EPISODE OVERVIEW

Duration: Approximately 28 minutes

Best For: Trapped entrepreneurs who keep hearing about AI and feel left behind, overwhelmed, or sceptical about whether it actually applies to their business

Key Outcome: A clear mental framework for deciding what to automate, what to keep human, and how to stop wasting time on the wrong tasks

He spent 15 years teaching machines to understand the world. Then he realised most business owners still can't understand AI.

THE BOTTOM LINE

You built your business to create freedom. That said, you're now answering emails at 5am, missing family dinners, and wondering if AI is just another thing that's going to add complexity to your already overwhelming life. Sairam Sundaresan has worked at Intel Labs, Qualcomm, and now leads engineering teams at Valeo teaching cars to drive themselves. He's also the author of AI for the Rest of Us and reaches 100,000 followers who trust him to cut through the noise. In this conversation, he reveals something that will change how you think about AI. It's not about learning 500 tools. It's about understanding which of your tasks are worth $10, which are worth $100, and which are worth $1,000, then letting AI handle the first two so you can finally focus on the work that actually moves your business forward. Because the trapped entrepreneur who figures this out doesn't just save time. They get their thinking back. They get their judgment back. They get their life back.

WHY THIS EPISODE MATTERS TO YOU

You'll discover why less than 2% of the world is actually using AI properly, which means you're not late, you're early, and there's still time to get this right before your competitors do

You'll learn the $10, $100, $1,000 task framework that instantly clarifies what to automate and what requires your irreplaceable human judgment

You'll understand why the business owners who go all in on AI risk feeling like robots, while the ones who stay strategic will dominate their markets

You'll stop feeling guilty about not keeping up with every new tool, because Sairam reveals that even people on the bleeding edge feel overwhelmed, and that's actually fine

KEY INSIGHTS YOU CAN IMPLEMENT TODAY

The algorithms have been around since the 1940s. What changed is data and infrastructure. This means AI isn't magic, it's pattern recognition at scale. When you understand this, you stop being intimidated and start seeing it as a tool you can actually control.

Sort your tasks into three buckets. $10 tasks, $100 tasks, $1,000 tasks. AI should completely automate the first two. It should be a sparring partner for the third. If you're spending hours making AI do clever things on $10 tasks, you're losing money.

Human judgment matters more now than ever before. These models learn biases from data. They can't bring context the way you can. The trapped entrepreneur who understands this knows exactly when to hand off to AI and when to step in personally.

Writing by hand is now a superpower. When you outsource your thinking to AI, you get a dopamine hit without earning the knowledge. It disappears in 10 minutes. The business owners who write their thoughts first, then use AI to synthesise and challenge them, will outthink everyone else.

The principles change slower than the tools. Stop chasing every new model. Learn how to communicate your intent clearly to these systems. That skill transfers across every update, every platform, every new release.

GOLDEN QUOTES WORTH REMEMBERING

"AI should be helping you completely automate the $10 and the $100 tasks. It should help you brainstorm and be a good sparring partner for the $1,000 tasks." - Sairam Sundaresan

"We are writing the user manual as we go. So it's a fantastic time to get on." - Sairam Sundaresan

"People are outsourcing the judgment and thinking and that's a terrible thing that's happening." - Sairam Sundaresan

"The human touch is more important now than it was ever before." - Sairam Sundaresan

"If you're able to take this stuff out of your mind and put it somewhere else, your thinking becomes much wider and much more advanced." - Roy Castleman

QUICK NAVIGATION FOR BUSY LEADERS

00:00 - Introduction: How a childhood fascination with cameras led to teaching machines to see

03:45 - The Real History of AI: Why the algorithms have existed since the 1940s and what finally made them work

07:20 - The 2% Stat: Why almost nobody is actually using AI properly, and why you're not late

10:15 - The Overwhelm Problem: Even bleeding edge engineers feel this, and here's how to handle it

14:30 - The $10, $100, $1,000 Framework: The mental model that clarifies everything about what to automate

18:45 - Human vs AI: When to hand off and when your judgment is irreplaceable

22:10 - The Sameness Problem: Why AI is flooding the internet with slop and how to stand out

25:00 - Writing as Superpower: Why going analog for your thinking gives you an unfair advantage

27:30 - Conclusion: Where to find Sairam's book and newsletter for ongoing AI clarity

GUEST SPOTLIGHT

Name: Sairam Sundaresan

Bio: Sairam Sundaresan is an Engineering Leader at Valeo, where he manages teams teaching automobiles to understand the world. An alumnus of Intel Labs and Qualcomm, his work in deep learning has led to numerous patents and consumer applications. He's the author of AI for the Rest of Us published by Bloomsbury and creator of the Gradient Ascent newsletter, reaching 100,000 LinkedIn followers and 25,000 weekly readers. He's one of the industry's most trusted voices for making AI accessible to people without technical backgrounds.

