In this episode Rich and Paul assess AI risk. They discuss them in the 2 chapters - that where the anxiety should be versus where it is overblown. This episode is sponsored by Aboard, a place where you can take a breath from your AI anxiety.

Mail bag.

I love good mail.

Bag.

Mail

Mail.

Mail.

bag.

Mail.

What is that, Paul?

Is that a Restoration Hardware catalog in your hand?

Yes

is.

Love.

Mail

Mail bag.

All right, so Richard, we got a, we got a letter.

You ready?

Subject.

AI risk.

Oh Lord.

It's from Peter Wilson.

Peter Wilson.

Okay.

sound.

Hello Peter.

Thank you for writing in and thank you for listening.

Pete, how you doing?

How you doing, Pete?

All right.

Hello, Mr.

Ziti Ford.

I appreciate the advisory service you've provided Thus, Far in regards

to the prolonged hysteria the world is experiencing as relates to the

latest incarnation of the term ai.

This statement and the signatories do give me pause as I suspect all the way

these folks talk about AI encompasses all the things you've been discussing on

your podcast from posts like days to now.

Could you perhaps provide in a future podcast your perspective on this

statement, quote, mitigating the risk of.

Extinction from AI should be a global priority alongside other societal scale

risks, such as pandemics and nuclear war.

And that comes from safe.ai/statement-one-ai-risk.

Thank you.

Yours sincerely.

So Rich,

All right.

are we extinct yet?

Can we end this?

Is it over?

Is the AI apocalypse upon us?

Let's get one thing out of the way.

Okay.

Okay.

Okay.

Um.

To use the term extinction,

Uh, is a hell of a thing.

I don't know who the Center for AI Safety is, AKA safe ai.

It's the regular suspects of, you know, the wild thing with

AI is everybody's like, man, ah, boy, you gotta regulate this.

And it's just, it's essentially like someone just constantly

just, just huffing paint.

And then turning to you with their face all covered in silver

going, we gotta regulate it, man.

Yeah, I think, I think, uh, I think standing up a, I don't know what

this is, I don't know if it's a for-profit consultancy or a non-profit foundation.

It's,

It's a, it's a non-profit foundation, but let's be clear, all

boundaries such of that are, are kind of meaningless at a certain scale.

you know what the, where the alarm bell goes off is extinction.

It's like, let's all just take a breath.

But I understand the anxiety and I, I wanna make, I want to essentially

break this response into two chapters.

Okay?

One chapter is, um, Where most of my anxiety would be.

And another is where I think it's overblown.

So the chapter where I think my anxiety would be is, here's what

we have learned around humans and innovation and advancement in

Okay.

Uh, we are really good at creating these things

that tend to branch out in a like exponential fractal-like manner.

And get away from us.

And then we all of a sudden wake up, we're like, whoa.

Oh boy.

I didn't expect that.

I thought we were just gonna share like tuna casserole pictures on social media.

I didn't expect an Egyptian revolution that was ill founded.

It

There was an amazing, um, onion article at the

beginning of Trump's campaign.

Yeah, like years and years ago.

And the article is just, it's a picture of Donald Trump and

the headline is, admit it.

You just wanna see how far this thing will go.

Exactly.

Exactly.

And so what you had was, you know, uh, was there an evil cabal that

was behind Facebook all of a sudden upending governments and spreading

misinformation from the get-go?

No.

It was a bunch of knuckleheads who were good at php.

Hey, I, I can make this real simple, like, we know this is a culture.

What is the road to hell paved with?

Like buttons,

good intentions.

Here's where I end up with this.

There is a nerd of fantasy of absolute power because the computer seems to be.

It's

a magical device and it, you see yourself reflected in it.

And then you see that these tools, you can hypothesize how if the brain

is just a computer, how a computer could become brain-like, and once

brain-like, it will simply absorb all world resources, take control.

And why would it need humans?

Yeah.

I, I,

but I think that, like that, I'm on year 25 of that fantasy and

I'm, I've seen communities literally come and go around that fantasy.

It, it fulfills an emotional and religious need that humans have an.

It's also when you go out to the, God bless our West Coast friends,

but they often don't read books that don't have numbers in the title.

Like four days something, and like, you know, 25 methods too, like.

So in a month

Yeah.

Everything in a month, right?

I'm at like, okay, is it an extinction level event?

And what I love with this is they always tend to neglect.

It's like, oh really?

Cuz you, you've, you kind of knocked all the other extinction level events out

of the park, like nuclear war asteroids.

We solved that climate change.

No, no.

This comes first.

you know that a seven, A 7 47 can land itself?

Yes.

I did know that because Can

fully land.

I don't think it could taxi to the gate.

That's a little much, but it'll land itself just perfectly.

it can't see people.

