0:00

Whether you're on a cross-country drive or on your

0:02

daily commute, time in the car is perfect

0:04

for listening to podcasts. T-Mobile's network

0:06

can help keep you connected to all your favorite

0:08

podcasts when you're out and about. T-Mobile

0:10

covers more highway miles with 5G

0:12

than anyone. So if you need great coverage,

0:15

especially when you're on the go, check out T-Mobile.

0:17

They're the largest and fastest 5G

0:20

network. Find out more at t-mobile.com

0:23

slash C-Y. That's S-E-E-W-H-Y.

0:27

Pack lightly? Not a chance. In

0:29

the all-new 2024

0:30

Chevy Trax, you and your squad have the space

0:32

and versatility you deserve and

0:36

need for your next big adventure. With large,

0:38

modern display screens, wireless phone

0:40

connectivity, and affordable pricing, this

0:42

small SUV is kind of a big deal. But

0:45

if you're looking for a car that's fast, reliable, and has

0:47

a great range, this is the car for you.

0:50

It's the T-Mobile C-5G. It's a 5G-based 5G network,

0:52

and it's a 5G-based 5G network. With affordable

0:54

pricing, this small SUV is kind

0:57

of a big deal. And with the option to

0:59

choose between stylish designs like

1:01

the cool, competent Active or the ultra-sporty

1:04

RS, you're sure to find the all-new Chevy

1:06

Trax fits your budget and your brand.

1:09

Oh, hey, it's the bunny that you swear

1:11

you saw on the lawn, even if no one else believes

1:13

you, Allie Ward. And here's all the

1:16

Gs. Hey, am I a real person? Unfortunately,

1:18

I am. Am I intelligent?

1:21

That's up for debate. But this week, we

1:23

are taking a dive into artificial intelligence

1:26

and brain data with a scholar

1:28

in the matter. So listen,

1:30

the past few months have been a little surreal.

1:33

Photoshop's out there generating backgrounds

1:36

to cut your cousin's ex-girlfriend out of your wedding

1:38

photos. ChatGPT is

1:40

writing obituaries, and frankly,

1:43

a lot of horse bucky. There's

1:45

also groundbreaking labor strikes in the

1:47

arts, which we covered in the field trip

1:50

episode from the WGA Strike Lines.

1:52

If you haven't heard it, I'll link it in the show notes. But

1:55

I heard about this guest's work, and

1:57

I said, please, please, please talk to me about

1:59

how to feel about AI. Are

2:02

we farting around the portal to

2:04

a new and potentially shittier

2:06

way of living? Or will AI

2:09

say, hey, dipshits, I ran

2:11

some simulations and here's what we have to do

2:13

to unextinct you in the next century.

2:15

We're going to find out. So this guest has studied

2:18

law at Dartmouth, Harvard, and Duke,

2:20

and been a professor at Vanderbilt University

2:22

and is now at Duke's Institute for

2:25

Genome Sciences and Policy. She

2:28

recently delivered a TED Talk called

2:29

Your Right to Mental Privacy in the

2:32

Age of Brain Sensing Tech,

2:34

and just authored a new book called The Battle

2:37

for Your Brain, Defending the Right to Think

2:39

Freely in the Age of Neurotechnology.

2:42

But before we chat with her, a quick thank you

2:44

to patrons of the show who support at patreon.com

2:47

slashologies for a buck or more a month

2:49

and submit their questions for the second half. And

2:51

thank you to everyone inologiesmerch.com,

2:54

shirts and hats and such. Of course,

2:57

you can also support the show just by leaving a review

2:59

and

2:59

I may delight you by

3:02

reading it, such as this one left this week by

3:04

environmental lawyer, Harrison, Harrison,

3:06

Harrison, who wrote a review calling

3:08

ologies and ooey gooey

3:11

ratatouille rip roaring. Good

3:13

time. So yeah, I read them all. Thank you,

3:15

Harrison, for that. Okay. Neurotechnology.

3:18

Let's get into this. How the brain interacts

3:20

with technology and also techno

3:22

neurology, how tech is striving

3:25

to replicate and surpass human

3:27

intelligence and what that means for us all.

3:30

So let's be bump our way into a talk about

3:33

texting, scrolling, cheating, brain

3:35

implants, mental health, doomsday

3:38

scenarios, congressional hearings,

3:40

apocalypse potential, medical

3:42

advances, biometric mining,

3:45

and why suddenly artificial intelligence

3:48

is

3:48

on our minds with law

3:50

professor and neurotechnologist,

3:53

Dr. Nita Farhani.

4:01

Nita Farahani,

4:04

it's she, her.

4:13

So

4:15

good to meet you. Terrifying to

4:17

meet you. Are you the scariest person at a dinner party because

4:19

of how much you know? I'm not. I'm

4:21

not a scary person. And I

4:24

find that people think that

4:26

it's equal parts fascinating and terrifying. So

4:28

if anything, I think I'm a great dinner guest, right? Because

4:30

they're fascinated. I definitely should clarify that.

4:33

You are, there's nothing scary about you.

4:35

The information that you hold is

4:37

like, no, I know. Do I want

4:39

to look? Do I not want to look? Do I want

4:42

to look? It's thrilling, like a horror

4:43

film. Yeah. It's like people can't

4:45

look away. Yes. Right? I

4:48

want them to know. But at the same time, what I usually get is like,

4:50

wait, this is real? Like what you're talking about is

4:53

it actually exists and people are really using it and

4:55

employers are really using it and governments are really

4:57

using it. And

4:58

wait, what? Yeah. Do you spend a lot

5:00

of your time chatting with people trying

5:03

to warn them or calm them down? Yes.

5:08

So on

5:10

the one hand, I am trying

5:12

to raise the alarm and

5:14

to help people understand that this whole area

5:17

of being able to decode and really

5:20

hack and track the brain is a new

5:22

frontier and the final frontier of

5:25

what it means to be human and privacy

5:28

and freedom. And at the same

5:30

time, I don't want to make people

5:33

have the reactionary approach to technology,

5:36

which is like, okay, then let's ban it because

5:38

the promise is also extraordinary.

5:41

And so I am very much equal

5:44

parts. Like let me help you understand

5:46

not only what the promise is and

5:48

why you're likely to adopt it, but

5:50

why before you do so and before we

5:52

as a society at scale adopt

5:55

this technology that we make some really important choices

5:57

that will actually make it good

5:58

for us and not.

5:59

the most Orwellian, frightening,

6:02

scary thing possible. I feel like there's few

6:04

topics that have this much true ambivalence

6:07

of so much good and so

6:09

much potential for misuse. Did

6:11

your brain become a lawyer brain because

6:14

of those sort of like philosophical conundrums?

6:17

What drew you to this kind of deep, deep

6:19

thought? Yeah, I've always been driven to the questions

6:22

that are at the intersection of philosophy

6:25

and science. In high school,

6:27

I was really interested in the science, but I was a policy

6:29

debater.

6:29

In college, I was a government

6:32

minor and science major. I

6:35

did in lab stuff, but largely

6:37

things that were policy. So

6:39

Nita got several graduate degrees studying

6:42

law and science, behavioral genetics

6:44

and neuroscience, the philosophy of mind, neuroethics,

6:47

bioethics, and even reproductive rights

6:50

and policy in Kenya. And she said

6:53

all her work seems to gravitate toward

6:55

this intersection of philosophy and law

6:58

and science because she had

6:59

fundamental questions like, do

7:02

we have free will? And

7:04

do we have fundamental autonomy

7:07

and freedom? And how do we put into place

7:09

the protections? But I've always been fascinated

7:11

and really interested in the science

7:13

and the technology itself. I've never been a

7:15

Luddite. I've always been somebody who's an early

7:17

tech adopter, but clearly see what the downsides are

7:19

at the same time. Where was tech at when

7:21

you were getting that roster of

7:24

graduate degrees? Where were we at? Were

7:26

we at emails? Were we at video calls? Yeah.

7:28

So we were not at video calls.

7:29

We were at emails. The internet

7:32

existed. We used it. We all had computers,

7:34

but we didn't have cell phones. I

7:36

got my first cell phone after I graduated from

7:39

college, like the year after, and I had

7:41

a flip phone. And I

7:43

thought that was super cool. I could

7:45

type out a text message one character

7:48

at a time. Oh, T9? Yeah. I

7:50

was a little bit metal in T9. Nice. I could

7:52

do it without even looking at

7:55

the phone, where I found it harder when we

7:57

had a keyboard. Yeah. And

7:59

then I had a Palm Pilot.

7:59

like as the precursor to the iPhone.

8:02

And then I stood in line the first day that the iPhone

8:05

was being sold and got one of the first

8:07

iPhones in my hand. So I've seen

8:09

the evolution of tech, I guess, as I was

8:11

getting all of those degrees. And what about

8:13

in

8:14

terms of neurotechnology? Have you seen

8:16

kind of an exponential growth pattern in

8:19

terms of technology? Is that growth

8:22

pattern still valid or have we

8:24

surpassed it? Slowly over

8:26

the

8:27

past decade or two,

8:29

neurotechnology has been getting better. And the ways

8:31

in which neurotech has been getting better has

8:33

largely been kind of hardware-based, which

8:36

is the sensors are getting better. Sometimes

8:38

the software has been getting better to be able to filter

8:41

out noise, the algorithms

8:43

to be able to pick up brain activity

8:46

without having muscle twitches

8:48

or eye blinks or interference

8:50

from the environment to pick up different information.

8:52

All of that's been getting better. But suddenly

8:54

we've gone from what was

8:57

improvements to just the past

8:59

five years, seeing much more rapid advances.

9:02

Generative AI is making things move

9:05

and these seismic shifts, like where you suddenly

9:07

have just a massive leap in capabilities.

9:10

Just real quick, before we descend into

9:12

the abyss of ethics and possible

9:15

scenarios, what is generative

9:17

AI? What is AI and what's just

9:19

a computer computing? Okay, I looked

9:21

this up for us and then I took a nap because

9:24

it was confusing and then I tried again. And here's what I sussed out.

9:27

So artificial means it's coming

9:29

from a machine or software and intelligence,

9:31

fuck, I mean, that depends on who you ask,

9:34

but broadly it means a capacity

9:36

for logic, understanding, learning,

9:39

reasoning, problem solving and retaining

9:41

facts. So some examples of

9:44

AI are Googling or search engines,

9:46

the software that recommends other things

9:48

you might like to purchase, navigating

9:51

via self-driving cars, your Alexa

9:54

understanding when you scream, Alexa

9:56

stop

9:57

because she tried to get you to subscribe to Amazon

9:59

Prime.

9:59

again. It also includes

10:02

computers being chess nerds, that's AI,

10:05

and generating artwork. And according

10:07

to some experts, AI can

10:10

be separated into a few categories including

10:12

on the base level, reactive machines,

10:14

and those use existing information

10:16

but they don't store or learn anything. Then

10:19

there's limited memory AI that can

10:21

use precedent to learn what choices

10:23

to make. There's something called theory of mind

10:26

AI and that can try to figure out

10:28

the intentions of a user or even

10:30

acknowledge their feelings like if

10:32

you've ever told Alexa to get bent

10:35

in a lot of other words and then she sasses

10:37

you back. There's also a type called self-aware AI that

10:39

reflects on its own actions

10:47

and then fully autonomous is

10:49

kind of the deluxe model of AI and

10:53

that does its own thing. That sets its own goals,

10:55

set it and forget it if you can. So when

10:57

did things start speeding up? When did they

10:59

start careening toward the future like this? When

11:02

computers got faster and smaller and

11:05

better in the last 10 but

11:07

really kind of 2 or 3 years. So

11:09

better hardware means more processing

11:11

power. There's also cloud storage

11:14

and that adds up to something called deep learning

11:17

which kind of sounds creepy like a hypervigilant

11:20

mannequin but deep refers

11:23

to many layers of networks

11:25

that use what look like these complicated

11:27

flowcharts to decide what actions

11:29

to take based on previous learning. So that's

11:31

kind of what led up to these startlingly

11:34

human-like generative AI outputs

11:36

and deep fakes where they

11:38

can just straight up put Keanu Reeves

11:40

face on your mom and then confuse

11:43

the bejesus out of me on TikTok or

11:46

chat GPT which is one language

11:48

model chat bot. Computers are

11:50

starting to pass bar exams.