Connect with Sairam:

Website: https://newsletter.artofsaience.com

LinkedIn: https://www.linkedin.com/in/sairam-sundaresan/

YouTube: https://www.youtube.com/@artofsaience

Instagram: https://www.instagram.com/artofsaience/

Book on Amazon: https://www.amazon.com/AI-Rest-Us-Illustrated-Introduction/dp/B0F29THNLT/

YOUR NEXT ACTIONS

This Week: Open ChatGPT or Claude on your phone. Take a photo of something in your business, your schedule, your inbox, your whiteboard, and ask it to help you sort tasks into $10, $100, and $1,000 categories. Notice what you've been wasting your expensive time on.

This Month: Identify three $10 tasks you do repeatedly and create simple AI prompts to handle them. Document what happens to your energy and focus when you're not doing low-value work.

This Quarter: Build a system where AI handles customer enquiries up to a certain threshold, then gracefully hands off to you or your team for the complex situations. Watch how your customer experience improves while your time opens up.

EPISODE RESOURCES

Book: AI for the Rest of Us by Sairam Sundaresan (Bloomsbury)

Newsletter: Gradient Ascent at https://newsletter.artofsaience.com

Tools mentioned: ChatGPT, Claude, Claude Code, Anthropic Dispatch, Notion, Obsidian, Supabase

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

READY TO ESCAPE THE TRAP?

Take the Freedom Score Quiz: https://scoreapp.atpbos.com/

Discover how trapped you are in your business and get your personalised roadmap to freedom in under 5 minutes.

Book a Free Strategy Session: https://www.atpbos.com/contact

Let's discuss how to build a business that works WITHOUT you.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

CONNECT WITH YOUR HOST, ROY CASTLEMAN

Roy is the founder of All The Power Limited and creator of Elevate360, a business coaching system for entrepreneurs ready to scale without burnout. As a certified Wim Hof Method Instructor and the UK's first certified BOS UP coach, Roy combines AI automation, wellness practices, and business operating systems to help trapped entrepreneurs reclaim their freedom.

Website: www.atpbos.com

LinkedIn: https://www.linkedin.com/in/roycastleman/

YouTube: https://www.youtube.com/@allthepowerltd

1

Good morning, good afternoon, good evening, wherever you are in

2

the world. I'm here today with might get this wrong.

3

Siram, that correct? Yeah. And Siram. Yeah. So

4

Siram has been in AI for a lot longer than

5

I thought AI was around. And we're gonna have a

6

chat today about AI and business and what it's doing

7

at the moment, what people are thinking about it. I

8

believe you've written a book as well. Yes, that's right.

9

And you teach. So interested to hear a little bit

10

more about your story. But first, how did you start

11

in this reopening into the world of AI? It started

12

actually many years ago when my parents took me for

13

vacation. Back in that time we had film cameras, so

14

you had to be very careful with the pictures that

15

you took because there was a finite amount of film.

16

It fascinated me that you could press a button and

17

capture a moment forever. There was another process of bringing

18

that moment to life in the lab. That was my

19

first foray with images and image formation. Later on I

20

was involved in that. I figured that you could teach

21

computers to understand images. That was mind blowing to me.

22

I decided that's what I wanted to do for the

23

rest of my life. I wanted to teach computers to

24

understand the world like we do. That's how I got

25

into AI. Since then, there's been one shift after another

26

and it's been an amazing journey. You talk about the

27

Fujifilm roles, you take the picture and then you put

28

it in the camera. So you send it to the

29

man, he delivers it, and then he gives you these

30

pictures and you're like, oh, okay, that was three years

31

ago. Yeah. So you just walk us through a little

32

bit of your history and then we'll dive into business.

