SPC#79 – And Don’t Call Me Shirley

Hello, SPC followers.

2

Welcome back to another episode.

3

I've got a little explaining

to do for this one, but I

4

think you're going to like it.

5

So bear with me a little bit.

6

Once I bring you up to speed, we'll get

onto the main attraction for episode

7

79

8

And that's a very fascinating

discussion with Alex from flyshirley.

9

com, an AI startup.

10

That's all about aviation.

11

So let's get into it.

12

All right, so a bit of an explanation

before we dive into it I'll be brief

13

because this is gonna be a long episode

anyway But I decided when I published

14

episode 78 the last episode in late

September that I needed a bit of a

15

break So I decided to take my birthday

month, October, off from the podcast

16

while I recharged a little bit We had

some changes happening with the almost

17

launched fledgling flightline podcast

that I'll get into in a little bit, but

18

I just needed to take a little break.

19

So I got some episodes partially

ready and held off on publishing

20

supposedly until November.

21

As you know, At that same time,

I've been trying to get my CFI

22

checkride scheduled for ages, and

I wasn't sure when it would happen.

23

But then, last couple of days of

October, I finally got the call.

24

I got a checkride scheduled at an

airport about 60 nauticals miles

25

away from my home airport for a

couple of weeks in the future.

26

So now About the time I was going to

be starting the pet podcast publishing

27

again, my focus turned to relearning how

to fly and how to pass a CFI check ride.

28

As you can imagine, all my

focus turned towards that.

29

And once again, I had to forgo

publishing for a little bit.

30

I had a lot of cool stuff

already in the can, but.

31

Now didn't have time to finish the edits

and do final publishing, et cetera.

32

So I decided to wait another month.

33

The good news is I passed my check.

34

Right?

35

So I'm a brand new CFI.

36

The bad news is I ended up taking two

months off from the publishing instead

37

of the one that I had originally planned.

38

That said, I'm now back and I've

been wanting to get this one out for.

39

Quite a while.

40

It's really interesting.

41

Now, there's one more thing.

42

Tiffany and I originally recorded

this for, like I said, the

43

Fledgling Flightline podcast.

44

So that's what you'll hear

when we're talking to Alex.

45

A couple of months ago, though, Tiffany

and I decided to pause the launch of

46

that podcast, at least for a while,

as her situation was changing and we

47

couldn't, you know, Put the focus on

it that we had hoped we'd be able to.

48

We want to do it at some point somehow,

but the good news is I can focus on

49

the student pilot cast even more.

50

And since we already have a community

of listeners here, I'm glad that this

51

interview will get more eyeballs and

more years on it anyway, by releasing

52

it as a student pilot cast episode.

53

As you'll see from the discussion, It

does have a bit of a student and learning

54

orientation anyway, so it fits right in.

55

So you're basically up to speed at a very

high level and we're back to publishing

56

weekly for the student pilot cast.

57

I certainly appreciate everyone's patience

as I recharged and then super charged

58

for my check ride and then relaxed

a bit over the Thanksgiving holiday.

59

So I hope everyone also had

a good relaxing holiday.

60

As always, please reach out

with questions, suggestions,

61

ideas, feedback, Everything.

62

You can find me via email

at bill at studentpilotcast.

63

com or use the contact form on the

website or even via X at, at billwill.

64

That's Bravo, India, Lima,

Lima, whiskey, India, Lima.

65

So with that said, I hope you enjoy

episode 79, and don't call me Shirley.

66

right.

67

Welcome back everybody to

the flight line podcast.

68

We've got a really

special episode tonight.

69

we're sitting here with Alex List,

who's part of the team at FlyShirley.

70

And Alex and I met.

71

Uh, sort of in happenstance, uh,

while we were looking at one of the

72

other vendors at Oshkosh that had to

do with large language models and,

73

and what we sometimes refer to as AI.

74

And so we got to talking and

said, Hey, you know what, we

75

should really get together and

talk about this on the podcast.

76

So Alex, welcome to the

flight line podcast.

77

Alex: Thank you very much.

78

It's good to see you and Tiffany.

79

Bill: Excellent.

80

tell us a little bit about.

81

Alex, and then we'll get to FlyShirley.

82

Alex: Sounds great.

83

So my background is actually

in apps and technology.

84

So in high school, I taught

myself how to program iPhones.

85

I had some of the first iPhones

in the app store and ended up

86

winning an Apple design award.

87

I got into MIT and I decided to study

aerospace because I felt I already

88

knew computer science somehow.

89

And so I, uh, so I studied

that for three years.

90

I ended up dropping out

and making a startup.

91

Coming back and switching

to computer science.

92

And then I, then I got my private pilot's

license, back, got my master's in kind

93

of a combination between aerospace

engineering and computer science.

94

And I ended up having the opportunity

to lead flight simulation at a company

95

called Beta Technologies up in Vermont.

96

an electric protocol takeoff and

horizontal flight airplane company.

97

They make some of the, I would say

some of the coolest electric airplanes,

98

but also have some of the best sims

that you'd ever see on the planet.

99

These amazing full dome, uh, half dome,

you know, wrap around flight sims.

100

Really, really sweet.

101

I worked on enabling the simulated flight

test and instructor operating stations.

102

Helping with the avionics transition.

103

I had the opportunity to bring some

of my thesis work into the avionics.

104

When you, when you're flying these new

airplanes is a little bit different.

105

And, and we ended up building about

eight different VR sims and flights in

106

flight domes throughout the country.

107

I left when the airplane is entering

production and worked at a leading

108

aircraft autonomy company and had

the opportunity to work with some

109

really great folks over my career.

110

Ended up starting by Shirley in March

after, um, after some, and we can

111

get into this a little bit more after

there was a little bit of an accident

112

that my friend had, and I wanted to

be able to build something that could

113

a difference for aviation safety.

114

And so.

115

and so we started, I started by building

something that you could fly with in a

116

cockpit, and so you could fly with Shirley

while you're, while I was flying my DA42,

117

I could ask questions and things like

this, but there are some limitations.

118

And so one of my friends actually, uh,

I have a couple of friends at X Plane,

119

and they said, Hey, you know, why

don't you try launching for Sims first?

120

You know, you wouldn't have

the issue with hallucination.

121

This wouldn't be a problem.

122

You wouldn't have issues trying to

keep things online versus offline.

123

And you would be able to sort of deal

with all these noise issues really easily.

124

anyway, we, uh, we launched for flight

sim starting at flight simulator expo.

125

And then, um, and then ended up,

uh, and then ended up here and I'm

126

really happy to see you guys and,

and talk to you a little bit about.

127

this sort of journey and so

128

Bill: uh, some of that tech that you're

talking about, but I'm going to hold off.

129

Because I want to ask, why aviation?

130

What got you into aviation

in the first place?

131

You, you said you started with

aerospace at MIT, and then

132

you became a private pilot.

133

How did that happen, and why?

134

Alex: that's actually, that's actually

a really good, that's a great question.

135

So I always was interested in space as a

kid, a huge sort of, you know, buff for

136

all the sci fi movies runs in the family.