11:52

Maybe they're writing the

11:55

quippy flirtations on your dating

11:57

app. Who knows? Meanwhile, less than 100% of the users are using the

11:59

app. So that's pretty much it.

11:59

years ago, a lot of the U.S. didn't

12:02

have flush toilets in case you

12:04

feel weird about how weird this feels because

12:06

it is weird. Evolutionarily, our

12:08

flabby, beautiful little brains can barely

12:11

handle the shock of a clean river

12:13

coming out of a garden hose, let alone

12:16

some metal and rocks that are computers

12:19

that were training to potentially kill

12:21

us.

12:21

We don't know how to deal with that. So pattern

12:24

recognition using machine learning algorithms

12:26

has really pushed things forward rapidly.

12:29

A lot of brain data that happens in characteristic

12:31

patterns and those associations

12:33

between what is a person seeing or hearing

12:36

or thinking, how are they feeling,

12:38

are they tired, are they happy, are they sad, are

12:40

they stressed, those things have been

12:42

correlated with huge data sets and

12:44

processed using machine learning algorithms

12:46

in ways that weren't possible before.

12:48

I can read your mind. Then you

12:50

have generative AI and chat GPT

12:52

that enters the scene in November.

12:55

All of a sudden, the

12:58

papers that are coming out are jaw-dropping.

13:00

Data that's being processed by generative

13:02

AI to reconstruct what a person is thinking

13:05

or hearing or imagining or

13:07

seeing is next level.

13:09

My book came out March 14th, 2023. All

13:11

of a sudden, what was happening was continuous

13:14

language decoding from

13:18

the brain in really,

13:19

really high resolution

13:22

using GPT-1, not even the most advanced

13:24

GPT-4. Visual reconstruction

13:26

of images that a person is seeing in ways

13:29

that were much more precise than anything

13:31

that we had seen previously. That's

13:34

happening at this clip that is just,

13:37

I think, extraordinary. It's just so much

13:39

faster than even I would have imagined. Even

13:41

I could have anticipated even

13:43

having written a book about the topic. That was

13:46

literally going to be my next question because

13:48

when a person writes a book, that doesn't happen overnight.

13:51

Even working on this book probably for a couple

13:53

of years, did you have any

13:55

idea that your book would be so

13:58

closely timed to such a giant

14:00

leap in terms of public

14:03

perception and awareness of AI. I

14:05

mean, it couldn't have timed it better. Well,

14:07

I mean, of course I'm a futurist. I

14:10

was predicting it perfectly, right? No. No.

14:12

I mean, I wish, right? In truth, my book

14:15

is like a year and a half late from when I

14:17

was supposed to turn it into the editor, into

14:19

the publisher, but there was a global

14:21

pandemic that got in the way and a bunch of other

14:23

things. But I'm grateful that it didn't

14:25

happen sooner because I was both able to

14:28

be part of what is a growing conversation

14:31

about the capabilities

14:31

of AI and to see when

14:33

you say to a person like, oh yeah, also AI

14:36

can decode your brain. You

14:38

know, that really puts a fine point on it for people

14:40

to understand how quickly these advances are coming

14:43

and to see how it's changing everything

14:45

in society, not just how people are writing essays

14:48

or writing emails, but fundamentally unlocking

14:51

the mysteries of the mind

14:53

that people never thought before possible and

14:56

the risks that that opens up, and the possibilities

14:58

of mental manipulation and hacking and

15:00

tracking, those are dangers

15:02

that I think a year ago,

15:05

before people really woke up

15:07

to the risks of

15:08

AI, they would not have been having the

15:10

conversation in the same way that they are around the book

15:12

and now they are having that conversation, seeing

15:14

the broader context and seeing the

15:16

alarm bells everywhere, right? Like, oh

15:19

wait, we really do need to regulate or recognize

15:21

some rights or do something. So futurists

15:23

are urging some foresight. Congressional

15:26

panels have aired on C-SPAN and there

15:28

seems to be this kind of collective side-eye

15:31

and like a hope that someone's

15:34

on top of this, right? So I mean, I think

15:36

people are looking for some guidance

15:38

and to have somebody

15:40

come at it from a balanced perspective, like, wait

15:42

a minute, there's a lot of good here and

15:44

there's some serious risks and here's a potential

15:46

pathway forward. I think instead

15:49

of like pause, which everybody says like, of course,

15:51

we can't just pause or a doomsday

15:53

scenario without any positive, like,

15:56

oh, let's regulate AI. I think we need voices

15:58

at the table who are thinking about.

15:59

it both in a balanced way but also are coming forward

16:02

with like here are some concrete things we could

16:04

do right now that would actually help

16:06

the problem. So we know a few types

16:08

of AI from Googling a source

16:11

for a research paper or digitally

16:14

removing your cousin's ex from

16:17

your wedding photos, but what about technology

16:19

that's gathering data from our

16:21

brains? Let me give you the spectrum. There's

16:25

medical grade neurotechnology.

16:28

This is technology that people

16:30

might imagine in a doctor's office

16:33

where somebody puts on an EEG, electroencephalography

16:36

cap that has a bunch

16:38

of different wires coming out of it and a bunch of

16:40

gel that's applied to their head and a bunch of sensors.

16:43

That's picking up electrical activity, which we'll get back to in a minute.

16:46

Then there's the clunky giant machine,

16:48

a functional magnetic resonance imaging machine,

16:50

which can peer deeply into

16:52

the brain and somebody

16:55

might have already undergone an fMRI test

16:57

for something like a brain tumor to kind of

16:59

look more deeply into the brain. What that's picking

17:02

up is changes in blood flow

17:04

across the brain, which tells us something

17:06

about different areas that are activated

17:08

at any particular time and what those patterns

17:10

might mean. So if you've never had an MRI,

17:13

I guess congratulations, that's probably good, but

17:15

this is magnetic resonance imaging.

17:18

It's pretty exciting how these strong

17:21

ass magnets all line up the

17:23

hydrogen atoms in your body to go one direction

17:25

and then they release them and

17:27

from that they can see inside of your body.

17:30

Now, an fMRI is a

17:32

functional MRI and to put it in super

17:34

simple terms, it's kind of like animation

17:37

instead of a still picture, but it's of your

17:39

brain. So when you see imaging

17:41

examples of how someone's melon

17:44

illuminates like a Christmas tree

17:46

to certain stimuli, that's fMRI

17:49

technology tracking blood flow to

17:51

different regions of the brain. This fMRI

17:53

technology is used in a lot of

17:55

neuro and psychology research.

17:57

And then there's something like functional near infrared spectra-

17:59

which is more portable

18:02

and it's also measuring changes in the

18:04

brain, but it's using optical and

18:06

infrared lights in order to do so. And

18:09

that functional near infrared

18:11

spectroscopy looks for changes

18:13

in oxyhemoglobin and deoxyhemoglobin

18:17

in the brain. These words might not matter to

18:19

you right now as you're cleaning your

18:21

shower grout or your carpooling. But in clinical

18:23

settings it comes in handy for patients

18:26

with strokes or learning about Alzheimer's

18:28

or Parkinson's or even anxiety

18:29

or a traumatic brain injury,

18:32

which my brain would like you to know I've had

18:35

and I will link the traumatic

18:37

brain injury or the neuropathology episode

18:39

about my Helen Narnar concussion I

18:41

got last year in the show notes. But yes, there

18:43

are a lot of ways to get data from a brain

18:46

including CT scans and PET scans

18:48

with radioactive tracers. But what

18:51

about non-medical uses? Do

18:53

they exist? Oh boy, how do you do they?

18:55

If you then look at what's happening in the consumer

18:58

space in the consumer space you take

18:59

the 64 or 120 electrodes

19:03

that are in a big cap and then you have a couple of them

19:05

that are put into a foreheadband

19:07

or a baseball cap or increasingly

19:10

what's coming is brain sensors that are embedded

19:12

in everyday technology. So you and I are

19:14

both wearing headphones and

19:17

the soft cups that go around our ears are being

19:19

packed with sensors that can pick up brain activity

19:22

by reading the electrical activity through our

19:24

scalp. You want my tinfoil hat? Or

19:27

if we were wearing earbuds inside of our ears

19:29

instead embedding brain sensors

19:31

inside of those that can pick up electrical activity

19:34

in our brain activity as we're thinking or doing

19:36

anything and those

19:39

become much more familiar and much more commonplace

19:41

very quickly. So there's just a few of those products

19:43

that are on the market but that's where most

19:45

of the big tech companies are going is to embed

19:48

brain sensors into everyday devices like

19:50

earbuds and headphones and even watches

19:52

that pick up brain activity from

19:54

your brain down your arm to your wrist and picks

19:57

up your intention to move or to type or to

19:59

swipe or something like that.

19:59

So to use like a medical

20:02

analogy, you know continuous glucose

20:04

monitors. These are a powerful tool

20:07

for diabetics to monitor their blood sugar

20:09

levels and their insulin needs and we covered those in

20:11

the two-part diabetology episode with Dr. Mike Natter.

20:14

But now continuous glucose monitors are

20:16

starting to become available to people without diabetes

20:19

just to better understand their metabolisms

20:21

and their dietary responses, their mood

20:24

and energy. So all of these neuroimaging

20:27

and all this data was just used

20:29

in clinical

20:29

and research settings by people in crisp

20:32

coats carrying metal clipboards

20:35

but it's starting to pop up in the market now. This

20:37

is great news, right? You understanding your brain?

20:39

Yeah,

20:40

yeah, but not

20:41

all the research in consumer applications

20:44

is solid and some make some wild

20:46

claims of efficacy. Others

20:49

argue that if a device can enhance

20:51

our moods and sharpen us

20:54

cognitively but cost some

20:56

serious cash, doesn't that just widen

20:59

a privileged gap even further? But I

21:01

guess so does college. I

21:03

don't know. In the US, you need a GoFundMe

21:06

to pay for chemo. So we've got a lot

21:08

of pretty large systemic

21:10

fish to fry.

21:11

But if you've got money, you

21:14

can buy EEG headsets

21:16

that track your mood and emotions and stress

21:18

for a few grand. There's others

21:20

that track your heart rate and brainwaves

21:23

for sleep and meditation. There are

21:25

VR gaming sets that track

21:27

brainwaves and even a Mattel game

21:29

called MindFlex. You can buy for like 120

21:32

bucks. But Nita says, all

21:35

of those consumer-based technologies pick

21:37

up a little bit of like kind of low resolution

21:40

information right now.

21:41

They pick up if you're stressed, if

21:43

you're happy or if you're sad, if you're

21:45

tired, like it maybe picks up that your

21:47

mind is wandering and you're kind of like

21:49

dozing off and the

21:52

things like fMRI pick up

21:54

much more precise information. Now

21:57

that could just be a matter of time. It could

21:59

be...