33

Yeah, sure. I pursued my masters in a field called

34

computer vision, which is the science of teaching machines to

35

see and perceive the world like we do. So your

36

self driving cars, or your robots that are grasping things,

37

or even your phone that's able to recognize what kind

38

of plant you're taking a picture of. All of that

39

is computer vision. And I pursued a master's degree in

40

that. And that brought me into the subfield region of

41

AI. AI is a very blanket term. After my master's,

42

I got a job at Qualcomm and I began my

43

career in machine learning, which is another subfield of AI

44

and computer vision. And over the course of the years,

45

I've lived through this modern AI boom, where everything is

46

now solvable with AI, apparently, and agents are taking over

47

the world and what have you and along the way,

48

there's been a common thread, which is that there's always

49

a set of people who feel left out of the

50

conversation. That happens because they either don't understand the jargon

51

that's typical of the field, or they feel like they

52

need to have a really strong software background or mathematics

53

background. That was something that bothered me as well when

54

I was getting in. That's how I began to teach

55

and explain things. Eventually, some of the folks through my

56

newsletter said, why don't you put this into a book?

57

That was the genesis of my book. Most people watching

58

this won't even have an idea that AI was around

59

15 years ago. AI has been around since the 1940s.

60

The reason that it didn't take off back then is

61

that there weren't powerful enough computers. There wasn't a sufficient

62

volume of data. Once the Internet came around, we had

63

this age of big data, which means that you had

64

volumes of data on a variety of topics, and then

65

you had machines that could do calculations at a rapid

66

rate and do it in a scale that was previously

67

impossible. And now you had the infrastructure, you had the

68

data, and the algorithms have been around for decades. AI

69

has been around since the 40s. Explain that, because that

70

just doesn't make sense to me. I'm an IT guy.

71

I'm not a historian, so I might get the facts

72

wrong, but around the 1940s, there were a couple of

73

scientists who discovered something called the Perceptron, which is a

74

model of how our brain works in a very loose

75

fashion. So what they figured out was that if you

76

connect these entities together and feed them examples,

77

they can figure out how to do things like solve

78

logic puzzles or detect or recognize a man or a

79

woman. So this has been around for a while, but

80

of course, the infrastructure. There's actually a video somewhere on

81

the interwebs where you can actually see this machine. The

82

interest was immediate, and everyone was excited about, okay, let's

83

go all in on AI. But then another group of

84

scientists disproved that the Perceptron could actually work on everything.

85

And so the interest waned. And then we had these

86

spells of AI summers and AI winters, where there'd be

87

glimpses of progress and then something that would dampen the

88

funding and the spirits eventually. A lot of the formative

89

algorithms that AI scientists use today, like back propagation

90

or neural networks, large language models, LLMs belong to this

91

class of algorithms called neural networks, which are an approximation

92

of how the human brain works. In very loose terms,

93

this. These have been around many decades. But we never

94

had the volume of data. Now imagine having access to

95

all of Wikipedia, all of the articles on the Internet,

96

millions of images and video, and then being able to

97

use these as examples to teach these algorithms to learn

98

rules. And you need a lot of infrastructure. Like the

99

data centers that they're building today were around in the

100

2010s for gaming. And rendering was of the core calculations

101

used in these algorithms is something we studied in high

102

school. Matrix multiplication and GPUs are fantastic at that. So

103

now you had all the ingredients coming together. You had

104

very mature algorithms, a large volume of data and the

105

machines that could put these two things together. And that

106

was the big bang of AI as we know it.

107

Amazing. I want to dig into how

108

people. There's so many people out there at the moment.

109

I saw some graphics where you look at the population

110

of the world and the number of people that are

111

actually using large language models at the moment. I think

112

it's something ridiculous. Less than 2%. It's amazing how it

113

feels like the entire world is using AI and yet

114

that's such a sobering stat. I think it goes to

115

show that a lot of people may not have caught

116

on to the amazing things that you can do as

117

a result of these models. At the same time, it

118

also is interesting that maybe this has been restricted to

119

a few areas in the world versus the vast population.