137

You know, I wanted to be like a

lot of folks, I wanted to be an

138

astronaut, but you do the sort of the

expected value calculation on that.

139

And it kind of falls a little bit short.

140

And so I thought to myself, what's

the opportunity that you'd have to.

141

Sort of the best shot that you

could still kind of, you know,

142

fly around and in air or space.

143

And of course, the answer to

that is being a private pilot

144

or becoming a pilot in general.

145

And so, you know, I, I decided to get my

private pilot's license and, um, and it

146

was probably one of the best decisions.

147

You have the ability to, you know,

you have the superpower, right?

148

You could just show up

somewhere, you know,

149

Bill: You have, you have a flying carpet,

150

Alex: do that.

151

You have a

152

Bill: A real, honest to

goodness flying carpet.

153

It's amazing.

154

Alex: It's just the coolest

thing you could possibly do.

155

So there's like, you know, you know,

we're blessed to be in the United States.

156

And I think just.

157

Having this many sort of GA

airports around in this country

158

is just such a phenomenal resource

159

Tiffany: Yeah,

160

Alex: you know, for, of course, all the

municipalities that they're in, but also

161

for the individuals who can end up being

pilots and sort of embracing this love of

162

flight that I had, you know, my teammates

and at Shirley's that Shirley's share and

163

that, you know, of course, all of us here

on the podcast and listening to as well.

164

Tiffany: I was looking at your LinkedIn

and I have to say that between you and

165

the other two co founders, you guys look

like a bunch of slackers and with this 5.

166

0 at MIT from Sam, so it

looks like there's three of

167

you guys who founded this.

168

Bill: You

169

Tiffany: Obviously incredibly intelligent.

170

Bill: Yep.

171

Tiffany: was this?

172

Alex: So, you know, I was

fortunate to meet, uh, Seb.

173

Seb, I actually met, he was an intern

of mine while at Beta Technologies.

174

He was a, sort of, he had just finished

his freshman year at McGill, and he was

175

sort of looking for something to do.

176

He grew up in Vermont,

Beta's up in Vermont.

177

And Kyle, the CEO of Beta, said, Well,

you should check out this guy, you know.

178

You know, you might not, you know, have

a ton of CS experience yet, but, you

179

know, he's going to work really hard.

180

He's trying, you should

like give him a shot.

181

And so, you know, gave Seb a shot and by

three, four years later, I'm a co founder

182

with him at my, at my current startup.

183

so, you know, I, uh, I left my

job at Merlin labs in March and

184

worked on the first prototype of

FlyShirley that you could fly with.

185

ended up, you know.

186

Getting some of my, some of my

friends and family to help me sort

187

of stretch this a little bit farther.

188

You know, I was really lucky to meet Sam.

189

There's a, you know, for people who

like to start companies and are sort

190

of interested in this sort of thing.

191

There are these sort of like Tinder for

co founder type of websites that are

192

as much of a crapshoot as any sort of

dating app you could possibly imagine.

193

In this particular circumstance, you know,

Sam wrote me the nicest message ever.

194

He said, well, actually, we went to

the same college, know, like all the

195

same sort of things that you, yeah.

196

So, you know, all right, we'll

leave it, we'll leave it at that.

197

But, you know, he just wrote

out a very respectful, you

198

know, very respectful message.

199

Of course, his background, he

interestingly was one of the first, he's

200

actually, he's, Actually in the army

for another couple of years, at least.

201

And, and he was actually one of

the first people to go through the

202

VR training sort of program they

had in the army, the initial one

203

and for, for helicopter training.

204

And so, you know, that's a

really interesting experience.

205

He flies the.

206

Um, the uh, a plane as a, as a fixed wing

operator there and, you know, his 1st,

207

his 1st sort of attachment in the army,

if you will, was to go to MIT and get a

208

master's and so, you know, just a really.

209

a really good group,

initial group of people.

210

And so that's, that's a little

bit how, how we got together.

211

I'm actually surprised that I

didn't meet Sam while I was at MIT.

212

By all rights, I should have,

213

Bill: were there at the same time?

214

Alex: things, things that, yeah, we were

215

Bill: Okay.

216

Alex: time.

217

Bill: All right.

218

you're all pilots.

219

Alex: yeah.

220

Hopefully.

221

So, yes.

222

So, you know, Sam flies

223

The, uh, said is a student pilot

and I'm a commercial multi engine

224

instrument pilot, know, I'm looking

forward to getting my instructor

225

pilots license and learning about some

of those fundamental of instruction.

226

Oh, yeah, we love flight.

227

Yeah,

228

Bill: And did I hear

you say you fly a DA42?

229

Is that what you said?

230

Alex: that's what I fly.

231

Yes, I fly

232

Tiffany: Nice.

233

Alex: out of Philadelphia

northeast airport.

234

That's where my, that's where my family

lives is in sort of the Philadelphia area.

235

So I end up going down there.

236

I get to fly the 42.

237

Actually, 1 of the coolest things that

I've done in that airplane was the at the

238

eclipse that kind of ripped through the

239

Bill: Yep.

240

Yep.

241

Right.

242

Alex: up there with, uh, My, my,

my partner, me and her ended up

243

flying up in there and seeing

this amazing 360 degree sunset.

244

And, you know, you know, I was still

looking through these little glasses

245

and I was, you know, looked down at the

instruments and then I look back and

246

see this amazing sort of white Corona.

247

That's

248

Tiffany: Wow.

249

Oh my

250

Alex: and most resplendent way

you've ever seen in your life.

251

So I would recommend it not to the point

where you don't listen to, you know,

252

your TCAS or anything like that, but you

know, absolutely go, go check it out.

253

There was, you know, you'd be in good

254

Bill: I, I wasn't in the air for that

eclipse, but I did go to Totality

255

in Texas and it was one of the most

incredible things I've ever experienced.

256

So I can imagine it, it must've

been incredible in the air too.

257

Alex: You see this sort of wall

of darkness kind of flying towards

258

Bill: Yes.

259

Alex: some, you

260

Bill: Yes.

261

Alex: You think it's just like just

rain, like the thickest rain you've

262

ever seen in your entire life.

263

But

264

Bill: Yeah.

265

Alex: know, it passes you and then

you start to see the sort of the

266

light coming around the same way.

267

And just sunset in all

268

Bill: Everybody made fun of me for

my reaction to it because it made me

269

feel so small and so at the same time.

270

It was, it was incredible.

271

Yes.

272

Alex: Like, like within

273

Tiffany: Oh, interesting.

274

Alex: the heat on.

275

Bill: Yeah.

276

Alex: Yeah, within moments.

277

Bill: Interesting.

278

Alex: yeah.

279

Bill: Well, I don't know that

I'll be around for the next one

280

that happens in North America.

281

Um, cause I am kind of

an old guy, but, um,

282

Alex: ha

283

Tiffany: here we go again.

284

Bill: and I still have a medical, I might

have to try it from the cockpit next time.

285

that's a great idea.

286

Alex: Let's go.

287

I'll brave you.

288

Tiffany: There we go.

289

Alex: what.

290

Yeah.

291

Bill: Excellent.