21:59

that as machine learning algorithms

22:02

and generative AI get applied to the

22:04

electrical activity in the brain, that it'll

22:06

get better and better and better. And it's

22:08

interesting, because in a way, you could think about AI

22:11

as being the convergence between computer

22:13

science and neuroscience. So

22:15

computer scientists have been designing

22:18

algorithms that can process

22:21

information in very narrow ways, and

22:23

they're very good at doing specific tasks. So

22:25

for example, a doctor or a pathologist

22:28

who's looking at many different samples

22:29

of tissue

22:32

to figure out if it looks cancerous or not

22:35

can only see so many samples

22:37

in a lifetime. And so they've marked them

22:39

and labeled the data. And a machine

22:41

learning algorithm can be trained on that

22:44

data, which is like, here's

22:46

thousands of images that are cancer

22:48

and not cancer. Now, here

22:51

are new images, predict whether or not they

22:53

have cancer. And they become very, very good,

22:55

because they can process millions and millions

22:57

of images and see far more images and get

23:00

much better at being able to do

23:02

that specific task of identify if something

23:04

is cancerous. So those tasks are relatively

23:07

simple for machines to learn and

23:09

execute. Computers are like, child's

23:11

play. But the human brain

23:14

isn't so narrow and task-specific.

23:16

And neuroscience has long understood

23:19

that the connections that the brain makes

23:21

are much more multifaceted.

23:24

They're much more complex. And

23:26

so the modern

23:29

types of AI are

23:31

built on how the brain

23:33

works. They're built on what are called neural

23:35

networks. So this is a deep learning

23:38

model, which is instead of that very

23:40

specific task of like, do this,

23:43

do that, it's meant

23:45

to take a huge amount of information

23:47

and to learn from that information and

23:49

then do what we do, which is to predict

23:52

the next thing or to kind of understand where

23:54

that's going or to make inferences from more

23:56

of a deep learning perspective.

23:58

So it's more than machine learning.

23:59

learning like the pathology example

24:02

she gave. So remember deep learning. So

24:04

neural networks are modeled after

24:07

biological brains and they have nodes

24:09

like neurons that consume input

24:11

that they learn from and then it's processed

24:13

in several layers or tiers aka

24:16

it's deep to come up with

24:18

a result or an action and

24:21

things like chatbots or facial

24:23

recognition or typing dog

24:25

into your phone's photo album to see what

24:28

goodness comes up or speech to text.

24:29

Those are all done by neural networks and

24:32

AI that we're already using and they seem

24:34

commonplace after having them for just a few years.

24:37

But since late last year we're

24:39

seeing them create more like

24:41

how the human brain might. And those insights

24:43

about the brain and neural networks have

24:45

informed this new class

24:47

of AI which is generative AI generative

24:50

AI is different in

24:52

that it is both built on a different model and

24:55

it has much greater flexibility in what it can

24:57

do and it's trying to not say

24:59

like this is cancer that isn't cancer

25:02

but to take a bunch of information and then be

25:05

asked a question and to respond

25:07

or to generate much more like the human

25:09

brain reasons or thinks or comes up

25:11

with the next solution. And that's exciting

25:14

and terrifying. What about

25:17

the

25:18

information that say the

25:20

artistic AI is getting? Are they scrubbing

25:22

that from existing art and

25:25

in the case of say the writer strike

25:28

where you see writers saying

25:30

you cannot

25:32

take my scripts and write a sequel on something

25:35

without me. And if you're curious about what

25:37

is up with these strikes, what is going on

25:39

in the entertainment industry including

25:42

the WGA or the Writers Guild of America

25:44

strike which started on May Day of this year

25:47

and it was joined in recent weeks on the picket

25:49

lines by SAG-AFTRA which is a screen

25:52

actor skilled. And again we did a whole episode

25:54

explaining what is going on. It's called field trip

25:56

WGA strike that will be linked to the show notes. So

25:59

if you want watch TV or movies

26:01

or you ever have, listen to that episode

26:04

because it affects us all. And

26:06

these entertainment labor unions are known as

26:08

the tip of the spear for other

26:11

labor sectors. Your industry may

26:13

be affected or might be next.

26:15

I'm really interested in what happens in

26:17

this space, not just because of the

26:20

writers themselves and hoping

26:22

that they

26:24

managed to succeed

26:26

and actually getting fair, appropriate

26:30

treatment,

26:31

but also because it's gonna be incredibly telling

26:34

for every industry as what

26:36

happens when workers demand

26:38

better conditions and better

26:41

terms.

26:42

And the result is

26:44

greater experimentation with generative AI to

26:46

replace them. But why is this such a sudden

26:49

concern? Why does it feel like AI has

26:51

just darkened the horizon and thundered

26:54

into view and we're all cowering

26:56

at its advance? Is this the first act

26:58

of a horror film? So where does it come

27:01

from? They're not totally

27:03

transparent. We don't know all of the answers to that,

27:05

right? But we do know that these

27:07

models have been trained, meaning

27:10

there's been billions,

27:12

potentially trillions, we don't know, right, the exact number

27:14

of parameters. That is prior

27:17

data which has been used. Meaning

27:19

the material that the machines learn from.

27:22

And that could be prior scripts, it could be prior

27:24

books. It includes a bunch of self-published

27:27

books, apparently, that are part of it, prior

27:29

music, prior art, potentially

27:32

a whole lot of copyrighted material that

27:34

has been used to inform

27:36

the models. Once the models learn, they're

27:38

not drawing from that information

27:40

anymore, right? That information was used to train

27:43

them. But in the same way

27:45

that you don't retain everything you've ever read

27:47

or listened to, and your creativity

27:50

may be inspired by lots of things that

27:52

you've been exposed to, the models

27:54

are similar in that they've been trained on these prior

27:56

parameters, but they're not storing or drawing

27:58

from or returning to the machine. them. It's

28:01

as if they have read and digested

28:03

all of that information. And

28:05

I was talking with an IP scholar who I

28:07

like and respect very much. And his perspective

28:10

was, how is that different than what

28:12

you do? Right? You write a book and

28:14

you read tons of information and

28:16

there's tons of information you cite. And there's also tons of information

28:19

that you learned from, that inspired

28:21

you, that shaped how you write and think that

28:23

you don't actually cite. And is that actually

28:26

unfair or violating

28:29

the intellectual property or somehow

28:31

not giving a fair shake to every

28:34

source that you have ever learned from or every input

28:36

that you've ever learned from? I mean, it's an interesting

28:38

and different perspective, right? I don't have the answers to

28:40

it yet. I'm really

28:41

interested to see how this particular

28:43

debate evolves. What do other people think

28:45

who aren't me? So a recent study reported

28:47

that about 50% of AI experts

28:50

think there's a 10% chance

28:52

of unchecked AI causing

28:54

the extinction of our species

28:57

with AI getting into little sneaky

28:59

elf on the shelf shenanigans like playing

29:01

God or establishing

29:04

a political dictatorship. And the

29:06

Center for AI Safety issued a

29:08

statement. It was signed by dozens of

29:11

leaders in computer science and tech, including

29:13

the CEO of Google's Deep

29:15

Mind and Bill Gates and

29:18

the guy who started ChatGPT

29:20

and the director of a center on strategic

29:23

weapons and strategic risks.

29:26

And this statement said very simply, quote,

29:29

mitigating the risk of extinction

29:31

from AI should be a global priority

29:34

alongside other societal scale

29:36

risks, such as pandemics and

29:39

nuclear

29:39

war. So that's a

29:41

pretty big statement. And other experts

29:44

draw parallels between humans and

29:46

chimps, but we're the chimps and AI

29:49

is us. So guess who's making who wear

29:51

diapers and live with Michael Jackson? Yeah.

29:54

Although of course there are computer

29:56

scientists saying that we need to calm our collective

29:59

boobies.

29:59

and that AI isn't advanced

30:02

enough to threaten us. Yet, I

30:04

love yet. Yet

30:07

is so comfy. Yet is the space between

30:09

the alarm clock and the panic

30:12

of racing out the door because you'll be late to a job

30:14

interview. Ah, yet,

30:16

just yummy, just fuck it. I think

30:18

from a governance perspective in society,

30:21

we have near term risk that we need to

30:23

be safeguarding against. And

30:25

this is near term risk like bias

30:28

and discrimination and inaccuracies.

30:31

I don't know if you saw the story recently about

30:33

a lawyer who filed a brief

30:36

in a case before a federal judge that

30:39

the pleading for the case had been entirely

30:42

written by Chad GPT, which included

30:44

a whole bunch of invented cases. And

30:46

the invented cases, like he hadn't

30:48

gone and sight checked them or

30:51

read them. In fact, he has this dialogue

30:53

where he's asking Chad GPT

30:55

if the cases are real or not, rather

30:57

than like, yes. And

31:00

he was not doing this to prove a point. Just

31:02

a bit of a dumbass. No, no, just straight up dumbass

31:05

just did it. And

31:08

then the other side comes back and

31:10

says, hey, judge, we can't find any

31:12

of these cases. And the judge says, you have to

31:14

produce it. And apparently he produces

31:16

the full citations

31:18

of the made up cases. And anyway,

31:20

it finally goes back with the

31:23

lawyer then admitting, I'm so sorry, this

31:25

is all apparently fabricated

31:28

and it's fabricated not intentionally, but

31:30

it's fabricated because I generated it all using

31:32

Chad GPT. Nita says, who knows

31:35

what will happen if and when where

31:37

people start using bots to kind

31:39

of cut corners and no one fact

31:41

checks it. And around Juneteenth, I

31:43

saw a viral tweet about Chad

31:46

GPT not acknowledging that

31:48

the Texas and Oklahoma border was in

31:50

fact

31:51

influenced by Texas desiring

31:53

to stay a slave state. I told

31:55

my husband, Jared, your pad mom, didn't believe

31:58

it could get things so wrong. And then he...

31:59

proceeded to have like an hour-long fight

32:02

and discussion with ChatGPT, hoping

32:04

to teach ChatGPT that

32:07

it has a responsibility to deliver

32:09

accurate information. I was

32:11

like, dude, you're fighting a good fight and I

32:13

wish you luck. Now, as for this lawyer

32:15

that Nita mentioned, according to a May 2023 New

32:18

York Times piece about it titled, Here's What

32:20

Happens When Your Lawyer Uses ChatGPT,

32:23

the lawyer in question pleaded his own

32:25

case within the case, telling a rightfully

32:28

miffed-off judge that it was his first

32:30

foray with a chatbot and that he

32:32

was, quote, therefore unaware of

32:35

the possibility that its content could

32:37

be false. And the New York Times explains

32:39

that ChatGPT

32:41

generates realistic

32:43

responses by making guesses

32:45

about which fragments of text

32:48

should follow other sequences based

32:50

on a statistical model that has ingested

32:52

billions of examples of text pulled from all over

32:55

the internet. So ChatGPT is

32:57

your friend at the party who knows everything,

33:00

and then you find out that they're full of shit and they're very

33:02

drunk, and maybe they stole your wallet and

33:04

they could kill your dog. Will they shit in the pool?