120

I guess it's a mix of both. And I think

121

Even of that 1.6 or 2%, only a very small

122

percentage of them have a paid version. Yeah. And even

123

a smaller percent of them are using it with any

124

force. My point is that you're not late. You're so

125

early at the moment. If you haven't even looked at

126

it yet. It's so early in the AI world. We

127

are writing the user manual as we go. So it's

128

a fantastic time to get on. Whichever is your tool

129

of choice or if you're just wondering where to start,

130

it's always useful to pick something and try it out,

131

play with it, spend time and see what it can

132

do. See where you want to dig in a little

133

bit more. And that's how you get in. There's no

134

curriculum that says do this first. It used to exist

135

for people trying to enter the field from a technical

136

perspective, but ever since ChatGPT came out, it was sort

137

of Pandora's box where everybody now has access to this

138

incredible class of models and you can do so many

139

things with it. You almost have to unlearn things that

140

you have learned like we used to always believe that

141

the model can't do this or the model can't do

142

that. And oftentimes I myself am very surprised by what

143

I have to reconfigure my biases because I have a

144

lot of things I think the model can't do and

145

I'm surprised when it does. So it's a great time

146

to get into the field for sure. And yeah, now

147

we have, we've spoken about this low level of people

148

that understand it and the capabilities are just doing this.

149

Yeah, week on week there's a new thing coming out

150

and that I think is the other piece of overwhelm

151

that people have got because it's okay, I'm going to

152

learn something that's going to change. I think there's a

153

different perspective to that. More that that's a

154

journey of mastery. You have to go through those four

155

stages of the journey of mastery. You have to start

156

as the incompetence unknown or whatever the first one is

157

and then go through and that only. Yeah, you don't

158

need to know anything, but you have to play with

159

it and do it. People that I've speak to that

160

haven't used it before, I said get your phone, put

161

ChatGPT on it, use the free version, go to your

162

fridge, take a picture and say, give me a Michelin

163

Star recipe that you can talk me through. Yeah. Or

164

when you go on holiday, go where you are and

165

ask where you should go and see, these are the

166

things I like. And it's this human usefulness that really

167

is going to come to the surface. Right? Absolutely. And

168

I think. Were you referring to unconscious incompetence, conscious incompetence

169

and the four items and the level of mastery is

170

like unconscious competence or something basically. That's very true. You

171

need to start using it. And if it's any consolation,

172

people working on the bleeding edge of this also feel

173

overwhelmed because everything is changing every day and it's very

174

hard to keep up with everything. Oftentimes you might be

175

working on something where the next day your neighbor next

176

door finishes it and goes on to the next. So

177

it's moving at such a rapid pace and it's fine.

178

It's totally fine. I think two points I want to

179

surface. The first one is just step back and look

180

at what is possible now versus what was possible possible

181

three years ago or even two years ago. It's mind

182

boggling to see the capabilities that have been unlocked. And

183

it's important to take that in because you feel like

184

there's so many things that I couldn't do previously that

185

I can do now. On the bare minimum, I'm already

186

ahead of where I was. That's the first bit. The

187

second bit is the principles change much slower than the

188

tools you need to figure out a grounding principle. Here

189

is for the common man. You need to know how

190

to converse with these models in a way that gets

191

your intent to them as quickly as possible. Because the

192

more back and forth you have, the more these models

193

begin to bloat and then they forget what was said

194

and then they start muddling and your results start falling

195

off a cliff. So those principles don't change. Those are

196

constant as far as LLMs are concerned. When you operate

197

in that kind of a mentality, it's always easier because

198

every time a new model comes out, you might have

199

to tweak your prompting a little bit so that you

200

get the results you used to previous. But it also

201

means that this model can do a lot more than

202

the previous one could. That's a good thing. You may

203

have to prompt less, give it less instructions. It may

204

grasp what you're saying a lot easier. There's two sides

205

of the coin. I would take it that way versus

206

feeling. Oh, here we go again. I have to now

207

catch up with this new skill. I feel like as

208

these models keep advancing, they're going to become easier and

209

easier for us to get things done with. Focus on

210

how best you can achieve your outcomes and revisit your

211

priors every time the model updates. We talk to business

212

owners and I like to take it back down to

213

what's the actual problem. There's two things I'll bring up.

