Episode 22

Decentralized Computing Revolution: Alessandro De Carli on Acura's Real-World Applications (Part 2)

In this episode, Dr. Jemma Green continues her conversation with Alessandro De Carli, exploring how Aura’s smartphone-based computing network is disrupting traditional data centers. Discover the sustainability benefits of repurposing old phones, practical applications in web scraping and AI, and the competitive edge of decentralized computing.

Highlights:

  • Overview of Aura’s real-world applications.
  • Hear firsthand experiences of onboarding and running Aura processors.
  • Learn how upcycling smartphones contributes to environmental sustainability.
  • Strategies for growing the Aura network and educating users.
  • Real-world examples: from web3 security to large-scale web scraping and AI.
  • How Aura’s decentralized approach challenges big cloud providers.
  • Addressing risks, bandwidth, and safety for users and providers.
  • Final thoughts, takeaways, and a look at the future of decentralized computing.

Key Takeaways:

  • Aura leverages old smartphones for powerful, energy-efficient computing.
  • The network supports a range of applications, including web3, AI, and data scraping.
  • Decentralized computing offers privacy, cost, and sustainability advantages over traditional cloud services.
  • Community growth and education are crucial for adoption.

Connect & Learn More:

  • Learn more about Aura: Aura Website
  • Guest: Alessandro De Carli

UnBlock'd podcast with Dr. Jemma Green

For more information on Dr. Jemma Green

Visit: https://www.powerledger.io/

Or connect on LinkedIn: https://www.linkedin.com/in/jemmagreen/


-------------------------------------------------------------

Produced by: Podcasts Done For You

View this episode on YouTube @PodcastsDoneForYou_clients

Transcript
Anthony Perl:

Decentralized computing revolution.

2

:

Alessandro Dali on Aura's Real

World Applications part two.

3

:

In this episode, Dr.

4

:

Gemma Green continues her conversation

with Alessandro, exploring how aura's

5

:

smartphone based computing network is

disrupting traditional data centers

6

:

from the sustainability benefits of

repurposing old phones to the practical

7

:

applications in web scraping and ai.

8

:

Alessandro reveals how his technology

is creating new possibilities

9

:

for decentralized computing.

10

:

They also discuss the competitive

advantages of mobile devices, energy

11

:

efficiency, and how they're addressing

security concerns while scaling to

12

:

challenge traditional cloud providers.

13

:

I'm your co-host Anthony Po.

14

:

And whether you're an investor or

a startup looking for insights,

15

:

it's time to get unblocked.

16

:

Jakob Stammler: I would just like to

add kind of an anecdote, not a question.

17

:

I've been successfully running Acura

processors for over a year now, and I

18

:

have to fully agree with you, Alessandro.

19

:

I think it's the smoothest onboarding

I've ever had, not just in Web3.

20

:

Also in web two to ever run

anything similar to this.

21

:

It's literally as simple as

scanning a QR code and basically

22

:

the whole setup is done.

23

:

And also what you said, kind of your

last point is every time I upgrade

24

:

my phone, I have a new processor.

25

:

So that means that even just on at home,

I started with one, maybe two, and when

26

:

I upgrade my phone, I add a new one.

27

:

And I remember very early on in your

story, I think it was only possible to

28

:

run a processor on a Google Pixel phone.

29

:

Initially, and I remember I was

looking for pixels in the European

30

:

market and it was almost impossible

to find anything in Switzerland.

31

:

Everything was sold.

32

:

Any type of pixel was completely sold out

because everything was bought by people

33

:

wanting to run accurate processors.

34

:

And when you were looking for pixels

within the European market, it was

35

:

really difficult finding anything.

36

:

And it was really interesting

to see because people understood

37

:

the value of running it.

38

:

And how cheap it is to actually

run this processor, even with just

39

:

old phones that they had at home.

40

:

That just made it so simple and almost

like no brainer to just onboard it.

41

:

And I think that that is definitely

part of the success that you've seen.

42

:

Alessandro De Carli: I mean, thank

you so much for the kind words.

43

:

Yeah, definitely.

44

:

We're, we're working hard to make

it as easy as possible, and I mean.

45

:

An aspect that we're not even advertising

throughout the entire website is the

46

:

aspect of the impact on the planet

in terms of sustainability, right?

47

:

Because people are effectively

buying out phones that otherwise.

48

:

It's quite crazy.

49

:

You have to think about it.

50

:

You have big tech investing billions

in r and d, printing out these amazing

51

:

main boards and compute, and then

people just throw them away after one

52

:

or two years or put them in a drawer.

