AI has been doing a huge chunk of our work for us, however it can't seem to get that last 10% right. Rich and Paul discuss how technology lacks the human traits of real-time judgement and unpredictability. So, it will have to negotiate with humanity. This episode is sponsored by Aboard (the new tech in town).

All right, rich, you're hitting the table.

You're so excited to tell me something.

What do you wanna tell me?

Tell me.

Tell me.

Tell me.

All right, so here's the deal, okay?

Okay, let's go back to 2015.

Boy.

Okay.

Pandemic's a thousand years away.

Everyone's thinking about self-driving cars and, uh, cars

are gonna drive themselves, right?

That's one.

When you say self-driving cars, I drew that conclusion.

we're gonna have self-driving cars and we're gonna get to do things like

watch a TV show while we get to the office

the car's

Yeah, that was, oh,

and here's the crazy thing that happened.

Do you think driving a car requires a massive amount of intelligence?

No.

Do you think coding a React application requires more

intelligence than driving a car?

Yes.

Yes.

Do you think.

Writing a business plan requires more knowledge and

intelligence than driving a car.

Yes.

So then answer me this.

We've stalled on driving cars mainly because of Chinatown and New York City

and our inability to see people whizzing between parked cars and smashing into us.

wait for people who don't understand because

let's just put it out there.

That could have sounded pretty racist if you don't do New York City, that

has nothing to do with it being

No, it has to do with

It

has to do with the Manhattan Bridge ending in Chinatown and just absolute

a large bridge ending in a small village is essentially what's

that, that's it is.

Chinatown is absolutely, it's, I used and I bike through the It's a disaster.

Yes.

But

what happened?

We were gonna have self-driving cars, something pretty straightforward, signal

left, turn, left park, the car that's stalled, and now we've got AI all over

the place telling me I don't need.

Health insurance.

I don't need to doctor anymore cause I could just ask the AI robot

suspicious because you're right.

It couldn't, we, we, we don't have self-driving cars, but it'll draw you

a picture of like a cat in a shoe.

I, I, I put in yesterday, um, Lebanese Civil

War as a Renaissance painting.

Hmm.

How to do is this mid journey.

It was mid journey using Discord, which is re preposterous, but

whatever.

In general, it's hard.

Like we, we has software gotten better in the last eight years.

Like what?

What, how are we using discord to interact with ai

If we can do these incredibly complex why?

What?

I don't think we're gonna get it.

We were, they were predicting, I think 10 years, like 20, 25.

If you told me it's not until like 2040 if ever.

I wouldn't be surprised.

what happened?

Well, this is the thing that always happens with AI and has been

happening with AI for 50, 60 years.

It's the right around the corner syndrome.

It it's this complicated set of interactions where like the people

who are true believers are like, you better get ready because the world's

gonna end and the people who aren't true believers are like o O, okay.

And then it's always pushed out just a little bit.

We're seeing it with this new wave, right?

Which is.

It's gonna write all our songs for us and write our documents and so on.

What it is gonna replace is the world's worst marketing

messages that you sent to spam.

Anyway, it's gonna do that.

is.

Let me throw out a hypothesis and you tell me if I'm right or wrong.

Okay.

Um,

the, the human brain is vastly complex in its ability to, um, synthesize a decision

in real time with like almost instantly.

Almost instantly.

And what I mean by that is it's that last mile of logic that I feel like

a lot of AI and self-driving cars, it's the one thing they have in common

is they kind of quickly sprinted the first like 90% of the race.

And then there's that last 10% where you pretty much have to decide how

you're gonna navigate Chinatown.

Are you gonna have to decide?

And, and, and that's why you're seeing like a lot of the, the chat G P T

stuff looks incredibly impressive, but if you really look carefully,

that last 10% is, goes off the

Let me make this easy for you.

I can give you one word and then you'll be able to completely clarify your point.

Neurosurgeons.

Whoa,

Okay?

They go to school for how long?

years and then they watch surgeries for like another 10

Mm-hmm.

And it's that last, it's, it's for those hours that they're in the room, right?

there's

computers everywhere.

There's robots everywhere.

They're assisted in a million different

There

actually are.

Yeah, it's incredible actually.

The tools they use now, um, don't allow them to screw up from a,

like motor skills perspective.

It's kind of wild.

That's right.

So they can't make a mistake because the robot will keep

them from making a mistake.

Correct.

Okay.

And uh, but could we get rid of the human in the loop?

Um, really, really hard because you need that realtime judgment.

You can't, right?

You can't, you can't.

And it's, it's context and it's humanity and so on.

And so like is it

and experience

is it inconceivable than a robot?

Could an expert system of some kind could eventually perform better

neurosurgery than most humans?

No.

I, I, I, you know, I, I think like, I'm sure like it's, it's, it's

conceivable, but what I, I think happens, what drives me bananas is that the.

And I think of this as a West coast thing.

The West coast vision of humanity is extremely narrow, and that aligns

with creating products that millions of you, you know what's funny?

You work to get the humans out of the loop until the point they're about to.