292

Alex: In any form.

293

Bill: Well, thanks for

all that background.

294

So, um, one of the things you

mentioned, and you've sent us

295

information about Fly Shirley.

296

It's clear that, like you said,

you started with, with Sims.

297

Tell us, tell us what your approach

was with Sims and then where you're at,

298

uh, with the product as it pertains to,

uh, flying in, um, actual airplanes.

299

Alex: Yeah.

300

No, thanks for that.

301

So, Shirley is designed to be

your AI co pilot from Sim to Sky.

302

It's set up so that pilots can

just talk to Shirley naturally,

303

just like a co pilot friend sitting

next to them in the cockpit.

304

So surely provides training in the

SIM to help people prepare for lessons

305

and check rides and proficiency.

306

so our goal has always also been to

bridge people from SIM And so in the long

307

term, we're working towards being able

to have surely apply back in the cockpit.

308

so, you know, our approach starting with

the SIM has been to sort of home SIM

309

flight users, people who actually just.

310

You know, Flight Simulator released

a metric, 15 million users Microsoft,

311

uh, at Flight Sim Expo, they made this

new metric book at 15 million users,

312

Microsoft Flight Simulator, of this

amazing sort of trend of getting more

313

and more people involved, interested in

flight through this sort of sim community.

314

And so, you know, we've started

with a co pilot that just.

315

Really easily connects to We're

starting with explain because we're

316

friends with the explain guys.

317

all you need to do is you open a web

browser and then you go to flyshirley.

318

com and then you hit start call and.

319

Immediately, Shirley's talking to you

and sort of says, sort of airplane

320

teammates up, mic check when ready,

and then you get connect to explain.

321

You don't have to install any

plugin or anything like that.

322

As long as you just explain 12 and

then it just immediately connects

323

and has a window seat into your sin.

324

so we just start with this sort of,

you know, this a I kind of copilot.

325

Shirley's your copilot.

326

You talk.

327

It's really immediately replies.

328

You end up having this kind of nice

conversation, no matter where you're at.

329

Surely.

330

an idea of where you are and give you

info about where you're flying to.

331

So we sort of have three

different sort of modes.

332

We're starting here with this sort of.

333

Home flights in.

334

And so folks there, some people just

want us, you know, try out different

335

airplanes, cruise around different areas.

336

There's all sorts of reasons I'd be

able to sort of home flight sitting.

337

And then there's sort of a, we've

introduced a sort of challenge

338

feature as well, that, you

know, we can make challenges.

339

Other people can build challenges as well.

340

So the first one is this Aaliyah 250

flight training type of challenge where.

341

You know, explain ships with the

model of the alia to 15 beetle mode.

342

And so with that challenge, essentially,

surely is stepping you through, you

343

know, the controls check is stepping

you through this sort of initial.

344

How do you hover?

345

How do you have control in that sort

of mode and then how do you transition

346

out and sort of transition back in and

so surely kind of walks you through

347

this and that sort of automated fashion

challenges are really set up for other

348

people to be able to kind of build

fun, little interactive capabilities.

349

And we have the sort of for people who use

chat or whatever we have already sort of.

350

You know, example, example, prompts

that you would just follow and be able

351

to create these things and things like

skull challenges that possible, sort

352

of this, you know, you can have landing

challenges, different sorts and explore

353

different areas that you're interested in.

354

also just launched on Friday.

355

We just launched a sort of the first

version of our private pilot curriculum.

356

so that's really the sort of

the first step into training.

357

And private pilot, uh, curriculum is

really interesting that we end up being

358

able to teach people about some level

of, know, taxiing, you know, from the

359

fundamentals of, you know, what do you do?

360

What does it feel like to

actually take a lesson?

361

But to also let people be

prepared for their lessons.

362

So let's say you just went out and did

sort of in terms of at a point with

363

your instructor, and now, you know,

you're going to go make sure you got

364

it by next week and you're going to go

either your instructor is either going

365

to test you at the beginning of the

lesson and then have you do something

366

else or they'll test you at the lesson.

367

And if you don't do it, they're

going to make you do it again.

368

And so maybe you'll go and prepare for

that lesson and be able to make sure

369

you really nail it when you get back.

370

And one of the things that.

371

We provide with surely is sort of

feedback while you're doing the

372

maneuver, but we also keep the sort

of a data log of you doing that

373

are able to sort of chart it out.

374

Surely is able to talk to you about

sort of the, you know, the sort of the

375

flight data and sort of your results.

376

You have sort of a indication of

what would be to the ACS standards.

377

So some people would really like to be

able to prepare for their, for their

378

sort of check ride, because it's pretty

expensive and it's pretty high stress,

379

you don't want to, you know, mess up.

380

And so, you know, to varying

levels, depending on the fidelity

381

of somebody's sim and their setup

and all these different things, you

382

know, we can provide value based

on our private pilot curriculum.

383

And so that's live right now.

384

And then we're also sort of working to

bridge our way into different levels

385

of kind of commercial environments.

386

have a sort of a pilot program with.

387

With actually beta starting September,

where we're going to be providing.

388

That's sort of, you know, training, sort

of familiarization to the employees of the

389

company so that they can, you know, when

they have VR seats throughout the company.

390

And so when an employee comes by, they

can just sort of sit down in the seat and

391

be able to try it out, be able to like,

how do I actually fly this airplane?

392

I kind of forget I had it on my

first day as a lesson in but I

393

really just want to, like, see what

it's like in order to fly again.

394

I went, you know.

395

And it's really fun and kind of

like interactive and and can be

396

pretty inspiring for the employees.

397

looking at commercial

opportunities at various levels.

398

For instance, if somebody has a, a

aviation advanced training device that.

399

You know, the, one of the really

interesting trends, I think, is that

400

those are starting to be reduced in price.

401

They're now also, you know, you

know, there's also VR type AATDs,

402

you know, loft dynamics is a company

that is certified, uh, FAA now

403

for one of their, for an Airbus.

404

And so one of the cool things is that as

instructors are really in high demand,

405

just like pilots are really in high

demand, surely can provide training that

406

prepares people for their instruction.

407

and one of the things that we're

looking to do is be able to provide

408

sort of both surely providing

instruction, but also having the sort

409

of channel to allow sort of remote

instructors for people to check in.

410

This is sort of especially interesting

for commercial operators, like, you

411

know, companies at the same scale as.

412

JetBlue or American airlines, United,

et cetera, where you have to spend

413

multiple weeks on site when you're

trying to get a new tech rating.

414

And so if you were able to spend some

time at home with your family, not

415

have to, you know, from the eyes of the

operators have to do the spend of being

416

at a hotel per diem and all this other

stuff running around the flights, et

417

cetera, you know, displacing revenue

fares, know, then, you know, you'd

418

be able to prepare, show up prepared.

419

Have check ins with instructors remotely

as you're preparing at home and then sort

420

of, you know, show up for the 1 week where

you need to have the full flight simulator

421

and be able to have your 20 hours.

422

And then, you know, you're good to go.

423

But, you know, this is sort of things that

we're looking at and sort of, you know,

424

on the track towards getting be able to

bring Shirley back into the cockpit where

425

actually you begin by sort of taking.