53

:

Now all of a sudden, all that engineering,

all that tech waste, you can put it

54

:

back into a productive system where

you're able to squeeze out some very

55

:

valuable compute that you can offer to

56

:

Anthony Perl: somebody else.

57

:

Right.

58

:

Just to that point again, do you

have to be on a network as such?

59

:

Because you know, we're all used

to pay jumping on a network for

60

:

our phones, and if you're using

an old phone, you don't need that.

61

:

So you will need

62

:

Alessandro De Carli: internet

connectivity, but you won't

63

:

need a separate SIM card

that connects to cellular.

64

:

So you can have, if you want, you can

connect your processor over that line.

65

:

But what we usually see is that

people utilize either wifi or

66

:

go over ethernet leveraging the

USBC connection off their phone.

67

:

Anthony Perl: It's a great way, as

you say, for sustainability as well.

68

:

And I imagine that's a very big argument

for moving forward and using this because

69

:

what do you do with all the old phones?

70

:

Where do they go?

71

:

Alessandro De Carli: Yeah, and we followed

them and uh, where they all end up is

72

:

basically in Hong Kong at a auction where

you can purchase Samsung S 23 by the kilo.

73

:

Basically, you can purchase

a hundred kilo of iPhones, of

74

:

Samsung devices and whatnot.

75

:

What then usually happens

is that some people.

76

:

Utilize them for the gold and

scrap value that it still has.

77

:

But there are some other people that

started really upcycling them to then

78

:

create also these kind of racks where

you have 20 devices in a single rack,

79

:

just the main boards thereof, and

then you can connect them to a network

80

:

like AIST to provide that compute.

81

:

Anthony Perl: And what's the plan in

terms of telling everybody about this?

82

:

Because obviously it's a, something that.

83

:

Anyone can take up the opportunity with.

84

:

So how do you get the word out

there and make it simple for people?

85

:

Alessandro De Carli: So getting the word

out there is extremely important, and I've

86

:

been very, very grateful and thankful for

the marketing team that we have that is

87

:

really challenged with a very technical

topic, and being then able to make this

88

:

consumable for everyone is not so easy.

89

:

One of the main challenges here

is translating the tech language

90

:

to a language that everybody is,

is capable of understanding it.

91

:

And also the other big barrier is that.

92

:

Many people are belittling the mobile

device thinking that it does not like

93

:

it's unthinkable for them that this is

a device that can be used for compute.

94

:

Right?

95

:

So educating people and having a messaging

that is very easy to consume is one

96

:

way how we make this very visible to

everyone but the network because it has.

97

:

Two main stakeholders.

98

:

So the consumers and the compute providers

needs to grow in a symmetric way.

99

:

So if we just would have a lot of people

onboarding, compute on the network,

100

:

that is great, but you still require

then someone actually utilizing that

101

:

compute for this to be sustainable.

102

:

So my short answer is that.

103

:

Maybe getting the word out to everyone

just about how they can onboard.

104

:

Compute is not the immediate

best next step because we need to

105

:

find the balance between the two.

106

:

So right now we have the issue

that actually most of the

107

:

computers already being used up.

108

:

So now it's a good moment again to

push for people to onboard compute.

109

:

And if there's listeners in here

that are interested, please go to the

110

:

app stores and search for accuracy.

111

:

Then you will be able to

onboard very, very quickly.

112

:

But there was a time where we had a

oversupply and then just onboarding

113

:

more devices wouldn't necessarily

translate to more value for the network.

114

:

Right.

115

:

Anthony Perl: Can I just ask, can you give

us some good examples of who is using it?

116

:

And what they're using it for.

117

:

Alessandro De Carli: Yeah, absolutely.

118

:

We have various verticals

being a Web3 company.

119

:

We do have a couple of larger foundations

like Crypto Foundations utilizing and

120

:

leveraging the security aspect of it

because it's basically a perfect fit.

121

:

Thanks to the trusted execution

environment that it has for running

122

:

on chain automations, for running

bridges, for running things that

123

:

basically require observation from

one chain or service and then bring

124

:

that to another service or chain.

125

:

So those things is what we call, like

our Web3 on chain automation vertical.

126

:

That's where people utilize that,

especially for the security aspect.

127

:

But we then also have a.

128

:

Very large usage from people that utilize

this for web scraping and data kind

129

:

of observations of, of the internet.

130

:

So those people utilize it

mostly because of the bandwidth

131

:

aspect of the service provided.

132

:

These are companies that offer

a service to Google, OpenAI, all

133

:

these large companies that require.

134

:

A view of the internet heavily rely

on web scrappers and residential

135

:

kind of access to the internet.