33:06

It's anyone's guess, but wow, they

33:08

are spicing up the vibe. This

33:11

is not a boring party at all. It raises

33:13

this complex question about, you

33:15

know, who is responsible? And we've

33:18

generally said the attorney is responsible, right? The attorney

33:20

is the one who is licensed to practice law. They're responsible

33:23

for making sure all of the work that they certify under

33:25

their name. Is there any liability

33:27

for generative AI models? Now, ChatGPT

33:30

says, like, I'm not here to provide legal advice

33:32

and it's prone to hallucinations. Is

33:34

that enough to disclaim any liability

33:36

for ChatGPT? Just a jacuzzi

33:39

of hallucinating chatbots

33:41

saying whatever sentence they think

33:43

you want to hear, maybe pooping in there too. So

33:45

what happened to that lawyer, though? Did he get

33:47

so disbarred? Did he have to grow a beard

33:50

and move to Greenland? Does he make felted

33:52

hats out of goat fur now? No, no,

33:55

he's fine. He kept his job. He was just fined

33:57

five grand,

33:58

which if he builds for the research, search hours

34:00

that a chatbot really did, he maybe still

34:02

turned a profit on that deal. But the lessons,

34:06

those are invaluable. Now, if you appreciate

34:08

nothing else today, I just want you to stare off at

34:10

the horizon for even 30 seconds

34:13

and just say, what a time we're

34:15

living in. Hundreds of thousands

34:17

of years of people getting boners and falling

34:19

in love made me a

34:21

person standing on a planet at

34:24

a time when there's plumbing, antibiotics,

34:27

electricity, there's domesticated cats. And

34:30

I have a front row seat to some real madness.

34:33

What an era. As for what we do, I don't

34:35

know. Aren't we being watched all the time anyway? What

34:37

are the watchers doing about this?

34:40

Well, forgive the patriarchal caricatures,

34:42

but where are Big

34:45

Brother and Uncle Sam? Are

34:47

they working together on this? Is there any

34:49

incentive from like a governance perspective

34:52

to say, to step in and say like, we don't

34:54

know how far this should go, or

34:56

does it just generate

34:58

kind of more income for

35:01

maybe big corporations that can misuse it?

35:03

So like, hard to fight against that. So,

35:06

you know, it's hard to know, right? There have

35:08

been hearings that have been held recently

35:12

by the government to try to look into

35:15

sort of both questions that you're asking, which is Uncle Sam

35:17

and Big Brother, right? So there were hearings

35:19

looking at whether or not to regulate

35:22

private corporation use of generative

35:24

AI models. It was, you know, a very

35:27

public hearing where Sam Altman from

35:29

OpenAI calls for regulation. If

35:31

you're

35:31

wondering why this is a big deal. So Sam

35:34

Altman is the CEO of

35:36

OpenAI, which invented ChatGPT.

35:39

And he spoke at the Senate Judiciary Subcommittee

35:41

on Privacy, Technology, and the Law Hearing,

35:44

which was called Oversight of AI, Rules

35:46

for Artificial Intelligence. That was in May

35:48

of this year. He also signed that statement

35:51

about trying to mitigate the risk of

35:53

extinction. And he told the committee

35:55

that AI could, cause

35:58

significant harm to the public.

35:59

world. Papa Chat GPD himself.

36:02

My worst fears are that we cause significant,

36:05

we the field, the technology, the industry cause

36:07

significant harm to the world. I

36:10

think that could happen in a lot of different

36:12

ways. I think if this technology goes

36:14

wrong, it can go quite wrong. And

36:17

we want to be vocal about that. We want to work

36:19

with the government to prevent that from happening.

36:22

And ultimately, Sam urged

36:24

the committee to help establish a new framework

36:26

for this new technology. It was

36:28

a surprisingly collaborative

36:31

tone for most of the federal officials who were

36:33

questioning him very differently

36:35

than in social media context of the past.

36:38

But meanwhile, in a different building, that

36:40

same day, a different hearing was happening, which most people

36:43

weren't aware of, which was federal use of AI.

36:45

And a lot of the discussion in that context

36:48

was about how the federal government needs to be innovating

36:50

to use more AI

36:52

in a lot of

36:54

what they do and to be modernizing what's

36:56

happening.

36:56

Today, we'll be discussing how AI

36:58

has the potential to help government

37:01

serve, better serve the American

37:03

people. Okay, so tonally, the Senate

37:06

Homeland Security and Governmental Affairs Committee

37:08

hearing, which was called artificial intelligence

37:10

in government, was a little bit more optimistic,

37:13

like, hmm, guy, give me some of that. And

37:14

that would include things like Uncle Sam, like improving

37:17

the IRS system and, you

37:19

know, what does filing of taxes look

37:21

like? And are there ways to ease the burden? Are there

37:23

ways to modernize and have different parts of the

37:25

government talking to each other? And hopefully those conversations

37:28

will converge. We won't be looking at like,

37:30

how do we regulate and limit the risks of

37:32

generative AI and then infuse it throughout

37:34

all of the federal government at the same time, right?

37:37

Like, hopefully, like, you have the left

37:39

hand talking to the right hand so that we actually come

37:41

up with a sensible strategy and a road ahead.

37:44

over

38:00

jobs because it's so smart, but at the same

38:02

time, it's worse at Googling than your 10-year-old

38:05

niece with a book report. And while this

38:07

is going on, the government is holding two

38:10

simultaneous hearings on the same

38:12

day, and one is Oppenheimer

38:15

flavored, and the other is Barbie Land. So

38:17

if you are confused by all of this and

38:19

you don't know how to feel, the answer

38:22

is yes, that's correct.

38:25

But it's happening so

38:27

quickly that it's not going to

38:29

be law alone that does anything to rein

38:32

it in. We're going to need a lot

38:35

of cooperation between governments,

38:37

between tech companies. And if you look in

38:39

the U.S., the U.S. has not been good at

38:41

regulating tech companies, right? I

38:44

mean, it has had lots of talk about

38:46

it, lots of very contentious

38:48

Senate hearings. I

38:50

started Facebook. I run

38:52

it, and I'm responsible for what happens

38:54

here.

38:55

And then they have so much money and so much power and

38:57

so much lobbying influence that the

39:00

result is nothing happens.

39:02

And that just can't be the case now. We can't

39:05

go into this era leaving

39:07

it up to tech companies to decide the

39:09

fate of humanity.

39:10

Right. What do you do if you're mad

39:12

as hell and you're not going to take it anymore? What does an

39:14

average person who does not own a $40 billion

39:17

tech company say

39:18

when they're like, don't scrub my brain

39:21

data through my headphones. I'd stop simulating

39:24

art. Some people make some art. Have

39:26

you seen that meme about how somehow we've gotten to a

39:28

place where human beings are still laboring

39:31

at

39:31

wages that don't increase, that

39:33

are not livable, yet computers

39:36

get to write poetry and make art? No,

39:39

but that sounds right. That's such

39:41

a heartbreaking way to look at

39:43

it, where

39:44

no one can afford to be an artist. So

39:46

the exact words from Twitter user Carl

39:49

Sherrow read, humans doing

39:51

the hard jobs on minimum wage while

39:54

the robots write poetry and paint

39:56

is not the future I wanted. So

39:58

that tweet was shared.

39:59

35,000 times because

40:02

it's true and it hurts my soul.

40:05

I haven't seen that meme and now

40:07

I'm reeling from thinking about it, which is like,

40:09

oh my God, that's so true. Suddenly

40:12

we've outsourced all the things that we like and

40:14

we're now doing all of the grunt work still.

40:16

And like, how horrible is that? We're gonna send like

40:19

generative AI to the beach next weekend and

40:21

say, you know, like, what,

40:23

we like stay home and toil and pay

40:25

for it, right? Yeah, I

40:27

mean, you know, the problem is that

40:30

on the one hand we get to like, oh, it's all happening

40:33

so quickly. And so we can't do anything about

40:35

it. On the other hand, that's just the nature of emerging

40:37

tech. It happens quickly. And so it's not

40:39

as if there have not been proposals

40:42

about what agile governance looks like or what

40:44

adaptive regulations look

40:46

like that actually changed based on changes

40:48

in milestones in technology. And it would not be impossible

40:51

to put some of those things into place. and

40:54

thinking about and proposing these models for a long time.

40:56

First off, what does agile governance look

40:59

like? And what does adaptive regulations

41:00

mean? I don't know. I'm

41:03

not a law professor. I'm a podcast

41:05

host who's jealous of a circuit board

41:07

that gets to a watercolor. So I asked my

41:09

robot machine Google and agile governance

41:11

means a process that brings the most

41:14

value by focusing on what matters.

41:16

Okay, but adaptive regulations, I

41:18

think mean like, watch the space, keep

41:21

making laws if shit seems

41:23

like it's getting out of hand. Now in June,

41:25

the European Union overwhelmingly passed

41:27

the EU AI Act, which

41:30

classifies

41:30

different types of AI into risk

41:33

categories. There's unacceptable,

41:35

there's high risk, there's generative

41:37

AI and limited risk. What

41:39

is in these buckets? You're wondering. So

41:42

the unacceptable bucket includes cognitive

41:45

behavioral manipulation and

41:47

social scoring, a la black mirror

41:50

and biometric identification

41:52

like real time public facial recognition.

41:55

High risk involves more biometric

41:57

uses, but after the fact with a few.

42:00

exceptions for law enforcement, but it curbs

42:02

AI stitching on employees

42:05

and doing emotional spying from when

42:07

I gather. Generative AI would have to disclose

42:10

that it's generative and the makers need to come clean

42:12

on what copyrighted material they're using

42:15

to teach generative neural networks.

42:17

Now that's in the EU. As for America,

42:20

we have not gotten that far yet. I mean, that is

42:23

if everyone could even agree on what

42:25

needs to happen, then they'd have to

42:27

agree on voting for that thing

42:29

to

42:29

be actually enacted, which is,

42:32

it's a beautiful dream that

42:34

I'm generating with my human imagination.

42:37

The problem has been, I think, the political will to do

42:39

anything about it and to figure out like, why

42:41

should we care about the cognitive liberty

42:44

of individuals? Why should we care about

42:46

leisure and flourishing

42:49

of humanity? Let's just maximize productivity

42:52

and minimize human enjoyment in life.

42:55

That just can't be

42:56

what the answer is in the digital age anymore. I

42:58

mean, we need an updated understanding of what flourishing

43:01

means, and it can't be that it is generative

43:03

AI making art and writing poetry while

43:05

we toil away. That can't be

43:07

the case. I'm a philosopher. I'm

43:09

going to go back to, we have all

43:11

of these philosophical conceptions,

43:14

lots of perspectives on what flourishing

43:16

is. None of those perspectives,

43:18

if you go back and look at them, contemplated a

43:20

world in which our brains and mental experiences

43:23

could so easily be hacked and manipulated.

43:26

The idea of happiness being

43:29

the primary concept of human flourishing,

43:32

what is synthetic happiness? Is that really happiness?

43:34

If it's generated by dopamine

43:37

hits from being on a predictive

43:39

algorithm that's sending you little notifications,

43:41

it's just the right time to make your brain addicted and

43:43

staying in place, that looks like

43:45

happiness, but I don't think that's happiness,

43:47

right? Even that all of

43:49

these presupposed world in which we actually

43:52

had cognitive freedom,

43:53

we need to realize we don't anymore, right? If

43:56

we don't anymore, we need to create a

43:58

space in which we do so that human...

43:59

flourishing in the digital age is what we're actually

44:02

after and trying to make happen. That

44:04

we could put some human rights in place for

44:07

it. We could put some systems in place that were actually

44:09

creating incentives to maximize

44:12

cognitive freedom as the precursor

44:14

to all other forms of flourishing. And hopefully

44:16

that cognitive freedom would be the right to

44:19

create art without having it appropriated, the right

44:21

to write scripts and poetry without

44:24

having it used to train models

44:26

without our permission and without

44:29

us being part of it

44:29

that then make us irrelevant

44:32

so that the models can play while

44:34

we work. So in her book, The Battle for Your Brain,

44:36

Nita writes that we must establish the right to

44:38

cognitive liberty to protect our

44:41

freedom of thought and rumination,

44:43

mental privacy and self-determination

44:46

over our brains and mental

44:49

experiences. This is the bundle

44:51

of rights that makes up a new right to cognitive

44:53

liberty, which can and should be recognized

44:56

as part of the Universal Declaration of

44:58

Human Rights, which creates powerful

45:00

norms that guide corporations

45:03

and nations on the ethical use of neurotechnology.