Yeah.

You can't, it can't actually get them on the

or give you like dry roasted peanuts at the right moment.

You

They don't know where to put your bag.

But it can land itself.

It can land itself.

Now here's what I think is happening.

So let,

let's, you know what I hate about air travel?

No, I'm sorry.

Just

don't do that.

Or

so here's what I think I, I think is being asked, like when you

say extinction, and the truth is we are extremely diligent around where humans

have to put their hands on the wheel.

Like to this day, even Aztec has gotten more and more advanced.

Um, there are very clear lines.

Um, around how that works.

I have the best analogy I can give you here.

Not an analogy, it's an example, a technology example is my power washer.

I'm going to take us on a little power washing journey

First of all, is it artificially intelligent?

It's not, no, but it has a mechanism that I think we're

gonna adhere to, is my guess.

Um, and that mechanism is this, the power washer.

If you are wearing, if you're barefoot or wearing open toed

shoes, we'll rip your toe off.

You can actually kill yourself with a power

It's extremely high pressure water shooting out to

clean your yard or whatever.

And the way my power washer works, and I think they all work this way,

unless you get like a professional

I'm gonna guess you bought a relatively good power washer

No, it was like 200 bucks.

I dunno.

Is that a lot for a power wash?

I have no

I really couldn't tell you

exactly.

No, there were, there are much more

I'm gonna guess 200 is like a nice middle of the road

nice middle look.

And here's the thing about the power washer.

You can't turn it on and wave it around.

You actually have to hold it down to.

Shoot the water out,

it turned like it won't shoot water

It's exhausting.

It's, you actually have to keep the trigger tight in your hand

to keep shooting water out.

Why?

Because if, if, if, if some, if my son yells, Hey dad, what are you doing?

And I turn and wave towards him, like nonchalantly, I can take out

a window and take him out too.

Yeah.

no, no.

This is, I, there's a switch on my, um, tree trimmer.

we're kind of nuts about safety

Yeah, yeah, that's

sophisticated circular saw mechanism that

if you put your finger on it

It like ruins the saw.

It has this mechanism that actually wrecks the saw in a split second.

It's actually pretty cool.

It's on.

So I guess what I'm trying to say is this, we're actually pretty paranoid about

machines getting away from us and we have mechanisms in place around it, and he let.

Another great example.

We've hit a wall with self-driving cars.

Well, that's very literal

I thought we would've turned the corner, but right now it

seems like only Arizona, where there is nobody except like 11 coyotes are

willing to really give it a proper go.

as of an hour ago, that's 10 coyotes because the self-driving

car test isn't going so good.

so I guess what I'm trying to say is a lot of these

policing mechanisms are in place.

We won't let the 7 47 land itself.

So many years ago, there was a book called Super Intelligence by a writer

named Nick Bostrom, which was about how the AIS could like, you know, let the, the

classic one is the paperclip maximizer.

You ever heard of the paperclip maximizer.

Uh, I didn't know we were gonna talk about Delta

Entertainment, but here we are.

No, I've never heard of that.

let's say you make a machine, you say like, your job is to

make as many paperclips as possible.

That's your job machine.

And then you get the AI module and you plug it in so that it makes

paperclips even smarter, And it goes, oh wait, that's my job.

I'm supposed to gather intelligence and make more paperclips, right?

My job is to turn every atom of existing things in the entire

universe into paperclips.

I'd better do that.

So it's like it gets intelligent and it sort of enslaves all of humanity because,

and puts in the work making paperclips and when they made enough paperclips,

it turns the people into paperclips and so on, until the entire world is just

one giant pile, or the entire universe, sorry, it's one giant pile of paperclips.

It's a great game about this.

Um,

Called, uh, I think Universal Paperclips.

Anyway, regardless.

Uh, so it was, it's, it's a book about that stuff.

And I had to call this AI guy to, to get like a source or just

to kind of riff on the review.

And, uh, you know, do other people believe this?

He's like, you just couldn't, it was just 10 years ago.

He's like, it's just coming tomorrow, right?

He was, it wasn't quite that, but it was just, he was just like, yeah,

you, you, you can't put blinders on here, you know, and that.

And, and, and so ever since then I've been kind of keeping my eyes open and

I just see pictures getting drawn in large language models and, and more

and more humans can get faked out.

But I don't see intelligence.

I don't, I just don't see actual intelligence emerging.

And everyone keeps telling me that, no, no, it's right around the corner.

But I it's been so long, I don't think more CPUs at it actually gets you there.

It doesn't.

And, and I think what's so, what's so interesting about it, when we say

the word intelligence, um, I mean there's knowledge which is sort of

the most basic form of intelligence, which is recall ability and just

sure,

You know, Google's very good at knowledge intelligence, right?