426

Some of the learnings and, you

know, working with some of, you

427

know, the, the, the, um, some of the

folks that are amenable to saying,

428

yeah, like I support the mission.

429

Like, I'd love to be able to make flight.

430

Safer, you know, you know, I'm in for,

uh, I'm in for sort of stuff like that.

431

Then we're able to make some of these

features that surely has in terms of

432

interactability, conversationality

available on, on an iPhone so that

433

pilots can talk to Shirley while

they're flying and that can provide.

434

A lot of sort of overhead, especially

in the GA context to begin with

435

where, you know, task saturation,

et cetera, is really hard.

436

And sometimes like if you're flying

in instrument conditions and then

437

something breaks and now you need this

number, you can't remember the thing.

438

And it's just, it just ends

up being a mess where you can

439

just ask for the information.

440

And this ends up being kind of a kind of

thing, but we're kind of working our way

441

back to there, which was where we started.

442

And I think along the way, we're

providing a really revolutionary

443

training product that.

444

a lot of value for people.

445

Surely folks have tried out at flyshare.

446

com, but it's, and there's a free

trial for that, but it's, it's a really

447

engaging sort of system and doesn't,

think quite a good job at being able

448

to provide a level of training that

helps people prepare for check rides

449

and help them with proficiency too.

450

Tiffany: Can you do a, could you set up.

451

Bill: into it.

452

Yeah.

453

Alex: Right.

454

So one of the things that we're

working towards, so we, then

455

there's two parts of that.

456

One of them is sort of the, the.

457

Oral examination part of that.

458

And no, that part is actually pretty,

pretty straightforward and pretty easy to

459

do because, you know, there's some aspect

of a question bank, surely has been,

460

you've been talking to Shirley, Shirley

has some ideas of things that you're,

461

are your weak points over, over time.

462

And so you can kind of have

a conversation and be sort of

463

grilled in the service, similar way

that you'd be grilled in person.

464

And, and then there are those

maneuvers aspects of it.

465

And the maneuvers aspects of it is

what's really neat about the ACS is

466

that it really lays out like the precise

parameters, you know, plus or minus

467

50 feet, you know, of course, they

say things in the ACS, like you can

468

temporarily exceed that,

you know, these thresholds.

469

But if you consistently exceed

them, and that's a, that's a fail.

470

So there's like, there's some

soft barriers, and that's

471

actually to be totally, totally.

472

On this, one of the really cool

things about using this sort of new AI

473

technology is that it has a lot more of a

holistic view on how things are working.

474

And so when you provide this sort

of graph that shows how tolerances

475

are exceeded temporarily, et cetera,

then you really do get kind of like

476

hear the whole story about, yeah,

this is if you're going to focus and

477

you had time to focus on something.

478

This is what you should focus on.

479

But overall, you might like

depending on the examiner, then

480

you probably would speak by.

481

And so, you know, there's stuff like that.

482

And, you know, there's so absolutely.

483

That's 1 of the things that

we're looking to provide.

484

I think 1 of the major sources of value

on top of helping people to prepare.

485

For their particular lessons

and and aspects like that.

486

Of course, there's different depending

on the equipment that they have.

487

And I think,

488

Tiffany: Yeah.

489

Alex: maybe we'll get into

this a little bit more.

490

There's different things that

are probably best suited to

491

help their training at home.

492

You know, a lot of people who are

pilots have iPads, probably the

493

majority of them by now or or iPhones.

494

And so there's sort of, you know,

equipment amenable to that, you

495

know, the intersection of people

who have amazing flight sims that

496

have monitors like wrapping around

or VR headsets with like world edge

497

leading kind of gaming computers.

498

That's a much smaller subset of folks,

but, you know, there's, there's these

499

sort of interesting opportunities where.

500

You start to be able to expand and sort

of the trends in the industry are such

501

that you start to be able to see that

really good representative Sims are sort

502

of making it out in the customer's hands.

503

Microsoft Flight Simulator gets better

and better, know, as does sort of

504

the number of computers that can run

something like X Plane and, you know,

505

There's sort of VR headsets get really

a lot cheaper and a lot more accessible.

506

You know, there's other aspects

related to, of course, related to

507

AI and also certified sims becoming

more affordable and things like that.

508

I think are really

interesting trends right now.

509

Yeah, thanks.

510

Bill: those are all sort of, those are

all sort of on your, I guess your roadmap

511

target list to investigate how Shirley

specifically can enhance that experience.

512

Yeah.

513

Yeah.

514

Yeah.

515

Alex: so you look at some, some folks,

even some streamers, this person named

516

Swiss001, who made a video about Shirley.

517

It's pretty funny.

518

could link it or something, but

the, the, The thing that he says

519

is that he got to start by, you

know, learning with infinite flight.

520

And so he got interested in being

able to fly because he had this

521

iPad app that showed him how to fly.

522

you know, they're sort of, they have

a, they have a data interface for sure.

523

They could very easily have a window

seat into that sort of experience.

524

don't, you know, their planes,

they don't have, they don't have

525

sort of a very interactive sort

of FMS or something like that.

526

But they do have sort of like a

navigation system and things like that.

527

So there's different levels and it's

sort of like also Microsoft lights.

528

And they've been around the Xbox.

529

You know, people can be able to fly

using a remote, but or like a game pad.

530

But, you know, they can also attach sort

of yolks and things like that to it.

531

you know, you can look at

the folks that have sort of a

532

sort of a single monitor or.

533

MacBook Pros now can all run X Plane

12 like really well because they have

534

these amazing processors on them.

535

And, you know, even with this either on

the, on the laptop or with the monitor,

536

that makes it a really, really quite

reasonable IFR platform for, for training.

537

if you want to be able to do something

where you have representative

538

maneuvering, then you either need to

get really fancy with the way you sort

539

of map, you know, keys on your yoke

for looking left or looking right, or

540

you need to have multiple monitors.

541

But then you can start to use something

like explain, which has a really

542

impressive flight model, even for

slow flight in order to be able to

543

see kind of realistically touchdowns,

takeoffs and be able to some sort

544

of stalls and things like that.

545

And so, you know, there's just, there's

just are a total variety of different

546

platforms where, you know, uh, where the

application of surely makes some sense

547

where, you know, surely can provide

helpless sort of instrument procedures,

548

give you, give you the numbers, help you

set up for the approaches, et cetera, sort

549

of judge how you did, where did you break

out, you know, where did you knock off?

550

Like things like that.

551

And, and, you know, um, and we're, and

we're also looking at how to engage.

552

With, you know, how to bring

flight instructors, of course,

553

to are such an important aspect.

554

We really see surely as a, as a tool

to help people prepare for those

555

lessons, especially since there is

such a pilot shortage, you know,

556

557

Tiffany: Yeah, huge.

558

Alex: that.

559

They're only getting worse.

560

Tiffany: Okay, I have a

quick layman question here.

561

So if I had a flight simulator right

now, which I do not, what kind of

562

guidance does it come with already?

563

Or is, all of a sudden, Fly

Shirley like this holy cow moment?