45:06

Neurotechnology has an unprecedented power to either

45:09

empower or oppress us. The

45:12

choice is ours." And

45:15

one liberty I've taken is

45:18

never using chat GPT, kind

45:20

of like my high school's football

45:23

rallies. I just don't want to participate.

45:26

And I don't like what it's all about, even

45:28

though literally no one cares that

45:30

a stinky drama student with dyed black

45:32

hair and braces is boycotting. Nobody misses

45:35

me. I've always been a little

45:37

bit creeped out and hesitant. Like, I've never

45:39

tried chat GPT.

45:41

And I have this absolutely

45:44

incorrect illusion that if I don't

45:46

use chat GPT, it won't get

45:49

smarter. And therefore, I

45:51

single-handedly by abstaining have

45:53

somehow taken down an entire industry

45:55

of AI. It's not true. Well,

45:57

it's not true, but there is something to this idea.

45:59

that we're not helpless and that there

46:02

is a demand side to technology just

46:04

as there is a supply side to technology. And

46:06

there is a sense in which consumers and

46:08

individuals feel like they're helpless. It's

46:11

the same way you see with voting. Well, what's the point

46:13

of voting because my state always goes

46:15

this way or that way or... And that kind

46:17

of apathy means that a lot of times elections

46:19

are decided by everybody else and you

46:22

know that you don't have an effect. But this is even more

46:24

so like collectively if we don't like the terms

46:26

of service, why are we all still on

46:28

the platforms? And you're right, the

46:31

models are going to continue to be trained with or without you.

46:33

Yeah, no, like it's not that radical an act

46:35

from just me to abstain. Well,

46:38

but that idea that collectively we

46:40

could act differently, if we

46:43

could motivate and actually

46:45

work collectively to act differently, we

46:48

could act differently. One individual

46:50

person silently protesting

46:52

against chat GPT isn't going to do it, right?

46:55

But loudly protesting against it and saying

46:57

like, look, the models train

47:00

based on human interaction and the more

47:02

human interaction there is, the more it is trained. And

47:05

so do you want to continue to feed

47:08

into that model? That's a worthwhile societal conversation

47:10

to have. You know, I was talking to my husband

47:12

this morning about how many brilliant engineers

47:16

end up working for bomb companies because they're

47:18

going to have the best benefits,

47:21

they're going to have the most stable employment, right?

47:24

How many people in the legal field

47:26

do you feel like get kind of scooped up by

47:29

tech companies because it's just an easier

47:31

way to live?

47:32

Do tech companies just have more pull

47:35

to get the

47:36

best lawyers to advocate for

47:38

them instead of for say greater

47:40

humanity? I think it's not just law,

47:43

right? If I look at some of the best tech ethicists,

47:45

many of them have gone in house to a lot of companies

47:48

that are not actually that invested

47:50

in tech ethics. And many of them got

47:52

laid off in the major tech layoffs that

47:54

have happened from 2022 to 2023. Because a lot of tech companies,

47:56

I think, have

47:59

put lip service to being serious

48:02

about ethics, but they haven't as seriously

48:04

grappled with it. And

48:06

the money and the power that these

48:08

corporations have and the influence on society

48:11

they have, I think both

48:13

makes it hard for some people to resist saying no,

48:15

but also this idea that like

48:17

if you're at a tech company where the transformation

48:20

of humanity is happening, maybe you can steer

48:22

it in the ways that you think are better for

48:24

humanity.

48:25

Are there any nonprofits or organizations that

48:27

you feel like are doing a good job? There are a lot. I

48:29

mean, I couldn't even begin to name them all. Like

48:32

I would say, first,

48:35

I admire what UNESCO is

48:37

doing. So UNESCO is the United

48:39

Nations Educational, Scientific, and Cultural

48:42

Organization. And on their Ethics

48:44

of Artificial Intelligence webpage, it

48:46

states, UNESCO has delivered global standards

48:49

to maximize the benefits of scientific

48:51

discoveries while minimizing the

48:54

downside risks,

48:55

ensuring they contribute to a more inclusive,

48:58

sustainable, and peaceful world. And

49:00

it's also identified challenges

49:02

in the ethics of neurotechnology. So

49:04

as a result, their recommendation on

49:06

the ethics of artificial intelligence was

49:09

adopted by 193 member

49:11

states at UNESCO's General Conference

49:14

way back in the olden times

49:16

of November 2021.

49:18

They're really trying to get out ahead of a lot of issues

49:21

and to thoughtfully provide a lot of ethical

49:23

guidance on a lot of different issues.

49:26

I think the OECD is trying to be

49:29

a useful and balanced

49:32

organization to bring important information

49:34

to bear. The OECD, I had to look this

49:36

up, is the Organization for Economic Cooperation

49:39

and Development. And it's headquartered in France,

49:41

but involves 38 countries. So

49:44

what are they doing? The OECD principles

49:46

on artificial intelligence promote AI

49:48

that's

49:48

innovative and trustworthy and that respects

49:51

human rights and democratic values.

49:53

And then of course there's the EU. I think the EU

49:56

is acting in ways that are really pushing

49:58

the conversations forward around

49:59

the regulation of AI and how to

50:02

do it and how to respect everything from

50:04

mental privacy to safeguard against manipulation.

50:06

And, you know, they get lambasted for like

50:08

going too far or not going far

50:10

enough. And those conversations are better than

50:13

putting nothing on the table, which is what's happening a lot

50:15

of times in the US. I think the Biden

50:17

administration has put out a lot of different principles

50:20

that have been helpful and that those kinds of principles

50:22

are things around like an AI Bill of Rights.

50:25

I went and took a gander at this doc and the

50:27

blueprint for an AI Bill of Rights

50:29

sets forth five principles, which I will

50:32

now read to you. You should be protected

50:34

from unsafe or ineffective systems.

50:37

You should not face discrimination by algorithms.

50:40

You should be protected from abusive

50:42

data practices and you should have agency over

50:44

how data about you is used. You

50:47

should know that an automated system

50:49

is being used and understand how and why

50:51

it contributes to outcomes that

50:53

impact you. And finally,

50:55

you should be able to opt out where appropriate

50:58

and have access to a person who

50:59

can quickly consider and remedy

51:02

problems you encounter. I don't know if that means

51:04

a helpline. I have no idea. But that

51:06

five point framework is accompanied

51:08

by a handbook called From Principles

51:10

to Practice and its guidance for

51:13

anyone who wants to incorporate those protections

51:15

into policy. So that's what the White House

51:17

has put out. They're like, y'all, we should

51:19

really like be cool and nice about

51:21

all this. And it's so sweet and I appreciate

51:24

it. My grandma had 11 children

51:27

and really just dozens of grandkids and she

51:29

still remembered

51:29

all her birthdays and would send a letter

51:32

with one dollar in it. And that dollar meant

51:35

a lot, even if it didn't get you far

51:37

in the world. But I appreciated it in

51:39

the same way I appreciate that AI

51:42

Bill of Rights. It's very sweet.

51:44

Don't know what to do with that.

51:46

There's a lot of different people coming at the problem from a lot of

51:48

different perspectives. If anything,

51:51

there are so many voices at the table that it's

51:53

in many ways becoming noisy where

51:55

we're not necessarily like moving ahead in

51:57

a really constructive or productive way.

51:59

of replication of efforts. But that's

52:02

better than having too little activity at the table.

52:04

So yeah. I think that a lot

52:06

of us on the outside of it think there's a tumbleweed

52:08

blowing through

52:09

a boardroom, and nobody cares.

52:11

So it's really good to hear. No, I

52:13

will tell you that I just feel like there are conversations

52:15

happening in every corner you can imagine right

52:18

now. And I'd like to see those

52:20

conversations be turned into useful

52:22

and practical pathways forward,

52:25

like calling for governance if you're

52:27

a major tech company and saying,

52:30

these technologies that I'm creating create

52:32

existential risk for humanity. Please regulate

52:35

it. Or if you think that they present existential

52:37

risk for humanity, don't just

52:39

rush ahead and come forward with

52:42

something positive rather than saying, my

52:44

job is just to create the technology. Your job is

52:46

to govern it. That's not the pathway forward

52:48

either. I have questions from listeners who

52:50

know you're coming on. Oh, great. Yeah, please.

52:53

But before we do, we'll donate to a relevant

52:55

cause. And this week, it's going to Human

52:57

Rights Watch, which is a group of experts, lawyers,

53:00

and journalists who investigate and

53:02

report on abuses happening in all corners

53:05

of the world. And then they direct their advocacy

53:07

toward governments, armed groups, and

53:09

businesses. And you can find out more at hrw.org.

53:13

And we will link that in the show notes. And thanks to

53:15

sponsors of the show who make that donation possible. Listen,

53:18

there's a lot of reasons to learn a new language.

53:21

Perhaps it makes you more competitive as a job

53:23

applicant. Maybe you're flirting with a stranger

53:25

you otherwise wouldn't have a chance with. Rosetta

53:28

Stone, they're there for you. They're available on desktop.

53:30

They can also be used as an app on your phone.

53:33

And you can choose from 25 languages, including

53:36

Spanish, and French, and Dutch, and Arabic. And

53:38

instead of memorizing and drilling a bunch of vocab

53:40

words, honk shoo, Rosetta Stone

53:43

teaches through immersion, reading stories,

53:45

and participating in dialogues. And it also

53:48

keeps it really interesting.

53:48

I use Rosetta Stone. It's been used by

53:50

millions. And it's been around for 30 years

53:53

because it works. I'm learning to speak Spanish. And

53:55

my neighbor Pablo says that my accent is

53:57

muy bueno. They also have a true accent feature. so

54:00

you get feedback on how well you're pronouncing words in case

54:02

you don't have Pablo to launch a compliment at

54:04

you. You can find lessons as short as 10 minutes

54:06

and you can start thinking in more languages.

54:09

How exciting is that? And I know you might be saying,

54:11

which language would I even choose? Well, start

54:13

learning that new language right now for a very

54:16

limited time. All of these listeners can get Rosetta

54:18

Stone's lifetime membership for 40% off.

54:21

So that's $179 for unlimited access to 25

54:26

language courses for the rest of your

54:28

life. Redeem your 40%

54:29

off at rosettastone.com

54:32

slash ologies today. That's rosettastone.com

54:36

slash ologies. Go for it.

54:38

Summer's here and no one expects you to remember

54:40

everything. Listen, maybe you're hosting a summer

54:42

cookout and then you realize, no marshmallows?

54:45

Do you cancel the party? No, use Instacart.

54:48

Maybe you're at a sunscreen. You can have that stuff

54:50

delivered in as fast as an hour. So with Instacart,

54:52

you can get everything you need for summer in

54:55

one app and you can check out with PayPal using

54:57

one login. So easy. Instacart

54:59

helps deliver the order in as fast

55:01

as an hour, which is giving time back to

55:04

enjoy summer. Hassle-free. And even

55:06

if you're hosting or you need large orders, it's easy,

55:08

you can just stock up and you can save on 1200 retailer

55:10

brands, 75,000 locations

55:13

across the country. I just used Instacart this

55:16

week. Not because I was hosting a party or having

55:18

fun, but I had pneumonia. And guess what? I

55:20

wasn't getting out of bed to go get soup. Nope, Instacart's

55:22

like, here you go. And I was like, ding dong, thank you. Instacart,

55:25

add summer to cart. Visit instacart.com

55:29

to get free delivery on your first three

55:31

orders. New Instacart customers also

55:33

get $25 off your order of $75 with

55:36

code PayPalNew25

55:38

when using PayPal through September

55:40

30th, 2023. Additional terms

55:42

apply. Have a summer, get your stuff.