136

:

Because, I mean, Google is

basically a web scraper, right?

137

:

That is what they do.

138

:

They, they scrape data and then

offer you the data for indexing and

139

:

for kind of search purposes, right?

140

:

If you don't utilize a

network like to do so.

141

:

You will quite quickly get

blocked because not everyone

142

:

wants you to collect that data.

143

:

So that is another web two use case where

we have about 20,000 devices that are just

144

:

web scraping every single day for data.

145

:

And now one thing that I'm super excited

about is the ability to basically

146

:

leverage the compute more, and that

is by leveraging this unified memory

147

:

architecture to run these AI models like

these LLMs, nowadays you have to run

148

:

them with, uh, centralized clouds, right?

149

:

That means that basically the centralized

cloud has full access over your data.

150

:

Which is also a little bit, the reason

why we don't see more adoption of this

151

:

technology, to be honest with enterprises,

because they don't necessarily want

152

:

their data to be shared with pic tech.

153

:

And with accuracy, you're actually

able to run these models privately

154

:

and on your own, and then leverage

also the compute aspect of it.

155

:

So basically it's a play where we

receive a resource that is incredibly

156

:

powerful, incredibly useful.

157

:

With various aspects and then leverage

those virus aspects to offer services

158

:

for people that then utilize those.

159

:

Jemma Green: I had a question

on the disruptive power of this

160

:

technology and whether it's caught

the attention of the big players.

161

:

Related to that is how much does

it cost to use Acura service for

162

:

compute purposes compared to.

163

:

Say traditional, and particularly for

ai, you know, we obviously know there's

164

:

a huge growth in demand for compute.

165

:

And electricity as well for servicing

AI models and in the energy space,

166

:

which power ledger works in like so

many very fascinating things happening.

167

:

For example, in the US real estate

is being brought up or lease that was

168

:

previously being used for industrial

purposes where they've got a big

169

:

electricity connection and an offtake

agreement to instead repurpose that.

170

:

For data centers to

meet the growing demand.

171

:

So there's no shortage of demand growth,

but obviously the players would like the

172

:

demand to accrue to their infrastructure

and their services as what you're doing.

173

:

Caught the attention of the

bigger cloud service providers.

174

:

And do you think that they will see

your technology and also pricing

175

:

as like real problem for them?

176

:

Alessandro De Carli: So this topic

alone would be a topic that we can

177

:

discuss for an entire afternoon, right?

178

:

The race for having larger and larger

data centers, and what it means for

179

:

things like energy infrastructure

and for humanity in general, right?

180

:

Because one of the bets of

these hyperscalers and like for

181

:

instance, also meta, who is now.

182

:

Planning to do this Hyperion data

center, which has the size of Manhattan

183

:

and basically is also consuming 5.7

184

:

or plans to consume 5.7

185

:

gigawatts, which is basically what

Manhattan is consuming, right?

186

:

So.

187

:

The, the reason for these players

to do that is because they all

188

:

think that will be soon in a world

where if you're able to monopolize

189

:

the access to this functionality

and this technology, you will be

190

:

ultimately able to control the world.

191

:

Right?

192

:

And it is quite scary because

the question is whether you want

193

:

that or not as a human being.

194

:

Now when it comes to the question whether

people have realized that accuracy is

195

:

something that could be used to counter

that, I do believe that yes, to some

196

:

extent we do have quite a lot of growth on

both the deployment and onboarding side.

197

:

So there are people that get the ideology

behind it and profit already today.

198

:

From the, the cheap compute.

199

:

I don't think that larger players

really see this as a threat yet, which

200

:

is a big advantage, I believe, because

it's better to be underestimated

201

:

than overestimated for sure.

202

:

That being said, though, there are

people in the, like in Microsoft

203

:

specifically, who did start to get

some interest in Acura to understand

204

:

what it's doing and whatnot.

205

:

So it's probably not because

they think that it is necessarily

206

:

already a threat today, but they

do see detraction behind it.

207

:

Jemma Green: What about like Oracle,

like, 'cause you know, they haven't

208

:

got their own models, so they're

not really a competitor, you know,

209

:

compared to the other tech companies.

210

:

And they just announced a deal,

I think yesterday, to be the

211

:

supplier for data centers for.

212

:

Are you talking to companies

like that about providing

213

:

your services in additional to

the traditional data centers?

214

:

Alessandro De Carli: So the

technology in the forum that you

215

:

have today is most valuable for its

confidentiality and security aspect.

216

:

So when it comes to an AI agent, as an

example, you want to be able to give that

217

:

AI agent your credit card, your logins

for your emails, and so on and so forth.