564

You know, my nephew, he's working on

his private right now at Cal Baptist.

565

So he takes his checkride in September.

566

Um, but, like, let's say that he

gets a flight simulator and he

567

jumps on it and he's not a pilot.

568

Is there any sort of, um,

tutorial or information?

569

Like, how do people learn how to fly

on a flight simulator besides just

570

game theory and trial and error?

571

Alex: you know, we don't, we don't

exist within a vacuum for sure.

572

You know, there's, there's actually

a lot of precedent for both, both

573

apps that people can add on for doing

training on their, on their sims.

574

And also, of course, the, there's a

whole cohort of people who have been

575

doing sort of training content for sims.

576

There's just so many like amazing

creators that do this sort of thing.

577

So.

578

Tiffany: But those would

live on YouTube or?

579

Alex: yeah.

580

And so actually a lot of people to

be interested in doing Sims things

581

because they were somehow kind of cross

marketed on YouTube and had sort of

582

a video that shows up and it's like,

Oh, here's this really cool person

583

doing like, you know, beetle VR, which

is this amazingly accessible sort of

584

You know, the fighter sort of like

but it's really easy to get going.

585

Or somebody sees an X Plane demo

where somebody is like, you know,

586

really being super concentrated,

flipping switches on, you know, seven,

587

three, seven, or something like that.

588

And people want to get involved with that.

589

And so, you know, what's

interesting is that for sure we

590

don't exist within a vacuum, so.

591

There are, you know, there have

been historically companies that can

592

provide a sort of standards based,

very specific kind of add on that

593

sets you up in a particular maneuver

and then looks to see your altitude

594

and sort of like really specifically

has these triggers and is just very,

595

very precise and gives you sort of.

596

Like a percentage like score based

on whether like how much and how

597

long you were within tolerances.

598

And if you're exactly at 100 percent the

entire time, then, you know, then the

599

best, but, you know, so, you know, there

was a whole generation prior of people

600

who have been trying to build sort of, you

know, tools to help people with training.

601

I think it's been hard to call those sort

of teammates because they're not really.

602

You know, interactive or copilots because,

you know, there are, there have been

603

sort of historically there's a, a tool

called Linda for the explain explain 11

604

and 12, which is sort of like a copilot

that's sort of for the zebo mod, which

605

is the popular kind of mod for a 737 that

adds a lot more interactivity and a lot

606

of really cool things and explain, but,

you know, what Linda does is be able to

607

sort of You know, you can use basic voice

recognition that sort of matches commands

608

be able to basically, you know, flip

609

switches for you and things like that.

610

You know, the, the previous company

I was talking about was sort of the.

611

Sort of, sort of kind of laid out

sort of, um, parameters is something

612

called, um, take play interactive.

613

Even like Sporty's sells

that sort of thing.

614

You know, there are really kind of

innovative companies that also sell You

615

know, Sporty's is a good example of that.

616

For instance, they.

617

You know, they are working with infinite

flight to try to provide lessons to

618

certain degrees in infinite flight,

like they have their own sort of

619

explained scenarios sort of bundle.

620

They, of course, have all their training.

621

I did my instrument sort of ground

school with them, and they also

622

sell Microsoft Flight Simulator

apparel for what it's worth.

623

But the, the, the interesting thing

about some of these opportunities,

624

I would say, in these companies.

625

Is that, you know, they

are not as interactive.

626

I mean, fundamentally what's, what's

been the new kind of breakthrough

627

with all this AI is that now you

can just have natural language.

628

Be able to map into the whole space of

possibilities for what you're going to do.

629

And so right now, surely, for

instance, you could just, know, for

630

the home flights and market, you can

ask Shirley to put down the flats

631

to like bring up the gear to turn

off the lights and things like that.

632

Like, you know, all these sort of,

sort of, you know, fundamental things.

633

And just by using natural language

and you can sort of describe things,

634

you can say, do this and that, and

that, and all those things happen.

635

Yeah.

636

Versus, you know, the issue with

sort of speech recognition up until

637

this point is that you just literally

are translating speech into words,

638

Tiffany: Right,

639

Alex: not sort of like an understanding

layer, whereas as people, we just sort

640

of like hear a bunch of stuff, and then

we have like an idea of what already

641

they wanted and sort of like, you know,

we walk into a coffee shop, and you

642

could probably mumble some things and

get a coffee, and, you know, that's

643

maybe not the sort of deal, just sort

of a basic sort of speech recognition.

644

So it's sort of the new,

sort of like the new.

645

Um, kind of, uh, standard is to

kind of combine speech recognition

646

plus some sort of intelligence.

647

Tiffany: right.

648

Alex: And that's absolutely what

we're doing with Shirley and have

649

sort of a, uh, a leg up on that.

650

Yeah.

651

Bill: Cool.

652

Tiffany: That's, this is awesome.

653

I'm loving it.

654

I, sorry Bill, I, I ended up getting A

complete sinus infection for two weeks

655

after my instrument, which I passed,

but I was just terrified of that one.

656

And, you know, as you're talking,

it's like, man, something like

657

this would have been so helpful.

658

Just to sit with somebody else by yourself

that you know is not a real person

659

that you have to face is going to go.

660

God, that chick's an idiot.

661

So, uh,

662

Alex: Yeah.

663

The

664

Tiffany: thing

665

Alex: the instrument that

can be brutal because

666

Tiffany: else.

667

Alex: have any random symbol on anything

and it's just like, Oh God, I don't, never

668

Tiffany: Yeah,

669

Alex: the DME arc and all, okay.

670

Yeah.

671

Totally agree.

672

Tiffany: I don't know if we want to

transition to this, but I am so curious.

673

Uh, you had made a comment

as to why you started this.

674

What really brought that idea on?

675

Um, I'd love to hear that story.

676

Alex: Yeah.

677

So.

678

Yeah, it was actually a kind of a, a

tragic, it was a tragic event, really.

679

I mean, I, a friend of mine

crashed his cozy and ended

680

up passing away through that.

681

And you know, when you, when

something like that happened, you

682

just sort of get kind of shaken out

of whatever you're doing currently.

683

For me, it was sort of, I was

running a flight sim team at,

684

at, Um, and Merlin was a great

aviation autonomy company in Boston,

685

I sort of saw that there could

be an opportunity to make

686

aviation even somewhat safer.

687

And, you know, I don't know, sometimes

when you feel really powerless, you

688

just try to figure out something you can

689

Tiffany: Right.

690

Alex: to help in some way.

691

Right.

692

And so I thought, like, maybe there's

some sort of a way to apply this new

693

technology and and be able to see if.

694

Get it make safer and while still

appreciating aspects of flight

695

being really worthwhile and fun

and sort of, you know, kind of

696

kind of wondrous in a lot of ways.

697

And I think that's 1 of the things that

that flying through the eclipse made me

698

remember was was that, you know, you know,

Yeah, flying is just incredibly beautiful.

699

We notice it every a lot of a lot

of afternoons and evenings when you

700

find in the sunset or something.

701

And so, you know, starting by trying

to see, even just from a very direct

702

way, does this actually improve fine?