55:46

This show is sponsored by BetterHelp. Sometimes

55:49

the most exciting time of our life is

55:51

when we're making a big decision, even though

55:53

it can be really scary, whether you're thinking about a career

55:55

change or feeling like your relationship

55:57

needs some TLC or you're going through grief.

55:59

or a crossroads, whatever it is, therapy

56:02

can help you map out the future and help

56:04

you understand why you're gravitating

56:06

towards certain decisions. So if you're thinking

56:08

of starting therapy, give BetterHelp a try. It's

56:10

entirely online, it's designed to be convenient,

56:13

flexible, and suited to your schedule. You

56:15

just fill out a brief questionnaire and they'll

56:17

match you with a licensed therapist. It's

56:19

quick, it's easy, if for any reason you're

56:22

not vibing, you can switch therapists anytime

56:24

for no additional charge, no drama. Everyone

56:27

needs a little help sometimes, and those big

56:29

exciting moments in life will change

56:31

where you go in the future. So let therapy be

56:34

your map with BetterHelp. Visit betterhelp.com

56:36

slash ologies today and get 10% off

56:39

your first month. That's betterhelp, H-E-L-P

56:42

dot com slash ologies.

56:44

Here's the deal, when it comes to money, sometimes you're like, I'm

56:46

just not going to think about it. And then other times you're like, I would like

56:49

someone else to think about it for me. Hence,

56:51

rocket money. If you don't know exactly how much you're

56:53

spending every month, you need rocket money. Do

56:56

you know how much your subscriptions really cost?

56:58

Most Americans think they spend around 80 bucks

57:00

a month on subscriptions. The total is

57:02

actually closer to 200. 80% of people

57:05

have subscriptions that they completely forgot about.

57:07

Chances are you're one of them. Rocket money

57:10

quickly and easily finds your

57:12

subscriptions for you. And then for any,

57:14

you don't want to pay anymore. You just hit cancel. Boom,

57:17

rocket money, gone. It's that

57:19

easy. Rocket money also helps you manage

57:21

all your finances in one place, and

57:23

it automatically categorizes your expenses

57:26

so you can track your budget in real time.

57:28

Super helpful. And it'll alert you if

57:31

anything looks off. Maybe I just bought a sweater

57:33

that cost $80 and it has a worm

57:35

on it. That was all me. Over 3 million

57:38

people have used rocket money, saving

57:40

the average person up to $720 a year. That's

57:42

a lot of worm sweaters. So that's rocket money.com

57:47

slash ologies. Rocket money.com

57:50

slash ologies. You can do this. Well, they

57:52

can do it for you.

57:54

Okay, on to questions written

57:56

by actual human listeners made of

57:58

meat and water. Let's start

58:00

with something optimistic. A ton of people,

58:03

Lena Brodsky, Nina Yvesy, Chris Blackthorn,

58:06

Meg C, Alexandra Kautoule, Adam

58:08

Silk, Katie McAfee, Madison

58:10

Piper, and Will Mac, want to know,

58:13

can we use AI for good? Ry

58:16

of the Tiger wants to know, what will AI's role

58:18

look like in the fight against climate change,

58:21

for example, or should we be using

58:23

AI for the toils like meal

58:25

planning and trip planning and things like that?

58:27

Yeah, so I think we can absolutely

58:30

use AI for good. And first,

58:32

I would say a friend of mine, Orly Lopel, wrote

58:35

a book recently called The Equality

58:37

Machine. And it's all about using

58:40

AI to achieve better

58:42

equality in society, and gives kind of example

58:44

after example of both how it could be done, and

58:47

how it is being done in some context. I

58:49

think recognizing that there is this

58:52

terrifying narrative about AI, but that actually

58:54

AI is already making our lives better,

58:57

in many, many ways, is an important thing

58:59

to look at. And that

59:02

we can put it to solving some of the biggest problems

59:04

in

59:05

society, right from climate

59:07

change and trying to generate novel

59:10

ideas to testing, and

59:12

identifying, and this is already happening, novel

59:15

compounds that could be used to solve

59:17

some of the worst diseases, to being

59:20

used to identify the causes

59:22

of different diseases, to

59:24

identifying better patterns that help

59:26

us to address everything from neurological

59:29

disease and suffering to the existential

59:32

threats to humanity like climate change. So I absolutely

59:34

think it can be used for good. It

59:36

is being used for good. It could be used for

59:39

more good. We have to better align

59:41

the tech companies with the

59:44

overall ways of human

59:46

flourishing. Right, I mean, if you were to

59:49

use AI to improve brain

59:51

health, instead of to addict and diminish

59:53

brains, that would be phenomenal.

59:56

And it could be used to do that. It can be used

59:58

for mental health treatment and for solving neuro- neurological disease

1:00:00

and suffering, or it can be used to addict

1:00:02

people and keep them stuck on technology. We

1:00:05

need to figure out a way to align the incentives

1:00:07

of tech companies with these

1:00:09

ideas of AI for good. It'll

1:00:11

be so interesting to see if

1:00:14

they are getting a lot of feedback

1:00:16

from our brains, any

1:00:18

mental health challenges or speaking

1:00:21

as someone who has anxiety and

1:00:23

is neurodivergent. Hello, hi. It's

1:00:26

like ADHD, autism, those have been so

1:00:28

overlooked in some populations.

1:00:31

It would be interesting to see people

1:00:34

getting a better understanding of their own brains that maybe

1:00:36

medicine has overlooked because of demographics

1:00:39

for a long time. Yeah. I have a TED Talk that

1:00:41

just came out that the first half

1:00:43

of the TED Talk actually focuses on all of the

1:00:46

positive ways that neurotechnology can be used

1:00:48

and all of the hope that it offers. Stress

1:00:51

tracking our everyday brain activity could

1:00:54

help us better understand what stresses us out.

1:00:57

The earliest stages of glioblastoma,

1:00:59

the worst and most threatening form of aggressive

1:01:02

brain cancer is, the earliest stages of

1:01:04

Parkinson's and Alzheimer's disease, better

1:01:06

solutions for ADHD and trauma,

1:01:10

everything from understanding the impact

1:01:12

of technology on our brains to the understanding

1:01:15

the impact of having that glass of wine or that cup of coffee

1:01:17

on the brain and how it reacts to it. Gaining

1:01:21

insight into our own brain activity could

1:01:23

be the key to unlocking much

1:01:26

better mental health and well-being.

1:01:29

I think if it's put in the hands of individuals

1:01:32

and used to empower them, that

1:01:34

will be tremendous and phenomenal. Long

1:01:37

as we don't overshadow those benefits or

1:01:39

outweigh those benefits with the dystopian misuses

1:01:42

of the technology, which are

1:01:44

very real and very possible, right, of

1:01:46

using in the same way that companies

1:01:48

are using all kinds of algorithms

1:01:51

to predict our purchasing behavior

1:01:53

or to nudge us to do things like watch

1:01:56

the tenth episode in a row of

1:01:58

a show rather than

1:01:59

you know, breaking free and getting some

1:02:02

sleep, which is important for brain health. If

1:02:04

the companies don't use brain data

1:02:06

to

1:02:07

commodify it, to

1:02:09

inform a more Orwellian

1:02:12

workplace, get back to work. If

1:02:14

governments don't use it to try to surveil

1:02:16

brains and to intrude on freedom of thought,

1:02:19

but instead it's used by individuals

1:02:21

to have greater power over their own

1:02:23

health and well-being and their own brains, it

1:02:26

will be tremendous. You just have to really

1:02:28

worry about those misuses and how we safeguard

1:02:30

against them. So the day before this interview,

1:02:32

a TED talk featuring Nita went

1:02:35

live and in it she discusses

1:02:37

the loss of her daughter and the

1:02:39

grief that overwhelmed her. And she tells

1:02:41

of how using biofeedback to

1:02:44

understand her own sorrow

1:02:46

and trauma from the experience helped

1:02:49

her so much, but how individuals

1:02:51

brain

1:02:51

data should be protected. And

1:02:53

this wrenching personal story that she tells,

1:02:56

plus her long backgrounds in ethics

1:02:58

and science and philosophy make her very uniquely

1:03:00

suited to see this issue from a lot of angles.

1:03:03

And a lot of patrons had questions about

1:03:05

surveillance and brain data and even

1:03:08

neural hardware, including Katie

1:03:10

McAfee, Ryan Marlowe, and Sandy Green,

1:03:12

who asked about things like

1:03:14

medical devices like brain implants

1:03:16

being used for surveilling or for

1:03:19

commerce. I was curious, so are some

1:03:21

listeners to

1:03:21

PAVCA34, Aminek, David,

1:03:24

and Alex Ertmann's words, if

1:03:26

we were to implant chips into human brains, what

1:03:28

would they most likely be capable of? Would

1:03:31

they be more in the realm of modulating real inputs

1:03:34

or would they be capable of generating

1:03:37

new thoughts? Alex says it seems

1:03:39

far-fetched, but also the truth can be stranger than fiction.

1:03:41

So is that a really big leap

1:03:44

philosophically and legally and

1:03:46

technologically? I think it might be easier

1:03:48

to interrupt thoughts than to create new thoughts.

1:03:51

However, I guess philosophically that

1:03:53

is creating new thoughts if you're interrupting thoughts, right,

1:03:55

because you're letting other thoughts happen. But implanted

1:03:57

neurotechnology right

1:03:59

now is...

1:03:59

very limited. It's very difficult

1:04:02

to get neurotechnology into

1:04:04

people's brains. And there are 40 people who

1:04:07

are part of clinical trials that

1:04:09

have implanted neurotechnology right now. It's a

1:04:11

tiny number of people. If Neuralink,

1:04:14

you know, and Elon Musk has his way, there will be far

1:04:16

more people who are part of that. But

1:04:19

implanted neurotechnology is limited. What

1:04:21

it primarily is being used to do

1:04:23

is to get signals out of the brain. That

1:04:26

is to listen to intention to

1:04:28

move or to form speech

1:04:29

and to translate that

1:04:32

in ways that then can be used to operate other

1:04:34

technology. If you're like, what is Neuralink

1:04:36

again? It sounds like a commuter

1:04:38

train, but this is actually a side hustle of

1:04:41

Twitter owner and Tesla

1:04:43

guy and tunnel maker Elon

1:04:45

Musk. And he described this

1:04:48

cosmetically undetectable coin

1:04:50

sized brain accessory as a wireless

1:04:53

implanted chip that would enable

1:04:55

someone who is quadriplegic

1:04:58

or tetraplegic to control

1:04:59

a computer or mouse

1:05:02

or their phone or really any device just by

1:05:04

thinking. And he likened it to a Fitbit

1:05:06

in your skull with tiny wires that

1:05:09

go to your brain. So a

1:05:11

robot surgeon also invented by Neuralink,

1:05:14

so 64 threats with over

1:05:16

a thousand electrodes into

1:05:19

the brain matter, which allows the

1:05:21

recipient to control devices

1:05:23

or robotic arms or screen using

1:05:26

telepathic typing, which sounds pretty cool.

1:05:28

In early 2022, it

1:05:29

came to light that roughly 1500 animals

1:05:33

had been killed in the testing process since 2018.