Once they get their credit cards out, that's okay.

Then you want 'em back in the loop.

But up until that, no.

I'll tell you like remember you ever tried to call like Amazon for

customer service, early days of

aws?

like they didn't want humans in there.

Everything needs to be as automated as possible cause it's all of your

efficiency and all your margins.

The minute you add humans to the computers, you ruin everything.

And it's less goes.

That is a West coast ideology in the West coast, but the West Coast ideology, God

bless it, often doesn't include things like read, you know, they're always ready

to kill the humanities on the west coast.

We've talked about this on the podcast before.

They're always like novels.

Nobody reads novels.

Yeah.

And let me be frank, I haven't been reading many novels these days.

I'm kind of busy, but nonetheless, I've read a lot in the past.

And there is a,

a.

real thing that happens, and it happens as you get older and it

happens, especially if you grow up maybe without a ton of money.

And so, but you realize that humans are unbelievably perverse and slippery and

in infinitely creative in the weirdest possible ways, and that you can do

anything you want to try to capture that.

In fact, some of the greatest minds in the world have done spent

their entire lives as philosophers, novelists, whatever, trying to capture

every aspect of human behavior.

And they just can't get it.

It's not the last five years.

It's the last 5,000 we've been working on this problem.

Yeah.

And computers don't solve it.

Not in this, I mean any more than like guns solve it or

like, I don't know, hammer solve

Yeah.

And, and, and that is just the creative mind.

I think so.

I think that, so why is this so hard to make a self-driving car?

It's not cuz of the cars, it's cuz humans are utterly unpredictable and they make

incredibly bad decisions all the time.

So what you're saying is if we.

Built a zone around Chinatown and said, only self-driving cars are

allowed.

I'll fix all that for you.

Absolutely.

I'll, if you let me put a one inch metal strip down the middle of the street and

then block all the other traffic off, the self-driving cars will work beautifully.

So the issue is other humans.

always

interesting

you can't create a model of humans.

Except in the absolute most aggregate, even then it falls through.

Right?

What's the way to accelerate this?

The way to accelerate this is to work with local governments and create like a

self-driving car lane throughout the city.

Mm.

So you're negotiating with humanity

Now?

We're in the, now it's so, it's like, I'm like, ah, there's all this

epistemological, philosophical stuff.

All it comes down to ultimately here is like regulatory, right?

So go ahead and go to the Adams administration and the Hoku administration

and say, Hey, we want to completely change the way traffic works in New York City.

And they'll say, we're trying to do congestion pricing,

and it's not going well.

right.

Like

So you're saying, Technology isn't the way out.

We're gonna have to negotiate with the humans.

It always ends up that way, and that's why Google has a, why would

Google the most automated organization in the universe have an extremely large

lobbying arm in Washington dc Like there's buildings of Google lobbyists.

Yeah.

Yep, yep, yep, yep.

So what, what did they figure out eventually?

Because honestly, with AI and all these systems and self-driving,

like why would you need to regulate?

Why would you need lobbyists?

Yeah.

You know, no.

You.

You.

That's not how it works.

You're gonna get in these systems for all the fantasies that we have on either

side, that somehow we can shut the platforms down or that the platforms are

gonna take over the world or whatever.

Ultimately, you still have these giant regulatory frameworks in the middle,

and you can look at them different ways.

They could be, you could see 'em as people with their hands out.

You could see 'em as people who are protecting our best interests.

You can see them as people who are selling out.

The working class like we have, but that's the framework we're in.

So you drop self-driving cars in as a, as a hypothesis.

It works fine.

You drop it into an environment with human beings interacting as human

beings do in a relatively justice oriented, law oriented society.

Yeah, no way.

Slow down there, buddy.

Yeah.

Yeah.

And, and, and you mentioned Arizona.

I mean, there are places where solving the puzzle is a lot easier than, yeah.

yeah.

I mean, New York City, if we don't have self-driving cars,

we, we can't, we can't do

Yeah.

Yeah.

We can't get new subways.

We, it's hard to get your trash picked

Right, right, right, right.

um, yes.

Arizona can have self-driving cars cuz the, it's like the, each street

is, you know, you know where you can You ever been to Minneapolis?

It's like 45 minutes to cross the street.

They're so wide.

Yeah.

Well that's a sprawl situation of a, of the first order.

Let's do it there.

Go find one of those like nicely constructed cities with

like a 1.2 million population.

Yeah.

I, I think, I think, I think you're kind of nailing it here.

I also think, you know, self-driving cars are capable of killing people.

AI isn't yet, maybe down the road,

God forbid we just have buses too.

Sorry, that's not

I worry about this with ai, is that we're really

good at sort of creating this wild machine that then gets out of hand

and then we gotta reign it back.

you don't mean like wacky robots with laser

No, I mean social media and it's like, oh my God, it's messing.

You know, messy kids are, are mess

I am

it and misinformation.

And now crypto and you had a lot of people losing crypto's, kind of making

a, like the, the, the shares are back up, but they'll, they'll be another day.