218

:

Today, if you do that, you would

effectively be giving all of those

219

:

credentials to a centralized cloud.

220

:

That is a big no-no.

221

:

That is a big problem, right?

222

:

With accuracy, however,

you're then able to say, okay.

223

:

It is my personal, confidential space

that I am renting, but nobody else

224

:

is capable of inspecting or tampering

with for that specific use case.

225

:

It's a perfect match already today

when it comes to running LLMs, there

226

:

are a set of limitations that we're

working on, one of which is the.

227

:

Limiting factor for running L

LMS is effectively the memory

228

:

that you have on your disposal.

229

:

If you want to run one of the

largest models, then you will require

230

:

something like 10 big, big GPUs, right?

231

:

And with Acura, the mobile device,

like the largest RAM allocation that

232

:

you get, is about 24 gigs of ram.

233

:

So.

234

:

Today, we are not able to chart the

model yet onto multiple devices.

235

:

This means that you can only load

models that fit into those 24 gigs of R.

236

:

Jemma Green: The container of one device.

237

:

Alessandro De Carli: Exactly.

238

:

Exactly.

239

:

These are all metrics,

multiplication that run in.

240

:

Need to synchronize between layers

but can be completely sharded

241

:

onto more multiple mobile devices.

242

:

It's just a matter of actually doing it.

243

:

And that's one of the r and d projects

that part of the team is working on today.

244

:

Jemma Green: Got it.

245

:

So that's a real kind of holy

grail to be on a quid crow

246

:

quo basis with the traditional

cloud service providers as well.

247

:

Alessandro De Carli: Yeah.

248

:

So.

249

:

I think in certain use cases, absolutely.

250

:

I think in certain use cases, we are even

winning already today with the offering.

251

:

I mean, being able to run these

data scrapers, you cannot run

252

:

them on centralized clouds.

253

:

Like it is effectively not possible.

254

:

Right.

255

:

So I do think that.

256

:

Being also small team and a project

that does have a lot of traction, but I

257

:

mean, a hundred thousand devices is, is

super exciting, especially in the short

258

:

amount of time that we onboarded them.

259

:

But it's just a little

drop in the ocean, right?

260

:

We, we need to grow this way, way

more and we need to do so in a

261

:

smart way by basically, um, first.

262

:

Catering and onboarding the kind of use

cases that work really, really well with

263

:

the tech, and then expand on the use

cases that are a perfect fit for the tech.

264

:

That is our strategy today, because if

we would just go after everything all at

265

:

once, you are master of nothing, right?

266

:

Jemma Green: Definitely.

267

:

Yeah.

268

:

It makes sense that you are, yeah.

269

:

Tackling the lower hanging fruit first.

270

:

How does the cost compare?

271

:

For using compute, like could you just

give a metric that might be more readily

272

:

understandable to put into contact

how much cheaper Acura actually is?

273

:

Alessandro De Carli: This is something

that we haven't even touched on before

274

:

when kind of highlighting the mobile

device, but these are devices that

275

:

never had the privilege of being

connected to a power supply 24 7.

276

:

So they need to be very, very efficient,

have a very high compute density

277

:

at a very low, but consumption.

278

:

If we take a look at what the cost

structure and main challenge for these

279

:

data centers is, is okay, you need

to be able to source the hardware.

280

:

That is a limiting factor with actors.

281

:

That's already cool because we're

able to crowdsource that and.

282

:

Upcycle devices that are already in the

market out there, we're not blocked by

283

:

anyone still needing to produce that.

284

:

But the second factor is definitely

the access to cheap energy because

285

:

these data centers consume so much.

286

:

Now, mobile devices have a way,

way lower cost consumption on the

287

:

energy side compared to a server.

288

:

In comparison of the

compute that they provide.

289

:

And this is also the reason why I believe

you can only run a decentralized compute

290

:

network that is capable of competing with

centralized offering on mobile device.

291

:

Like no other hardware will be suitable

for this because you will ultimately

292

:

suffer from economies of scale where

the others are just better off and if

293

:

you just use the same hardware there.

294

:

Gotcha.

295

:

Anthony Perl: I was just gonna ask as

well, I mean, just for layman's terms,

296

:

someone at a local level, they've got

their phone, they're installing this.

297

:

Are there any risks?

298

:

Are there any things that someone should

be aware of when they're doing this?

299

:

How do you educate at that level?

300

:

I mean, you obviously,

301

:

Alessandro De Carli: first

of all have a device, right?