703

Is there an opportunity to actually make?

704

safer and so things that I noticed

immediately for this from a GA context

705

were when there were issues that required

digging through a POH could be, you

706

a DA 42, you know, things that were

707

like, related to numbers that you had

to keep on recalling that, you know.

708

You know, might not be able to do if

your hands are actually on the controls,

709

or if you're on a pilot, it's maybe

unreliable, or if you're in the soup and

710

you don't want to like move your head into

weird directions, of opportunities where

711

folks in particular don't necessarily

have a teammate a copilot, right?

712

If you're flying up a buddy, or you got a

friend, or you can do a lot of this stuff,

713

and that's like, that's pretty great, but,

you know, having an opportunity to have.

714

You know, something you talk to about that

the flight help you kind of think things

715

through ended up being a pretty useful

thing, at least from my perspective,

716

and it'd be pretty be pretty helpful.

717

But you have to make sure that it's safe.

718

There's no exceptions.

719

Like, you have to make sure that.

720

It's not making stuff up.

721

It's not acting wild that it can

hear you, that it runs on your phone

722

without melting your phone or making

it like immediately drain and die

723

from battery or something like that.

724

And so, you know, starting with sims.

725

And working your way back to flight seemed

like the most responsible way to doing

726

that, because, you know, you'd be able

to have the advantage of having a huge,

727

a huge cadre of people in the sort of

sim land to be able to try out the things

728

that you're building in service of being

able to make something work in flight.

729

you know, you'd be able to see whether

those you'd be able to see really quickly.

730

People ask, you know, I've learned that

people sort of celebrate that surely can

731

talk about sports and things like that.

732

And

733

Bill: Yeah.

734

Alex: know, I see lots

of fun stuff, you know.

735

So I learned from Shirley about

football formations that like are,

736

Tiffany: that's funny.

737

Alex: and it's true, you know, Shirley

knows about all sorts of stuff because,

738

know, there's a enormous sort of

intelligence model behind it and,

739

and so, of course, like, you know,

even Shirley can tell you about the

740

best restaurants near the particular

airport that you're flying over.

741

Don't expect necessarily that

Shirley off the bat would be

742

able to do that on your phone.

743

But, you know, because it

has to run offline in your

744

pocket, um, on your dashboard.

745

But, you know, we're kind of building up

towards being able to provide things that

746

are safe and also still conversational

747

Bill: Yeah.

748

Alex: it's when, when you're able to

have that context, it's able to hear

749

hopefully, or, you know, ideally also be

connected into some message to the panel.

750

So there's a lot of

things to be figured out.

751

Along along that way.

752

However, I think the sort of the value

of having something or certain or

753

surely or something that you talk to

while you're flying is is absolutely

754

there, but it takes a lot of work.

755

You got to get it right.

756

Bill: Yeah.

757

Right.

758

Alex: to

759

Bill: takes.

760

Alex: way up.

761

Bill: Um, not only practice um,

learning on the model and the,

762

um, intelligence models part, but

it's going to take learning and

763

practice from the pilots as well.

764

So if you don't mind, um, I'm going to

branch off into a couple of different

765

areas and I'm going to start with

the product itself and the, the vast

766

opportunity tonight for me to geek out

with you is there, I want to do it on the,

767

on the Aviator and educator side, but I'm

going to hold off on that for a second.

768

I want to geek out on the actual

product a little bit, if you don't mind.

769

So I'm going to ask you a series of

questions that take us down that road.

770

Um, first of all, um, I know

this is an overloaded term.

771

You're using a large language model.

772

You're using, uh, you're

using an intelligence model.

773

Um, do you call it AI or do you

stay away from that overloaded term?

774

Alex: That's a good that's a

that's a reasonable question.

775

I think it is, okay, so I think

AI has become the suitcase

776

Bill: Yep.

777

Alex: defines the category of things

that use large language models, you

778

know, there's a whole kind of, there's

a whole cadre of people that go

779

into all sorts of rabbit holes about

artificial intelligence, and then they're

780

starting to worry about, you know,

plotting robots and things like that,

781

that don't make a lot of sense to me.

782

But.

783

Okay.

784

You know, starting and I think,

you know what I'm talking

785

Bill: Yeah.

786

Alex: there's, but there's absolutely

something that makes surely a,

787

I mean, you can talk to Shirley

about anything that you want and,

788

you know, there's totally the.

789

you know, surely we'll try to

bring you back into the light.

790

That's that's really, truly just kind

of wants to fly with you and maybe not

791

that you can see some of the videos

that creators make and they try to push

792

Charlotte in a different direction.

793

The trouble is like, maybe you want

to fly because that could make you

794

feel better and, and trying to kind

of get away from like something.

795

But, um, so, know, um, that's kind

of my philosophical point on that,

796

you know, And we're not, you know,

797

Bill: You're not against it.

798

You're not against using the

term, but you're also not

799

defining the product as just AI.

800

Is that fair?

801

Alex: Right.

802

It surely is your, yeah.

803

I mean, we say it's on our poster.

804

I mean, surely it's your

AI co pilot from Sim to

805

Bill: Okay.

806

All right.

807

Alex: so, you know, You know,

we're providing a service

808

where surely it can be your co

809

Bill: Perfect.

810

Well,

811

Alex: say something like

it's, you know, it's Shirley.

812

Bill: this is the definitive

test on if it's AI.

813

Do you use Lisp anywhere

in your, in your product?

814

Alex: So,

815

Bill: I couldn't resist.

816

Like I said, I'm an old guy.

817

So,

818

Alex: written,

819

I have written programming languages

and compilers and I programmed

820

in a language called Haskell

821

Bill: yes.

822

Yeah.

823

Yeah.

824

Alex: nope, we're not, we're not using

825

Bill: That, that was a joke.

826

I did program in list points, but

it was, but it was in college.

827

So that was a long time ago.

828

Alex: We're trying to be a successful

829

Bill: Yeah.

830

Right.

831

Right.

832

Alex: So

833

Bill: All right.

834

I couldn't resist.

835

So a serious question though,

how much, how much of your model

836

is running on the edge and how

much is, um, bound to the cloud?

837

And how are you, how

are you navigating that?

838

I mean, it's a, it's a tough balance.

839

Alex: Yeah.

840

So some of the things that are

really interesting about that.

841

So right now, you know, the quickest way

to start a company is to, is to use all

842

cloud services in this space for sure.

843

Right now, you know, what we're,

what we're doing is we are sort

844

of setting ourselves up to be able

to progressively run things more.

845

Offline and sort of what we're beginning

with that is to build a system that,

846

know, kind of encourages surely to be a

little bit more judicious about replying.

847

You know, 1 of the interesting things

about these sort of large models, you look

848

at either, you know, if you go to tattoo,

you go to claw or something like that.

849

you ask a question, it's always

going to reply to something.

850

And if you say, let's simulate a

conversation and you say, okay, thanks.

851

And it says, you know, see ya.

852

And you say bye.

853

And then it says And then you say, okay.

854

And it says, all right, let

me know if you have a state.