And even so, even though they're back up, there are a lot of people

who like got wiped out, like people who put their savings in it and,

and.

what happened with crypto too is the media's just done.

Like, it can't take, nobody wants any more press releases.

Nobody's gonna cover it.

So like,

And now we're on ai and then we're gonna be like, uh oh.

Um, that drone that was supposed to deliver your Taco Bell smashed

into like, you know, a home.

It's that.

And it's not just that, just like look.

Do I even have to tell you it's the 2024 election will be the AI

bot election, and we'll have to like, that story's coming and, and

just it's, I'm already, I'm, yeah.

Yeah.

I'm pre exhausted.

fake articles that are generated by ai.

By

ai,

yeah.

Can you imagine?

Like, I mean, who's gonna, it's Trump or DeSantis and then their,

their team will start using bots and

oh, God,

we're just, it's gonna suck.

And then Facebook will be like, well, this is our new policy.

Everything is, look, I think what is different is that even at the

giant platform level, everybody leans into a regulatory framework.

Even if they have to improvise

sometimes later than they need to.

But even, I don't think Facebook knows, like, I kind of doubt that

2024 will be a replay of, of like 2016.

I think that's

right.

And it could be other bad actors that are leveraging tech, but

Yeah.

I just like, yeah.

But where else?

There's only like a few platforms.

Where are they gonna go?

Like, here we are.

So I, I like that, that part, it's hard to get

Because it feels like the excitement is so manufactured, and if you look at

it, it, it kind of all ends up rolling back to a few incredibly wealthy

individuals who sort of keep raising the stakes and raising the stakes and, and,

uh, or very, very wealthy companies.

And so

My pleasure and excitement about technology is typically

an inverse relationship to that.

Cuz I like seeing people mess stuff up or frankly, like, this goes back

to the product you and I are building.

I like seeing people clean stuff up, like, you know, make sense

of the world for themselves.

Well, I mean, you're making an important distinction here

and I think this is a good sort of closer, which is distinguish technology

that tries to automate people away.

Versus technology that actually empowers people and amplifies our, ourselves

Yeah, but I'm gonna make one, I've been thinking about this a

lot cuz our, our, the product we're about to launch has a lot to, like

our particular culture, our culture of like internet nerds who believe the

computers are incredibly empowering and make smarter, better human beings.

Mm-hmm.

What do people actually do on the web all day?

shop for stuff?

for stuff, they watch videos, and often they're watching

videos about shopping for stuff.

If the web has taught me anything, it's that we are extremely economic

humans, like things that you think are incredibly meaningful and deep, like

your book club tie back to a product called the book or things like, I'm

gonna, I'm gonna buy a beautiful home and decorate it with my, my wife that I love.

You know, that involves a large transaction and a mortgage, right?

Yeah.

There's this infrastructure, the web is infrastructure around

those transactions just as much as

around life decisions,

distribution and document creation and so on and so forth.

And I, I feel like.

There's a product manager ethos of the greater human being that I'm just

increasingly suspicious of as well.

Right?

Like, I feel like if you're gonna help people, you want to help 'em.

And that's what we're trying to do.

Like, we're gonna like pat me on the back before our product launches and

everybody looks at us from blinks.

Um, but, but what we are consciously trying to do in this very room is

look at people where they are and say, how can I make that better for you?

Not just buying, also creating documents, also doing all that

Very east coast

very east coast, humans are gonna be in the loop.

Yeah.

Humans are gonna be in the loop.

Put lots of humans in the loop and, and hope

and, but not too many at once.

That's the other aspect.

Well that's not, that's a different podcast, but

that is a very different take from why are humans doing that?

We can make machines do it.

Right.

And

funny, the West coast ideology is like, don't let humans in the loop.

But if you do.

Let 'em all in.

Yeah, let,

yeah, yeah, yeah.

just assume they'll be fine.

Which is another

It'll work out the details

Yeah.

You kind of learn when you're walking around New York City that you don't

want to be in a room with everybody.

So what you're saying, Paul, is that the best self-driving

car is the one that uses a human

That's a, that's an Uber rich,

and a gas pedal

that's a

and a brake.

No one would love self-driving cars more than me.

No one is more ready.

Yes.

uh, I have one though.

It's called a bus.

It's not quite self-driving.

There's a human in the loop, but it's solving

well, but it's, it's one to 50 ratio, so it's pretty good.

pretty good.

It's pretty good.

my bus.

I love my bus driver.

this, this podcast has been sponsored by a board.com.

We talked about this, uh, throughout this podcast.

Um, it's a new tool that we're making available real soon.

Sign up to get in line.

We're gonna let wave a ton of people in very shortly.

Um, it's a, it's this really, really cool platform that helps

you organize your passions.

Can't say any better than that.

Check us out at, uh, hello.

It's yai ford.com at Ford on Twitter, but also add a board on Twitter.

Follow us.

We're gonna start tweeting soon.

Tweeting.

Um, and, uh, yeah, can't wait to show it to you.

Anyway, all Rich.

Let's, uh, let's get back to work

Have a good week.