Like it's very, very good at it.

If I type a movie name in, it gives me the world about that movie.

Um, and then if, then there are other aspects of intelligence.

But I wanna, I want to.

I wanna demystify I, and I think it, it's what's throwing us off

is the fact that this latest trend in AI is, um, very conversational

and is sort of throwing us off

Oh, the chat

the same way that it.

bots and image creators and you say, Hey, draw me a picture of a

the chat bot, which I think what's, what's happening is like a cat

seeing itself in the mirror kind of vibe.

It's like, oh, who invited you here?

That's

Are you, oh, you're moving, you're actually moving when I move.

I'm gonna kill you.

I'm gonna kill you.

And so I think, I think that's throwing us off.

It's it,

it, it's feigning intelligence in a, in not just intelligence, but almost

a little bit of arrogant personality in the whole vibe of the thing.

Right.

And I think what we're doing as humans, which to humans tend to

do is we're inferring intent.

Yes.

And when you infer intent, wait a minute, you seem to have a plan.

And when you infer intent, you can start to quickly draw up scenarios

of malignant intent, like Ill intent.

Negative intent as if the machine is like, oh, human, you thought you had

boxed me in, but I will take this further and it won't take it further.

In fact, it's just going to do what these machines do and it,

and what's happening right now.

And I think a lot of the, the, the, the anxiety is born out of this illusion

that's in front of us right now, which is like, this thing seems to have

real agency and obviously it doesn't,

I gotta tell you too, the people who are into this stuff, they

either have no kids or they have like 20 kids and, and like, and no, because if

we, you spend a lot of time with a child.

Yeah.

You realize how weird humans are.

It's not like, can you draw me a picture of a snake and it

draws you a picture of a snake?

You say, can you draw me a picture of a snake?

And suddenly you're like getting someone to stop climbing a tree 6 65 feet away.

Yeah.

Right.

It's just humans are perverse bananas like, like social creatures.

You cannot,

what's, what's, what makes 'em beautiful in a lot of

well, I, being a dad is something else, but Yes.

Yes.

Now look.

So let's park, the machines are gonna turn on us for a minute.

That's something we just have to train our brains to accept that.

Just because what is essentially a Google sentence writer

is, is throwing words at us.

It is not have a hidden agenda.

Like that's not

It's also, the problem is never the machine turning on you.

It's the machine making things so easy that you sit there

and consume carbohydrates.

Well, I thought you were gonna say something else.

Yeah, but that's true.

That's right.

Like the real, the real risk of AI is that you just will never leave your house.

Cuz it'll be like, Hey, looks like you want more hostess

That's right.

That's right.

So, so if we park the,

the like, you know, Evil machine has turned on me narrative, the

classic science fiction narrative.

Then a lot of the like anxiety goes down.

But there is still another issue.

There is still another issue is that humans tend to put

those protection mechanisms.

I'm gonna, I'm gonna go ahead and venture that the first power washers

didn't have safety mechanisms built

Well, there's al, there's always a reason why the fence is there.

I remember I had a, a teacher in high school and he came up with the concept.

He says, I always like to think about food heroes.

Who is the, the, you know, like literally the, the prehistoric man

who's like, that mushroom looks good.

No, no

bless his heart.

Gronk didn't make it

Joe

Yeah,

we named it after him.

that guy though, he was like, he was right.

Right.

yeah.

And so, so we tend to step in the shit first before we put

those mechanisms, mechanisms in

I mean, you know, I do.

I can't imagine like, let's riff for one sec cuz the world

that I just described, right?

Here's what I could see.

I could see like, this is the number one ai, uh, hospice care tool, and

it'll, it'll deliver morphine to you.

I mean, you could see like, and then it's like, well,

granddad died six months sooner.

He was actually fine.

Right?

And

He sneezed.

He sneezed.

the, the microphone on the AI machine picked up a

and he was really, he, he was happy, but it was over very fast.

And so we're gonna sue you for a hundred million after that suit.

After that suit goes through.

Yes.

Then suddenly it'll be like, you know, maybe hospice isn't what ais

are supposed to do, so we're going to try a combinatorial range of options

with all these new technologies to see where we can jam 'em in.

That's what we do.

That's capitalism and like we like that.

I, I think, look, I, I, I think.

I think the real

you and me personally, but we collectively asci society.

I'm sure everybody's like hospice and ai.

Man.

Those are two tastes that go great together.

I look you may be right because look, we, we, we let people

start up companies and say, you can do this with a third less staff.

Yeah.

Right.

And then here's the tool.

Right?

And we tend to do that, but

Remember Theranos is like the perfect example there, right.

I'm gonna close it with two points.

One is my, where my real anxiety li lies and, and then counter it with

why I'm more optimistic than not.