214

The first thing, if you find out what the problem

215

is and you just fix that problem, then you can

216

move on to the next thing. The second thing is

217

the example. You decide you're going to do some social

218

media posting on AI. You realize how easy it is

219

to do the LinkedIn post. So suddenly now, in your

220

wisdom and in your mindset as an entrepreneur, 15 different

221

things you can do. So you try and do all

222

of them. So the next thing, you're doing the same

223

amount of work without the outcome. You're overwhelmed because you're

224

learning all these things and you're losing focus on things.

225

Yeah. So that's the one side of it. I think

226

there's a real risk of understand the problem and just

227

fix the problem and don't get draw. I learned probably

228

500 tools in the last three years before I realized

229

it's not about the Tools, Right. I see that. I'm

230

already silent about that. I'm like, I didn't do that.

231

I can do this and I can have all the

232

different things that are on there and then it's okay.

233

And the other thing is, as you go through this,

234

I think it's important to bring this in. If you're

235

able to, from a business owner's point of view, you

236

understand what the 10x is. What's the one thing that

237

you do, you think in the next 12 months or

238

24 months that's going to make tech companies stand out?

239

I feel like there's two parts to this question. Let

240

me first answer it from the entrepreneur's perspective and then

241

we'll answer it from the company's perspective. You're absolutely right.

242

It can be very addictive to try out all the

243

capabilities of a model as soon as it comes out.

244

I can do this for my business, I can do

245

that for me. But then it's distraction, right? So your

246

context is now split and you're spreading yourself very thin.

247

A very good entrepreneur friend of mine has this mindset

248

that I always think about, which is you can put

249

your tasks into three buckets, the $10 tasks, the $100

250

task, and the thousand dollar tasks. And AI should be

251

helping you completely automate the $10 and the $100 task.

252

It should help you brainstorm and be a good sparring

253

partner for the thousand dollar task. Anything else? If

254

you're spending a lot of time trying to do cool

255

stuff on the $10 task, you're already at a loss

256

because you're wasting money. And that is a very nice

257

mindset for an entrepreneur to have because then you're focusing

258

on your business now, coming to the companies and product

259

side of things. There's this notion of AI powered, which

260

seems to be a buzzword right now. And I feel

261

like you should think about whether a product is existing

262

in the world because of AI or is enhanced. A

263

good example is an email client. You don't need AI

264

for an email client. However, it enhances the experience. Auto

265

completing your email, summarizing your emails. But at heart, an

266

email client is supposed to send and receive messages for

267

you. On the other hand, if you remove the language

268

model that's powering these things, you'd have a terrible experience.

269

And AI is at the core of those things. And

270

so the real 10x value comes from two things. One,

271

is AI central to your product or service? And if

272

it's not, are you wasting your time? Because you should

273

not be attaching AI just because everybody else is doing

274

it, there should be a clear purpose and a reason

275

for it. The trickier bit, if the models get better,

276

and they will, does it enhance your product or service

277

or replace it? And if the answer is the latter,

278

you really need to think about how you want to

279

approach your business and product or service. So the ones

280

that will 10x are the ones that have AI at

281

the core and get enhanced every time a model update

282

happens. I believe that you can do

283

10 times more work with the same number of people.

284

If you're doing 10 times more work with the same

285

number of people and you've optimized things, what does that

286

work look like? I believe that you should be focusing

287

on how do you bring more human into your business?

288

So there's the 70% that you can automate, and you

289

can automate it step by step. Sops are right, you

290

know what you're doing. You, you do that. The customer

291

conversations, the salespeople, the interactions, customer service line should stay