855

And so like, you know,

that can be sort of a mess.

856

And

857

Bill: It's like saying goodbye to

your significant other in high school.

858

Tiffany: You hang up.

859

Bill: Takes all night.

860

No, you hang up.

861

No, you hang up.

862

Alex: Shirley.

863

No, you hang up, Shirley.

864

Enjoy your flight.

865

I am.

866

Yeah.

867

Great.

868

And so, you know, that's, that's

something that, you know, you watch

869

them, some videos, you know, there's a.

870

Wonderful channel, virtual aviation

aviator, uh, Frank's providing a

871

ton of sort of feedback to us and

just, you know, watching his videos.

872

Of course, you learn that things that can

be a little bit challenging, like surely

873

being over ambitious about replying.

874

And so, you know, the 1st,

kind of the 1st sort of.

875

Segment for building things that go more

towards the edge are sort of models that

876

sort of tell us whether or not she needs

to reply or not, and things like that.

877

And so, and when and things like that.

878

And so, you know, start with that.

879

And then, you know, of course, the

model, which is going to run on on the

880

iPhone is going to run entirely offline.

881

So so that's really good.

882

But I think what's What's neat is

the, the amount of progress that these

883

sort of sort of hosted cloud models

have been making is just phenomenal.

884

I mean, right now you can almost, uh,

you know, right now, surely can totally

885

comprehend these sort of data graphs that

we're, that we're providing and there's

886

sort of charts and things like that.

887

And, you know, all sorts of

things you find in the AFD that.

888

Just by looking at them, you know,

there could be some sort of evidence.

889

So, you know, there's lots of real

benefits to using things that exist on,

890

on the cloud, as well as they have really

well defined sort of, you know, they are

891

respecting the rights of, you know, users,

et cetera, related to privacy and not,

892

not necessarily training on their data.

893

So we could kind of do it in a very

judicious way where, you know, um, you

894

know, where we're, where we are, of

course, we're collecting some of the

895

text of the conversations, but before we.

896

Do anything with it, you

know, we're we're, yeah.

897

Anonymizing it

898

Bill: Yeah.

899

Alex: and you know, sort of hitting

it with the past to make sure

900

there's nothing, wild in there.

901

If we kind of encourage users not

to share social security numbers

902

and things like that, Shirley, but

you know, just 'cause it did then

903

sort of hit it, hit it with that and

we'll, we'll also, we're still early.

904

They both figure, figure out a way

of letting people sort of opt out.

905

But I think

906

Bill: Right.

907

Alex: is something that people

will hopefully opt, to opt into.

908

'cause we're looking

to be able to provide.

909

Sort of training, they were

training, but also surely in flight

910

Bill: Yeah, yeah, yeah.

911

Alex: to do that is by being able to

use some of these conversations and make

912

something that's really small behave a lot

913

Bill: Right.

914

I mean, these are all kind of typical

challenges that you have around AI and,

915

and large language model learning, right?

916

Um, so has latency proven to be

somewhat of a, uh, an issue that

917

you're dealing with in your product?

918

Or are you finding that it's not a, it's

as big a challenge as maybe it would seem?

919

Alex: latency has been so it's

a little bit longer when an

920

image is in the conversation.

921

You know, then it can start

to feel like, you know, you're

922

waiting for a second or something

923

Bill: Mm hmm.

924

Alex: but our current latency is something

along the line of is less than a second.

925

It's like, you know,

600, 800 milliseconds.

926

it ends up being pretty fast.

927

In fact, you know, one of the, one

of the issues and the reason why

928

we're focusing on building something

that tells Shirley when to not reply

929

Bill: hmm.

930

Alex: is, is that latency is so low.

931

And so, you know, it ends up being a

932

Bill: Yeah,

933

Alex: very conversational

sort of interaction.

934

In fact, the videos that I sent her,

935

Bill: yeah.

936

Alex: on, on doctor, there was

no time warping in them and,

937

and, or anything like that.

938

And so.

939

You know, I think people are very, you

know, very pleased with the latency.

940

But, you know, there are, there

are issues where it's a little too

941

low and so, you know, so, you know,

interjecting at the wrong times, you

942

know, things interrupting or things

that shouldn't be happening like that.

943

And so we're, we're focusing on, you

know, fixing those issues without.

944

Making latency too much, although

those might increase the latency

945

temporarily while we sort of

figure out how to do that better.

946

So this sort of thing is kind of

bread and butter, but you know, it's,

947

it's an important aspect of making

a copilot that actually feels like a

948

copilot as opposed to somebody who's

the high school, as you say, the high

949

school kind of lover saying goodbye at

every time you're having an exchange.

950

Tiffany: When you're planning

On making the version for,

951

uh, actually being in flight.

952

We'll, we'll surely be

listening to the radio as well.

953

Will that be the helper that says,

Hey, I think they called your,

954

called you and you didn't reply.

955

How's that working?

956

Alex: So, you know, you could see it

with some of the existing companies.

957

In fact, that's, you know, there

are some very good examples of

958

companies that are providing that

you can talk to via bluetooth you're

959

while you're flying and they can

listen to you while you're flying.

960

The challenge is if you also

want to listen to the intercom,

961

so you have to start to have

adapters and things like that.

962

And as a software company, you don't

want to get into really the business of

963

building adapters or things like that.

964

You want to use sort of off the shelf

adapters and And so, you know, there

965

are some sort of hiccups to doing that.

966

However, you know, I've talked

to some avionics makers.

967

very interested in being able to provide

high quality transcriptions of things that

968

are happening, you know, over the radio.

969

And so one of the hopes that we

have is also through our process of.

970

to develop surely that we can provide

something maybe to some avionics makers

971

that would allow them to better do the

2 stage sort of process of 1st speech

972

to text and then sort of like a kind

of a rough, like an overview pass and

973

editorial pass of, you know, turning

that don't make sense in certain

974

places and the things that make sense.

975

so, you know, if we could sort of provide

that along the sort of along the way,

976

I'll provide a lot of value to people

being able to just sort of see whether or

977

not something was for you or maybe being

able to incorporate a, a directive into

978

your FMS directly or something like that.

979

So there could be a lot of value there

980

Bill: Yeah, that first step is to

digitize it, um, and then you can start

981

analyzing it, but I guess taking that

first step is where you're at right now.

982

Alex: right.

983

Bill: so how, how does the, uh,

intelligence model, uh, behind

984

Shirley specifics about the POH for

your airplane, for, for example?

985

how does it, how do you teach Shirley?

986

or how will you in the future, um, so that

it's getting the right version of the POH.

987

You can't just get this

generically off the internet.

988

It has to be for your

serial number, right?

989

It has to be for any number of things.

990

And I've got a follow up question

on other data as well that it needs.

991

So go ahead.

992

Alex: absolutely.

993

No, that's that's great.

994

So if we're talking about the version.

995

Of so if you're working with like a

flight school and you're trying to give

996

representative training to you know,

the students at the flight school,

997

or if you're working with providing a

version of an app that then pilots fly

998

with, you have to get the right numbers

just sort of, you know, make something

999

up that, you know, I heard on the

Bill: It can't be a chat

GPT type answer, right?