My real anxiety lies in the fact that most of the damage

done is by people to this day.

Like there's a lot of stuff Facebook did.

Whether intentional or not that caused damage.

And what humans tend to do, and this is where my fear actually is,

you want, you want to use the word existential risk, like extinction.

I look at humans more than I look at machines.

Humans with machines that have incredible capability, they will use them right.

Foreign actors have used our platforms to try to create division

and disruption and whatnot.

We probably do it to other countries like humans are really kind of the story

I, I see this a lot with the climate stuff where everybody's

like, we have to make change.

We have to make change.

We do.

And in retrospect, we will have needed to, but we have never learned anything the

easy way as a species ever, ever, ever,

ever,

And so it, and it's like, well, will the death

toll be hundreds of millions?

Yes, very possibly.

Pause.

Like, and I, I don't mean to like bring everybody down here,

but like that's what we are.

And every time we try to change it, somebody comes up and is

like, you can't change it.

Yeah.

You can't change it.

And, and here's, and so what am I optimistic about?

What I'm optimistic about is that the adversarial system of law that

we have tends to keep us in check.

I'm gonna tell you who's not going to tell lies about voting machines in the future.

Other media is not going to do it because the signal that was sent

out, and frankly, I will sue the, the

oh oh meaning cuz Fox News got such a big pickle

with the voting machines.

I had to spend something 700 million to get out of it, right?

know, I, I, I looked at that.

I was, and I didn't feel sad.

Wow.

I'm glad they got theirs and whatnot.

What I felt was, I'm like, my God, we needed that signal out

Yes.

Because now like if you're a true propaganda outlet, you have to actually

still go check some basic facts.

That's right.

That's

Once they got, once they got those text messages where

it's like, I don't care what the truth is, that was, that was bad

The warning label on the medicine bottle is not because the pharma

company is really worried about your kids.

They're worried about getting sued.

Yeah.

the system that we have allows us to hold accountable.

People that have things that are powerful in their

Thank God for lawyers.

Really.

I mean, really every

really like ultimately Pete, you're welcome.

Yeah, you're welcome.

I'm an attorney.

I don't practice.

But man, thank God for lawyers and thank God for our, like

grossly litigious society.

Um, uh,

God bless you.

Just, I, I appreciate the way that you just took two of the worst

things like AI culture in 2023.

Yeah.

And you're like, I know the answer.

Lawyers.

Lawyers.

So yeah.

Everybody just, if I was you, I would just go find a swimming pool

Paul, I, I'm a realist man.

I'm not gonna rely on people's good graces to behave themselves with powerful tools.

It, it's never happened in the history of

honestly.

Even secondary tools, like you could, I'm, I'm, you could start

a nuclear war with a Commodore 64.

Exactly.

Exactly.

Now look, I, I.

I think the more subtle example is like, whoa, I coded it to do a certain thing and

I didn't expect it to get away from me.

And we do let, tend to let tech get away from us.

that is real.

here's what I would say.

We all know the existential re risks and we're ignoring them too.

Might as well ignore the one about ai, but also deep down, I do believe

that we underestimate nuclear risk.

I believe that we.

Extremely underestimate climate risk.

Culturally, I don't think we're underestimating AI risk right now, today.

I think there could be some sea changes in how technology is

deployed throughout the world.

That could definitely give you the willies.

But for right now, man, it's a jump.

It's a jump to get

It's a jump.

But you know someone that isn't that.

Tech savvy watching that thing answer as if like you just attach voice to

chat GP team, like, oh, good god,

And look, we gotta close this one out.

But what actually happens?

There's that.

And then there is the community that gets high on their own supply

and they're like, this is it.

We've created the ultimate super weapon,

r i p landscapers, five AI tools that are gonna

change the way you mow your lawn.

Right?

exactly.

Oh my God.

It's, it is an extinction level event for people who write

really crappy sales emails.

Yeah.

Good riddens.

Yeah.

Anyway, to our future AI overlords.

Um, I hope you enjoyed this podcast when you

Peter.

Don't sweat it, man.

We're good.

yeah.

Relax, worry about climate

Yeah.

Uh, well, you have been listening to the Zian Ford Podcast,

sponsored by a board@aboard.com.

The podcast before this one dives deep into a board.

It's this wonderful startup.

It's also the sponsor of this podcast that's also our venture.

We're the founders of a board.com.

Check it out.

Sign up for the beta.

And give us five stars if stars.

Five stars is still a thing on the internet.

Hey, you wanna know something that happened with

podcast listeners last week?

I said If you dmd us at Citi Ford or at aboard

Yeah.

That we would wave you in early to the demo.

Yep.

And, uh, that happened.

Yeah.

Got a couple.

I let you jump the line.

So come on in.

come on in.

Have a lovely week.