publicized or
whatever through our channels.
And as well, we're also givingyou a sneak peek into what's
being developed and alreadyeither at market that we're in
alpha testing on with ourinitial alpha partners, or we're
leaning into bringing into betafor you to use.
(38:21):
Um, it's time we're here, andyou're gonna hear more and more
about it.
So thanks for being on.
Be sure to follow us in all theplaces you find us, whether it's
X, YouTube, Rumble, or others.
And then stay connected, guys.
There's so much news coming out.
March is based, April's gotanother one.
April's gonna be hot.
May's got a we've got slogansfor each one.
(38:42):
You're gonna want to be a partof it all the way through the
rest of this year.
It's gonna get funner.
Brayden, thanks for being on.
Everybody, thanks for joiningwith us.
Y'all have a great one.
Thanks, everybody.
Take care.
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.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Kingdom of Fraud
It’s the unlikeliest of criminal partnerships: a devout polygamist from an insular Utah sect joining forces with a shadowy Armenian tycoon from LA. The result - a billion dollar fraud conspiracy. In Kingdom of Fraud, investigative reporter Michele McPhee traces the origins of the extraordinary alliance between Jacob Kingston and Levon Termendzhyan. Together, the two men trigger the largest tax investigation in American history and weave around themselves a web of dirty cops, influential political relationships and transnational money laundering. All this is set against the backdrop of Jacob Kingston’s clan – The Order. A powerful and secretive polygamist organization in Salt Lake City. To whom Jacob is desperate to prove his worth. Kingdom of Fraud is produced by Novel for iHeart Podcasts. For more from Novel, visit https://novel.audio/. You can listen to new episodes of Kingdom of Fraud completely ad-free and 1 week early with an iHeart True Crime+ subscription, available exclusively on Apple Podcasts. Open your Apple Podcasts app, search for “iHeart True Crime+, and subscribe today!