1:05:36

Some from human errors like incorrect

1:05:38

placement on pig spines or

1:05:41

wrong surgical glue used

1:05:43

in primate test subjects. And some former

1:05:45

employees reported that the work there was

1:05:47

often rushed and that the vibe was

1:05:49

just high key stressful.

1:05:52

But nevertheless, Neuralink announced

1:05:54

just a few months ago that they got the green

1:05:56

light from the FDA to launch their human

1:06:00

you're like, hey, I am

1:06:02

always losing the TV remote, so wire

1:06:04

me up, Musk. Please cool your jets, because

1:06:06

they added that recruitment is not yet open

1:06:09

for their first

1:06:09

clinical trial. More on that as it develops.

1:06:12

But I guess when I said that we could become bubbles

1:06:14

of chimp, that was really on the optimistic

1:06:16

side of things. What is possible, though,

1:06:18

and this is one of the things I talk about in my

1:06:20

TED Talk, is it's possible to use

1:06:23

neurostimulation in the brain. So I

1:06:25

describe, for example, the case of Sarah,

1:06:27

where she had intractable depression,

1:06:31

and through the use of implanted electrodes,

1:06:34

was able to reset her brain activity.

1:06:36

This side note was conducted at the University

1:06:38

of California at San Francisco, where neuroscientists

1:06:41

implanted what's called a BCI, or

1:06:43

brain computer interface, which was initially

1:06:45

developed for epilepsy patients into

1:06:48

someone with treatment-resistant depression. And

1:06:50

one surgeon on the team said, when we turned

1:06:53

this treatment on, are patients' depression symptoms

1:06:55

dissolved in a remarkably small

1:06:58

time she went into remission? And

1:07:00

the patient, Sarah, reported laughing

1:07:02

and having a joyous feeling wash over

1:07:04

her that lasted at least a year after

1:07:06

this implantation. So the specific

1:07:09

pattern of neural activity

1:07:11

that was happening when she was the most symptomatic

1:07:13

was traced using the implanted technology.

1:07:16

And then, like a pacemaker for the brain, those

1:07:18

signals were interrupted and reset

1:07:21

each time she was experiencing them. That

1:07:24

doesn't create a new thought. What

1:07:26

it does is interrupt an existing thought.

1:07:28

But philosophically, you could say that creates a new thought. It

1:07:30

creates, for her, an experience of being

1:07:32

able to have a more typical range of emotions. I

1:07:35

think specific thoughts would

1:07:38

be very hard to encode into the

1:07:40

brain. I won't say

1:07:41

never. So brain hacking

1:07:43

and hacking into your brain may

1:07:46

radically change the way that we think

1:07:48

and feel

1:07:49

if we don't blow up the planet first, which

1:07:51

is not an intelligent thing to do. Speaking

1:07:54

of intelligence, many patrons wanted to know

1:07:56

what is in a name. Alexis Will-Clark,

1:07:59

Zomba.

1:07:59

who proposed the term OI

1:08:02

or organic intelligence for human thinking

1:08:05

and history buff Connie Brooks, they

1:08:07

all had questions about

1:08:09

AI and the term AI. Is

1:08:12

it intelligent? Is it artificial?

1:08:15

Are they ever going to do a rebrand on

1:08:17

that? Does it give people the wrong idea of

1:08:19

what it is? Yeah. So I mean, a lot of

1:08:21

the technologists out there were computer

1:08:24

scientists saying, this isn't artificial

1:08:26

intelligence because that assumes that there's intelligence.

1:08:28

These aren't intelligence.

1:08:30

They are task specific algorithms

1:08:32

that are designed to do particular things. And

1:08:35

that if we ever get to the point where you start to see

1:08:37

more generalized intelligence, then

1:08:40

that's the point at which it makes more sense to talk about

1:08:42

artificial intelligence. But not everyone

1:08:44

is so casual

1:08:45

about that assessment.

1:08:47

Interestingly, Eric Corvitz, who is

1:08:49

the chief science officer at Microsoft,

1:08:52

who has partnered with OpenAI for chat

1:08:55

GPT, he just published his

1:08:57

essay on this AI and anthology series.

1:08:59

And he talks about how his

1:09:02

experience with GPT-4 was to see a lot of threads of

1:09:07

intelligence, of what we think of as intelligence.

1:09:10

And you see increasingly a lot of

1:09:13

examples of reasoning more

1:09:15

like humans. I think one of the examples I've

1:09:17

seen out there is giving GPT-4

1:09:19

a question of like, okay,

1:09:22

you have some eggs, a laptop,

1:09:24

it's like five or six items, how would you stack

1:09:27

them? And then comes

1:09:29

out and explains how you would stack them. And like you would

1:09:31

put the book on the bottom, and then you

1:09:34

would put a set of eggs that were spread out so

1:09:36

that they could be stable. And then you would put the

1:09:38

laptop in a particular configuration and

1:09:40

blah, blah, blah. And why that

1:09:42

kind of reasoning was

1:09:45

more like human intelligence

1:09:47

than it is like an algorithm. And those are

1:09:49

really interesting to think about. Like what is intelligence

1:09:52

is really the fundamental question, I think when somebody

1:09:54

is saying, is it really artificial intelligence? It

1:09:57

is to have a particular perspective on what intelligence

1:09:59

is.

1:09:59

is and means, and then to say,

1:10:02

well, that isn't intelligence. Or

1:10:04

if a

1:10:06

generative AI model says it's happy,

1:10:08

that it can't really be because that's

1:10:10

not an authentic emotion because it's never

1:10:13

experienced the world and it doesn't have sensory

1:10:15

input and sensory output. Or

1:10:17

if a generative AI model

1:10:20

says, here's what the ratings of wine are

1:10:22

and what an excellent wine is, it can't possibly

1:10:24

know because it's never tasted wine. And

1:10:27

then there's a question of, is that kind of intelligence

1:10:29

what you need, which is experiential

1:10:32

knowledge and not just knowledge built

1:10:34

on knowledge. There are some forms

1:10:37

of intelligence, like emotional intelligence, which you might

1:10:39

think really requires experiencing

1:10:41

the world to authentically have that kind of intelligence.

1:10:45

I don't know shit about wine, and sometimes

1:10:47

I'm bad at my own emotions. Oh, well, we

1:10:49

can learn. Speaking of learning, many

1:10:51

patrons who are students had thoughts

1:10:54

and questions like Handy Dandy Mr.

1:10:56

Mandy, Natalie Jones, Josie Chase,

1:10:58

and Slayer, as well as educators,

1:11:00

including Nina Bratzke, Julie Vollmer,

1:11:02

Leah Anderson, Jenna Cong-Ben Theodorovician,

1:11:05

Hudson Ansley, and Nina Yvesi. There

1:11:08

were several teachers who wrote in with

1:11:10

questions. Katie Bauer says, I'm a middle school

1:11:12

teacher and I just started having students

1:11:14

use AI

1:11:15

tools to write essays for them. Help,

1:11:17

talk me down. How do we embrace

1:11:19

new tech but also teach students how to navigate

1:11:21

this new landscape with solid ethics

1:11:24

and an understanding of the need to develop skills

1:11:26

that don't revolve around AI technology? And

1:11:29

Liz Park, first time question asker, they're a teacher

1:11:31

and they feel that teaching along with a lot of other jobs

1:11:34

just can't be handed off to AI and expected

1:11:36

to have the same impact because machines, no matter

1:11:38

how advanced, won't be able to individualize

1:11:41

education and provide warmth and et cetera.

1:11:43

Well,

1:11:43

you know, it's funny because I hear that almost

1:11:45

the same question in both, right? What

1:11:48

is the role of education and human

1:11:50

to human education in a world

1:11:52

of generative AI? And I

1:11:55

think that's a great question to be asking. And I would

1:11:57

say first, I'm so glad

1:11:59

that they were giving. their students the assignment

1:12:01

of working with Chat GPT and trying

1:12:03

to understand it because I think there

1:12:06

are skills that you can't

1:12:08

learn from generative AI and if you don't learn

1:12:10

them we will not be able to interact

1:12:13

well with them and use them well. And these are critical

1:12:15

thinking skills and if the same old

1:12:17

assignments are how we're trying to teach students

1:12:20

then yeah students are just going to go

1:12:22

to Chat GPT and say here's the

1:12:25

book generate a

1:12:27

thesis statement for me and write my essay

1:12:30

right but they will have lost out

1:12:32

on the ability to generate a thesis statement

1:12:34

and what that critical thinking skill is and lost

1:12:37

out on the ability to build an argument in how you

1:12:39

do so lost out on the ability

1:12:41

to write and understand what good writing is

1:12:44

and they won't be able to interrogate the systems

1:12:46

well because they won't have any of the skills necessary

1:12:49

to be able to tell fact from fiction and what is good

1:12:51

writing or anything else. So then the question

1:12:53

is what do you do and

1:12:55

it's the

1:12:56

teachers and higher education and K

1:12:59

through 12 education needs to be thinking

1:13:01

about okay what are the fundamental

1:13:03

skills of reasoning and

1:13:05

critical thinking and empathy

1:13:08

and emotional intelligence and

1:13:10

mental agility that we

1:13:12

think are essential and that we have been

1:13:14

teaching all along but we've been teaching by

1:13:18

tasks that now can be outsourced and

1:13:20

then how do we shift our teaching

1:13:22

to be able to teach those skills

1:13:25

and you know if you go back

1:13:26

to like the Socratic dialogues

1:13:29

there's an art to asking the question to seeking

1:13:32

truth and there is an art to asking the question

1:13:34

of generative AI models in

1:13:36

seeking the truth or in seeking good outputs

1:13:38

and we have to be teaching those skills if we

1:13:40

want to move ahead. I wasn't sure

1:13:43

what the Socratic method of questioning was

1:13:46

so I asked the literature via

1:13:48

computer and I found that it involves

1:13:50

a series of focused yet open questions

1:13:53

meant to unravel thoughts as you go

1:13:55

and according to one

1:13:56

article instead of a wise person

1:13:59

lecturing

1:13:59

teacher acts as though ignorant of

1:14:02

the subject. And one quote

1:14:04

attributed to Socrates reads, the highest

1:14:06

form of human excellence is to question

1:14:09

oneself and others. So don't

1:14:11

trust my wine recommendations, but do cut

1:14:14

bangs if you want. Text a crush, ask

1:14:17

a smart person, a not smart

1:14:19

question, because worms are

1:14:21

going to eat us all one day. But yeah,

1:14:23

the point of education isn't to

1:14:26

get a good grade, but to develop skills

1:14:28

that in the future are going to

1:14:29

get you out of jam. So many

1:14:32

jams.

1:14:33

And I think your other person

1:14:35

talking about that they can never replace human

1:14:37

empathy, that's right. But

1:14:40

don't be blind to the fact

1:14:42

that they can make very powerful personal

1:14:44

tutors as well. And they may not

1:14:46

be able to tell when a student is struggling or

1:14:48

when they need emotional support or when they may

1:14:51

be experiencing abuse at home and need

1:14:54

the support of the school to be able to intervene,

1:14:56

for example. But they can

1:14:58

go beyond, a teacher

1:15:00

can go, a teacher doesn't have the capability

1:15:02

to sit down with every student for hours

1:15:05

and help them work through 10 different

1:15:07

ways of explaining the same issue

1:15:09

to somebody. And so you help them

1:15:11

learn how to ask the questions. And

1:15:13

then they could spend all night long saying, okay, well, I didn't understand

1:15:16

that explanation. Can you try explaining it to me a different

1:15:18

way? Can you try explaining it to me as if you

1:15:20

were telling my grandmother, I don't understand what that word

1:15:22

means. There's no teacher on earth

1:15:24

who has either the patience for that or

1:15:27

the time or is paid well enough

1:15:29

to do that for every student. And so I

1:15:31

think it can be an extraordinary equalizer,

1:15:34

you know, right now, like wealthier parents

1:15:37

are able to give private tutors to their kids.