302

:

So if you're able to upcycle that

device and buy it for relatively

303

:

cheap, you don't risk the situation

where you bought this device for,

304

:

let's say, accurate, and then you have

a a big cost expense on that side.

305

:

So that is, in my eyes, a potential risk,

but not necessarily a very large risk.

306

:

Other thing to definitely discuss as well

is that people will use your bandwidth

307

:

to scrape data and so on and so forth.

308

:

So today, um, accuracy is

blocking all sorts of malicious.

309

:

Website access and so on and so forth.

310

:

Tomorrow, we will also have a white

list approach where basically you're

311

:

able to say, it's okay to utilize my

bandwidth, but only if you go on these

312

:

websites that I know are not malicious.

313

:

Right?

314

:

But that is something that bandwidth

will be ultimately used, which is not.

315

:

Necessarily a big risk because if

you think about it already today, it

316

:

is possible that either your phone

or your router or something has been

317

:

infected with some malicious kind of

code, and then someone is utilizing

318

:

that bandwidth without you knowing it.

319

:

Right?

320

:

Dear, you at least know that someone

is deploying stuff and it's all

321

:

completely transparent, right?

322

:

You're, you're able to see who.

323

:

Deployed was, and it was really important

for us to have this still accessible for

324

:

everyone so that it is completely safe.

325

:

The last thing that I would say is

if you're planning to run hundreds of

326

:

devices, you will also need to start

thinking about the power consumption

327

:

and also the electricity lines, right?

328

:

Um, if you live in an old house.

329

:

And there's some houses here in

Switzerland that still insulated with

330

:

newspaper and I dunno, I mean, if you

start really getting a lot of energy

331

:

in those cables do get hopped, and then

maybe there is also a hazard there, but

332

:

with mobile devices, way less than if you

would run a server or something like that.

333

:

But those are the risks that I

can personally really think about.

334

:

Other than that, I

don't really see a risk.

335

:

Jemma Green: Alessandro, what is

your favorite song at the moment?

336

:

Alessandro De Carli: My favorite song,

so like, I'm from the Italian speaking

337

:

part of Switzerland, so Italian songs

are, have a special place in my heart,

338

:

and at the moment I'm listening a

lot songs from this band called Una,

339

:

which is not so known, let's say,

for the English speaking people,

340

:

but that's my favorite song for now.

341

:

Jemma Green: I like

listening to Italian music.

342

:

I like do Nazi a lot at the moment.

343

:

Nice.

344

:

So thanks for the tip.

345

:

I'll definitely be looking it up and

I'm sure there'll be some other Italian

346

:

song listening listeners to the podcast.

347

:

Alessandro De Carli: Yeah, let's

hope there are entire few, I hope.

348

:

Jemma Green: I really

appreciate you coming on the

349

:

podcast in this conversation.

350

:

The big takeaways for me is that.

351

:

What is could otherwise end up in

landfill or just be a lazy asset that's

352

:

not utilized, can be really have a true

circular economy repurposing effort.

353

:

And it's not just for the device

itself, but it's displacing

354

:

potential new infrastructure that's

been built not in the cloud space.

355

:

And we know.

356

:

How demanding that can be in terms of

energy consumption, but water as well

357

:

and the impact on local communities.

358

:

You know, it ticks a lot of boxes in

terms of it being decentralized and

359

:

a DAO and having new use cases for

cloud that aren't able to be used using

360

:

traditional, uh, compute services.

361

:

There's so many things about

it, which are really exciting.

362

:

Alessandro, and I just wanted to

acknowledge your real pioneering

363

:

work that you're doing in this and

the, the successes that you have.

364

:

Accrued so far and getting more

than a hundred thousand devices

365

:

on in such a short period of time.

366

:

So we look forward to having a

follow up conversation with you

367

:

and to see how things are tracking

in the not too distant future.

368

:

Alessandro De Carli: Thanks, and Pastor.

369

:

Yeah, thank you so much for having me on.

370

:

It's always a pleasure to be

talking with you, Jane Ma,

371

:

specifically, and now with Jacob.

372

:

Jemma Green: Awesome.

373

:

Anthony Perl: That's all for

this episode of Unblocked.

374

:

Please check out the show notes

for information on Power Ledger

375

:

and other contact information.

376

:

We welcome your comments and

feedback and please hit subscribe

377

:

wherever you are listening.

378

:

This podcast was produced

by podcast Done for You.

379

:

We look forward to your

company next time on Unblocked.

All Episodes Previous Episode

About the Podcast

Show artwork for Unblock'd
Unblock'd
Making sense of tech and money with Dr Jemma Green

Listen for free