292

more human because I think the companies that are more

293

human are going to be the companies that really stand

294

out. Because the companies that just go all in on

295

AI, they risk feeling like they're AI. So I feel

296

like there's a nuance here and you're touching on something

297

very important. And yes, I feel that the human touch

298

is more important now than it was ever before. And

299

not just because it feels like you're interacting with people

300

and that's always a better experience than interacting with a

301

machine. But the more important reason is that judgment, taste,

302

and just being able to make the decisions from a

303

contextual perspective and not a trained biased perspective. What I

304

mean by biased is in the context of AI models,

305

they're learning rules and patterns from data. And if your

306

data has biases baked into it, the model is going

307

to learn those. So having a human who also has

308

biases, but there's a higher likelihood that you can bring

309

in some context and smooth out the rough edges. So

310

from that perspective, yes, absolutely, human is important. If you're

311

operating at a lot of scale and a lot of

312

the queries that come in for customer support are trivial,

313

frequently asked questions that can be easily answered instantly without

314

them having to wait for hours on the phone, that's

315

a good place to inject AI. And then where you

316

introduce the human becomes a niche case none of these

317

models are able to address. And then they gracefully hand

318

it off to a human. And then when you speak

319

to a human, then the entire customer journey becomes so

320

much more pleasing. Because if a customer is contacting customer

321

support, it means they need help with something, they're upset

322

about something or they want to when something is coming,

323

maybe they're excited. And as business owners, your job is

324

to make that experience as frictionless as possible for them.

325

So there is a layer where AI makes sense and

326

there is a layer where that handoff is necessary. So

327

the answer isn't binary in my opinion. It's a gradient.

328

100% agree. I think it's which are those tasks that

329

just are repetitive and it's just look something up, give

330

somebody some information. 100% hand that off to AI. But

331

understand when someone starts getting irritated now you need to

332

hand it off to human. I've seen the onset of

333

all of these call answering services and call calling out

334

services. A lot of people, I'll put a few in

335

for people and you watch the people that drop off.

336

If it's not 100% perfect, you do get a drop

337

off. So it's like, okay, is this affecting my business

338

in a positive or negative way? And just go from

339

there and look, that's going to change, right? I'm working

340

with somebody at the moment, interesting guy who's built his

341

entire AI platform around human emotion so the AI can

342

understand human emotion. That makes that window smaller. But

343

then on the other side of that, you're also seeing

344

people doing posts and things now actually putting in the

345

errors, making it less flowery because now there's all this

346

perfect content out on social and that was just AI.

347

So in a sense now imperfections are now something that

348

we cherish a lot more than stale perfection. So where

349

do you see it going? I'm playing with Claude code

350

and I'm doing all sorts of other things and just

351

trying to keep my focus back in. Where do you

352

see it going in the next, even the next six

353

months to a year? It's very hard to say because

354

we're seeing things, as you said, that are changing almost

355

every day. What was previously not possible a week back

356

is suddenly possible. Just take, I'll give you an example,

357

right Anthropic just released something called dispatch and that allows

358

you to run cloud code and cowork on your computer

359

from the comfort of your phone. So if you're from

360

outside, you can actually just send. It's just insane convenience

361

and done at a level of execution and speed, very

362

hard to predict. But there are some trends that I'm

363

seeing. Some good, some not good. Let's talk about the

364

good. The good is that people beginning to build things

365

for themselves. Because now programming is resorting to language, whether

366

that's English, but whatever be your language. You can now

367

speak to a computer and have it build things for

368

you, whether that's a application, whether that's a dashboard, what

369

have you. A lot of these apps that do monitoring

370

or productivity can be built in house and so you'll

371

start seeing subscriptions there. Like you, the subscriptions will have

372

to earn their worth. That's the first bit. You'll also

373

start seeing one person companies become more common where one

374

person is just sitting at plot code and then hammering

375

away and building a product. And not everybody will succeed

376

at this because you still need to have very good

377

software engineering principles because of security, privacy, all that kind

378

of thing that really matters. You're seeing a huge uptick

379

in the number of people building things independently and then

380

launching them out in the world. This will accelerate even

381

further as these models get better. Because I'm blown away

382

by what you can. One shot. But on the not

383

so good side, you're seeing the sameness of social media

384

posts, of videos, of designs, of websites. Everything is becoming

385

same. If you've been working with these tools, you will

386

be able to spot things like instantly, ah, this is

387

written by AI or this was designed by AI or

388

the same art pitch all the time or it's not

389

this, it's. That's like the most common tell. And then

390

EM dashes. I feel really bad because I used to

391

love em dashes before chat gpc. But the point is

392

that's the gap that's becoming a problem because now the

393

Internet is being flood flooded with slop. So then that

394

becomes more of a thing going forward. And it's going

395

to accelerate in the next six months because people don't

396

have time to write their own posts, they're just going

397

to outsource it to a model and then spam all

398

social media platforms so that they're top of the mind

399

posting every day because the algorithms reward you for being

400

present on the platform every day and engaging and so

401

on. Negative spiral in my opinion. And the second part

402

is people are outsourcing the judgment and thinking and that's

403

a terrible thing that's happening. I don't mean just entrepreneurs,

404

students, anybody. When you had to turn in an assignment

405

in school, you would have to think, read a bunch,

406

do a bit of research. Yes, the Internet was around,

407

but you didn't have a 24. 7 tutor available on

408

the phone to give you the answers, literally. And I

409

feel like all of a sudden people have Stopped thinking.