1:15:39

Okay, now you can have a generative AI model

1:15:42

serve as a private tutor that can be customized

1:15:44

to every student based on how they learn. However,

1:15:47

that doesn't mean we don't need teachers to be

1:15:49

able to be empathetic and to help students

1:15:51

learn how to engage with the models

1:15:54

and learn critical thinking skills or to create

1:15:56

a social environment to help develop their emotional

1:15:58

intelligence and their digital intelligence.

1:15:59

intelligence.

1:16:01

But it does mean that there is

1:16:03

this additional tool that could actually be incredibly

1:16:05

beneficial and can augment how we're

1:16:07

teaching. Okay, but outside the classroom

1:16:10

and into your screens, folks had questions

1:16:12

including Michael Heiker, Kevin Glover,

1:16:14

Andrea Devlin, Jenna Congdon, credit

1:16:16

state of mind, Chris Blackthorn, RJ Dorridge,

1:16:19

and... One big question a lot of listeners

1:16:22

had is

1:16:23

Rebecca Newport says, what's your favorite or

1:16:25

least favorite portrayal of AI in media?

1:16:28

Chris Whitman wants to know, what is your favorite

1:16:30

AI storyline based movie and why is it

1:16:32

Ex Machina? Someone else said, Mrs. Davis, should

1:16:35

we turn off Mrs. Davis if we could? How do

1:16:38

we prevent Terminator 2? Whether or not

1:16:40

you watch Black Mirror, anything

1:16:43

that you feel like pop culturally

1:16:46

written by humans that you've loved

1:16:48

or hated? I love Minority Report.

1:16:50

It's an oldie but goodie, but it really

1:16:52

informs a lot

1:16:53

of my work and I think

1:16:55

it's great. I'm placing you under

1:16:57

arrest for the future murder of Sarah Marks. Give it a man his head.

1:17:01

The future can be seen.

1:17:03

I think that some

1:17:06

of the modern shows

1:17:08

that I like like Severance, Altered

1:17:11

Carbon, I thought was a great series, Black

1:17:14

Mirror, Yes. You know, all of those

1:17:16

I think are terrific and creepy.

1:17:18

I appreciate those stories and really

1:17:20

raising consciousness about some of the existential threats

1:17:23

but I would like to see stories that

1:17:26

give us a more balanced perspective sometimes. I

1:17:28

guess that doesn't make for good film but you know, look,

1:17:31

the fears of like we don't fully

1:17:33

understand consciousness,

1:17:33

let alone how emergent

1:17:37

properties of the human brain happen,

1:17:40

let alone how emergent properties could happen

1:17:42

in an incredibly intelligent system

1:17:45

that we are creating. I

1:17:47

share those fears. Like I don't

1:17:49

know where all of this is going and I worry about

1:17:51

it. I also

1:17:53

don't think anybody has an answer about how to safeguard

1:17:56

against those existential threats

1:17:58

and we should be doing things better.

1:17:59

to try to identify

1:18:02

them and to identify the points and identify

1:18:05

what the solutions would be if we actually start

1:18:07

to see those emergent properties and those emergent properties

1:18:09

are threatening. Like we need monitoring systems.

1:18:12

We also in the meantime need to

1:18:14

be looking at the good and figuring out

1:18:16

how to better distribute the good, how

1:18:18

to better educate people, how to change our education

1:18:21

systems to catch up with it, how to

1:18:23

recognize that the right answer for the writer's strike

1:18:26

isn't to outsource it to chat GPT and there's something

1:18:28

uniquely human about the

1:18:29

writing of stories and

1:18:32

the sharing of stories and the creation of art

1:18:34

and that that's part of the beauty of what it means

1:18:36

to be human. And so those

1:18:38

conversations about the role in

1:18:41

our lives and how to put it to uses

1:18:43

that are good and still preserve human

1:18:46

flourishing, like that I feel like is what we need to be doing

1:18:49

in the meantime, right before it actually torches

1:18:52

us all. That is great advice. And

1:18:54

the last questions I always ask is always like, what's

1:18:56

the worst part about your job? A lot of people say

1:18:58

might

1:18:59

be jet lag meetings, emails, but

1:19:01

I will outsource that to the patrons who wanted to

1:19:03

know, are we fucked? So many wanted

1:19:05

to know, are we fucked? So what is the most

1:19:08

fucked thing about what

1:19:10

you do or learn? So

1:19:12

I mean, we're fucked if we let

1:19:14

ourselves be and I

1:19:18

fear that we will, right? I mean, so

1:19:21

I can tell people until I turn blue

1:19:23

in the face about the potential

1:19:25

promise of AI and certainly the

1:19:27

promise of neurotechnology, if

1:19:29

we put it to good use and if we safeguard

1:19:31

against the Orwellian misuses of it in

1:19:33

society. But like we seem

1:19:35

to always go there. We seem to always like

1:19:37

go to the Orwellian and do the worst thing

1:19:40

and put it to the worst applications and be driven

1:19:42

just by profit and not by human flourishing.

1:19:44

And so if we keep doing that, then yeah,

1:19:48

we're kind of fucked. And if

1:19:50

we actually like heed the wake up call

1:19:52

and do something about it, like put into

1:19:54

place not only a human

1:19:57

right to cognitive liberty, but also

1:19:59

the system.

1:19:59

the governance, the practices,

1:20:02

the technologies that help cultivate it

1:20:04

in society. I mean, if we invest

1:20:07

in that,

1:20:07

we have a bright and happy future ahead.

1:20:10

If we don't, you know, it's not good. Yeah,

1:20:12

we're fine. We're talking. What

1:20:15

about, to be such a globally

1:20:17

recognized,

1:20:19

trusted voice on this,

1:20:21

obviously, I was so pumped to interview

1:20:23

you. Like, I came straight out of the gate

1:20:25

being like, I'm terrified of talking to you. What

1:20:29

is it about your work that gets

1:20:31

you excited? What kind of keeps you motivated?

1:20:34

I guess I'm also fascinated and terrified,

1:20:37

right? I mean, so like,

1:20:39

it's almost like the horror show where you can't

1:20:42

look away. And so I'm just motivated

1:20:45

to continue to look and to learn

1:20:47

and to research. And then I guess

1:20:49

at the end of the day, I am an eternal optimist.

1:20:52

Like, I just, I believe in humanity.

1:20:54

I believe we can actually find

1:20:56

a pathway forward. And that if

1:20:58

I just

1:20:59

try hard enough, right? If I

1:21:02

just like get

1:21:04

the message out there and work with enough other

1:21:06

incredibly thoughtful people who

1:21:08

care about humanity that

1:21:11

we will find a good pathway forward. So

1:21:13

I'm driven by the hope and the fascination.

1:21:16

I'm driven to continuously learn more. And

1:21:19

I'm just grateful that people

1:21:21

seem to respond. I'm encouraged

1:21:23

that in this moment, people

1:21:26

seem to really get it. They really seem to be

1:21:28

interested in working together collectively

1:21:31

to find a better pathway forward.

1:21:33

I feel like you walking into a room

1:21:35

or a conversation is like, have you ever seen a piece

1:21:37

of chicken thrown into piranhas? All

1:21:39

of us are just like, can you help me out? I

1:21:41

have a question, I have a question, I have a question. Like, I have

1:21:44

a question. The rest of us are like intellectual

1:21:46

piranhas being like, please don't be everything you know. Get it,

1:21:48

get it, get it, get it. And give me a hug while you're at it, thank

1:21:50

you. Well, that's

1:21:52

a good thing is I can give hugs too, right? And

1:21:55

so I'm also a mom at the end of the day.

1:21:57

I have two wonderful little girls

1:21:59

that I own.

1:21:59

who keep me grounded and

1:22:02

see the world full of curiosity and

1:22:04

kind of brilliance of all kinds

1:22:06

of possibility. And I want to help them

1:22:10

continue to see the world as this kind of

1:22:12

magical place. I want it to still be that place for them

1:22:14

as they grow

1:22:14

up. So ask actual

1:22:17

intelligent people some analog questions

1:22:19

because the one thing that we can agree on

1:22:22

is that there is some power in learning, whether

1:22:24

you're a person or a machine. And

1:22:27

now that you know some basics, you can keep up with some of the

1:22:29

headlines, but honestly, take news

1:22:31

breaks, go outside, smell

1:22:34

a tree, play pickleball or

1:22:36

something, or go read Nita's book. It's

1:22:39

called The Battle for Your Brain, Defending the Right to Think Freely

1:22:41

in the Age of Neurotechnology. We'll link that and

1:22:44

her social media handles in the show notes,

1:22:45

as well as so much more on our website at allyward.com

1:22:48

slash ologies slash neurotechnology.

1:22:51

Oh, also smologies are kid-friendly

1:22:53

and shorter episodes. Those are up at allyward.com

1:22:55

slash smologies, linked in the show notes. Thank

1:22:58

you, Zeke Rodriguez-Thomas and Jared Sleper of

1:23:00

Mindshare Media and Mercedes Maitland

1:23:02

of Maitland Audio for working on those. We are

1:23:05

at ologies on Instagram and Twitter, and I'm allyward

1:23:07

on both, just one L in ally. Thank

1:23:09

you patrons at patreon.com for such

1:23:12

great questions. You can join for a

1:23:14

dollar a month if you like.

1:23:15

Ologies merch is for sale at reasonable

1:23:17

prices at ologiesmerch.com. Thank

1:23:19

you, Susan Hale for handling that among

1:23:22

all of her many responsibilities as managing

1:23:24

producer. Noel Dilworth schedules for

1:23:26

us. Aaron Talbert, admin, Zeology's podcast

1:23:29

Facebook group. This is from Bonnie Dutch and Shannon Feltes.

1:23:31

Also happy birthday to my sister Celeste, who has

1:23:33

a great brain. Emily White of the Wordery

1:23:36

makes our professional transcripts and those are at

1:23:38

allyward.com slash ologies dash extras

1:23:41

for free. Kelly Ardwyer does our website.

1:23:43

She can do yours too. Mark David Christensen,

1:23:45

assistant

1:23:45

edited this and lead editor

1:23:48

and alarmingly smart. Mercedes Maitland of Maitland

1:23:50

Audio pulls it all together for us each

1:23:52

week. Nick Thorburn wrote and performed the

1:23:54

theme music. If you stick around until the end of the

1:23:56

episode, I tell you a secret and I'm going to treat

1:23:59

this space like a... a confessional booth, if you don't

1:24:01

mind. Okay, so once I ran into this guy

1:24:03

that I had dated who had dumped me and he was

1:24:05

with his lovely new girlfriend and

1:24:07

I pretended like I didn't hear his new

1:24:10

girlfriend's name right. I was like, what is it? As

1:24:12

if I hadn't been six

1:24:14

years deep

1:24:15

in her Facebook like the day they became official

1:24:18

and I still feel guilty about that. But I'm telling you that because

1:24:21

computers, wow, they've changed our lives and also

1:24:23

humans were so gooby and flawed.

1:24:26

But you know, everyone's code has bugs and

1:24:28

we just keep upgrading our software until

1:24:31

things work well enough. Okay, go

1:24:33

enjoy the outdoors if you can. Bye bye.