That's, that's what I'm getting at.

Alex: right.

Yeah, exactly.

Exactly.

Right.

Even though those sorts of

answers end up being quite

good at sort of a generic case,

Bill: Yes.

Alex: down like we do, and surely

we're just like, oh, you're flying

a 172, you're like, okay, 172.

Okay, well, it'll be about this, right?

So the process that we'll do

is to have a place where people

can ingest their own checklist,

Bill: Perfect.

Alex: sort of the POH.

We'll have a system that allows somebody

just to drop the PDF, of, you know, do

a first pass, sort of digitizing that.

something, clean it up for them.

You're like, here's the sort of

information that will appear.

If you ask questions, it'll be from here.

there's a lot of processes that you

can follow in order to make sure that

you're getting the right of data.

So one of those is, you know, tell the

system to not actually give them, give

a number because it can just make up a

number instead, it'll give sort of the.

Sort of index the place inside, like the

section number where that number is found.

And you sort of like, you know, label

that number inside of the POH somewhere.

And so you just, it outputs the

index and then you, before you speak

anything, you go and you grab that

information from the actual book.

And then, so you can kind of

make sure that things are, you

know, working appropriately.

And so this is the absolute stuff that

needs to be done prior to shipping in the

Bill: Yep.

Yep.

Alex: precisely, but also very

important to flight schools.

Before that ends up, you know,

know, shipping out for them, but for

sure, it's very important, you know,

before getting into the cockpit,

one of the things that will also be

implementing is, you know, feedback.

So somebody has a problem with the

number, they're surprised by something

is important to be able to capture that

and be able to be able to essentially.

Incorporate that and say, okay, maybe we

should take a look at how this performed.

you know, feedback in sort of human

Bill: Yeah.

And maybe

Alex: good,

Bill: a safe word, um, when

she's leading you astray, right?

Like, okay, leave me alone for a while.

Alex: The safe word in flyshirley.

com right now is by all

means, definitely mute.

Bill: Oh, okay.

Perfect.

Alex: is the safe word here.

Bill: Excellent.

Okay.

So my related question then is what

about, What about future integration?

Maybe you have it now, but I'm guessing

future integration with other real time

data that might prove handy to assimilate

into answers and so on, like ADS B data,

GPS data, um, data that's coming from

ForeFlight or, you know, things like that.

Kind of, how are you

taking that into account?

Alex: Right.

And so we're starting by using this

really convenient interface that explained

just launched called the web socket

Bill: Mm hmm.

Mm

Alex: not have to install a

plugin in order to be able

to have data go to Shirley.

And so that's been really very convenient.

We're looking at releasing

an app version of Shirley.

You know, there are other, there

are other sort of platforms that

provide these sort of data streams.

So ADS B is a good example, but

you can have a data stream from

infinite flight, which is the iPad

Bill: hmm.

Alex: in.

Sort of one of the leaders there, you

know, you can incorporate surely that way.

Of course, in the cockpit, you can have a

have the GPS from the sky onto your phone.

But of course, you and

the altimeter actually.

Now there's also altimeters

built into into phones.

Believe it or not, there's

barometric sensors.

But yes, the ADS B that comes

in has a lot more reliable data.

You should use that.

One of the things that we've been

focusing on as well, and one of the

reasons why we are sort of doing

this sort of data process is that, so

that surely will pay more attention

to the particular aviation context.

And so one of the things that we've

noticed is that existing really big sort

of language models are not very good at

paying attention to the state of flight.

And so even though you might have

information related to, of your altitude

and your descent rate and, you know, your

bank angle and things like this, you know,

the system might not fully be appreciating

the fact that you're about to land or

you already have landed and instead

relying more on the story to think about.

Bill: Mm hmm.

Alex: you've landed or not.

And so part of that is that these

sort of models like to reason,

sometimes you reason best by doing

this kind of chain of thought thing.

It's like, you know, you could tell

it this, you could say something

like, okay, surely using the

information, the information that

you see related to altitude and

rate of descent, what would you say?

We're in the herd phase of flight.

And then it could say

something along the lines of.

Descent rate is 500 feet per minute.

You know, altitude is 500 feet AGL.

We're about to land.

And, you know, what is really important

is to be able to skip those steps and just

for sure, to be able to understand that

Bill: Yeah.

Alex: about to land.

Bill: Just

Alex: And so bigger and

better and badder models

Bill: Yeah, right.

Alex: able to appreciate those factors,

but also by tuning the models using

that sort of information, you're

able to skip those steps and how

Bill: Yeah.

Alex: attention to the particulars.

Bill: we've, we've got,

Alex: opportunities to do this.

Bill: sorry, go ahead.

Yeah.

Alex: Well, there's other opportunities

also to do things called multimodal

models for, you know, essentially

1 channel is the flight data,

Bill: Mm hmm.

Alex: then the other

channels are the speech.

And, you know, you could have

two different channels feeding

the model so that it all points.

It has some concept of, you know, your

state of flight that it's generating

and also some concept of the story and

Bill: Right.

And that's kind of what I had in my mind.

Alex: it into text land too.

Bill: Right.

And that's sort of what I had in

my mind, because we have the, we

have access to these relatively

inexpensive adhars and things like

that, like in the century, right?

So it, that's kind of, is that

kind of part of your roadmap is

to have that multimodal approach.

Alex: and that's 1 of the cool

reasons for starting with Sims 2

is because they have all that same

Bill: Got it.

Alex: You're just getting

it streamed from the flight

Bill: Exactly.

Alex: could absolutely trade to pay

attention to this thing, especially

for a smaller model, which would

have a lot more trouble paying

attention to the correct things

without tons of prompting, et cetera.

And so being able to provide that

sort of these sort of pairs and being

able to train these certain models to

pay attention to their phase of life.

That's a, that's a huge

Bill: It's very interesting.

And you mentioned,

Alex: question.

Bill: you mentioned these massive

models that, you know, they're going to

get supermodels and things like that.

not the way I thought about

those supermodels in high school.

It's different.

But, um, so I thought

you'd like that, Tiffany.

Okay.

Um, so, so my question then is,

I think you're iOS only, right?

Today?

Alex: So right now we're X Plane

Bill: Uh, explain only.

Okay.

Alex: and so we're, so we're

Bill: Running in the browser.

Okay.

Thank you.

Thank you.

so you're probably thinking about how this

will, um, run in the cockpit and so on.

And so I'm wondering how important

the, the differences that you're seeing

between, you know, the A chips in, in

a, in an iPhone versus now, you know,

most of the iPads are getting M chips.

How important is that to you to be

able to do this edge processing?

as you go forward, is that going

to be a big deal for you like

you're you'll be able to fork

your or or change your application

based on the power that it has.

If it's running on an iPad that has

an M four in it, it's going to be

able to do different things than

if it's running on a phone with

an alpha chip with an H up, right?

Alex: Right.

No, I think that's, I think

it's massively, I think

it's massively important.

In fact, this is only the first year

where you might have a mobile phone,

which is powerful enough in order to

be able to run an AI model on it at

Bill: Right.