410

Okay, Let me ask ChatGPT, let me ask Claude. And

411

then they get the dopamine hit of knowing the answer

412

but not having earned the knowledge. And so it goes

413

away within 10 minutes. But nobody cares that lack of

414

critical thinking, lack of judgment and blindly relying on these

415

tools is becoming a huge problem from the grassroots. So

416

I'll stop here for questions, but these are the trends

417

I'm seeing. Yeah, I've finished a book that's been drafted,

418

which is how to think outside your brain. The key

419

concept is that we don't need to go and learn

420

these skills anymore, but you need to stay the thought

421

leader. You have to stay every project you do. You

422

have to stay the thought leader. You can't just. What's

423

my problem? Let me ask the questions now. If it

424

takes you. How many times have Chat PT taken me

425

down a totally wrong path? I love Court is much

426

better at that. But chatgpt, take me down this random

427

path and yeah, it's great to be able to learn

428

with it, and it really is, but actually learn with

429

it and actually understand what can you do inside your

430

brain? I've found for myself that I'm able to think

431

that much more, and I think that's an entrepreneurial thing,

432

is that your thinking becomes much wider and much more

433

advanced when you're able to take this stuff out of

434

your mind and put it somewhere else. I use Notion

435

or Supabase or one of them and I put my

436

ideas in and I said, just store them over there.

437

Then I don't have to think about them anymore. I

438

can come back to it. The way we think with

439

AI is different from the way we thought before. That's

440

going to come to the surface over the next year.

441

Writing is now more important than ever before because I

442

actually mean writing things by hand. It's a great time

443

to go analog for certain things. If you want to

444

sharpen your thinking, writing is the best way to do

445

it. If you rely on the computer to do it

446

for you, it's a problem because you fall into that

447

same trap if you have thoughts. If you're thinking about

448

an idea, just putting your thoughts on paper and then

449

maybe using a voice transcription to put it into obsidian

450

notion, what have you. Your favorite tool. That is a

451

very powerful thing to do for two reasons. First, you

452

have done the thinking and then the synthesis can be

453

done by these models. But the second bit is that

454

you can then ask these models to poke holes in

455

your thought process. You can ask, okay, what are the

456

rabbit holes? You can Go down. What are the contradictions

457

to what you have thought? What are the angles that

458

you can perceive this from? And that then opens up

459

a new world where you can expand your knowledge. So

460

I feel like writing is a superpower that everybody has

461

to start tapping into. Okay, I'm going to start bringing

462

it to a close, but I want to just hear

463

a bit more about you and what you're doing. So,

464

yeah, obviously you've written a book. What are you doing

465

day to day? You teach as well. Tell us a

466

bit more about what you do. Sure. Yes. I've written

467

a book. It's published at Bloomsbury. It's called AI for

468

the Rest of Us. It's an illustrated book where I

469

promise you, you can learn AI without having to get

470

a PhD. And I teach through my newsletter primarily. It's

471

called Gradient Ascent. And what I try to do is

472

cover the latest research and developments happening in the AI

473

field and then translate it into what actually, what are

474

the implications for people in the tech industry and around

475

the tech industry. It's read by a wide variety of

476

people, including VCs, managers, directors, engineers, students. That's

477

my favorite way because I get to learn stuff and

478

I get to teach it in my day to day.

479

I am an engineering manager and I lead a large

480

team of engineers building autonomous driving and parking software. That's

481

a really fun place to be in because you get

482

to see where reality meets imagination and you get to

483

see where things work and where things fail and lots

484

of lessons to be learned there. Thank you very much

485

for joining us. I'll put your contact details on the

486

show notes and wherever we post this and hopefully we

487

can chat again soon in six months and maybe our

488

AIs will talk to each other. That'll be fun. Thank

489

you so much for having me. Thanks, man.