We’re already using AI more than we realize

00:01

Of all the interactions you have with technology in a day,

00:04

interacting with artificial intelligence —

00:07

or not — feels like a choice.

00:10

But in some ways it isn't.

00:13

Over the past decade, we've become surrounded by AI systems

00:16

that perceive our world,

00:18

that support our decisions and that mimic our ability to create.

00:22

Whether we're aware of it or not is another story.

00:33

Imagine a day like this.

00:36

You do some exercise with a smartwatch, put on a suggested playlist,

00:38

go to a friend's house and ring their camera

00:40

doorbell, browse recommended shows on Netflix.

00:42

Check your spam folder for an email you've been waiting for.

00:45

And when you can't find it, talk to a customer support chatbot.

00:48

Each of those things are made possible by technologies

00:51

that fall under the umbrella of artificial intelligence.

00:55

But when a Pew survey

00:56

asked Americans to identify whether each of those used AI

01:00

or not, they only got it right

01:01

about 60% of the time.

01:03

Some of these applications of AI have become fairly ubiquitous.

01:07

They almost exist in the background, and it's not terribly apparent to folks

01:11

that the tools or services they're using are being powered

01:14

by this technology.

01:16

That's Alec Tyson, one of the researchers behind that Pew study.

01:19

When Tyson and his team asked respondents how often they think they use AI,

01:23

almost half didn't think they regularly interact with it at all.

01:28

Some of them might be right, but most probably just don't know it.

01:31

We know about 85% of US adults are online every day, multiple times a day.

01:36

Some folks are online almost all the time.

01:39

This suggests a bit of a gap where there seem to be some folks

01:42

who really must be interacting with a AI, but it's never salient to them.

01:46

They don't perceive it.

01:49

So why does that gap exist?

01:52

Part of the problem is that the term artificial

01:54

intelligence has been used to refer to a lot of different things.

01:58

Artificial intelligence is totally this giant umbrella tent term

02:02

that now has become a kitchen sink of everything.

02:05

That’s Karen Hao. She's a reporter who covers artificial intelligence and society.

02:09

In the past, there were distinct disciplines

02:13

about which aspect of the human brain do we want to recreate?

02:16

Like do we want to recreate the vision part?

02:18

Do we want to recreate our ability to hear? Our ability to write and speak?

02:24

Giving the machine the ability to see became the field of computer vision.

02:28

Giving the machine the ability to write and speak

02:30

became the field of natural language processing.

02:33

But on their own, these tasks still required a machine to be programed.

02:37

If we wanted machines to recognize spam emails, we had to explicitly program them

02:41

to look out for specific things, like poor spelling and urgent phrasing.

02:45

That meant the tools weren't very adaptable to complex situations.

02:50

But that all changed when we

02:51

started recreating the brain's ability to learn.

02:54

This became the subfield of machine learning,

02:57

where computers are trained on massive amounts of data so that instead of

03:00

needing to hand-code rules about what to see or speak or write,

03:04

those computers can develop rules on their own.

03:06

With machine learning, a computer could learn to recognize new spam emails

03:10

by reviewing thousands of existing emails that humans have labeled as spam.

03:15

The machine recognizes patterns in this structured

03:17

data and creates its own rules to help identify those patterns.

03:21

When the training data hasn't been structured and labeled by humans,

03:23

that method is called “deep learning.”

03:25

Most of the time people talk about AI now,

03:27

they're not talking about the whole field, but specifically these two methods.

03:32

We'll hear more about that

03:33

after a word from this video’s sponsor.

03:36

This episode is presented by Microsoft Copilot

03:39

for Microsoft 365, your AI assistant at work.

03:43

Copilot can help you solve your most complex problems at work,

03:46

going far beyond simple questions and answers. From getting up to speed

03:50

on a missed Teams meeting in seconds to helping you start a first draft faster

03:54

in Word, copilot for Microsoft 365 gives everyone an AI assistant at work

04:00

in their most essential apps

04:01

to unleash creativity, unlock productivity and up level skills.

04:06

And it's all built on Microsoft's comprehensive approach to security,

04:10

privacy, compliance and responsible AI.

04:13

Microsoft does not influence the editorial process of our videos,

04:16

but they do help make videos like this possible.

04:18

To learn more, you can go to Microsoft.com/copilotforwork.

04:23

Now back to our video.

04:25

Improvements in computing power, together with the massive amounts of data

04:29

generated on the Internet, made possible a whole new generation of technologies

04:33

that leveraged machine learning. And existing ones swapped out

04:36

their algorithms for machine learning too.

04:38

A lot of the “how” in the back has been swapped into AI over time

04:44

because people have realized, “oh wait, we can actually get an even better

04:47

performance of this product if we just swap our original algorithm,

04:52

our original code out for a deep learning model.”

04:56

Now, machine learning and deep learning models power recommendation

05:00

for shows, music, videos, products and advertisements.

05:04

They determine the ranking of items every time we browse search results

05:07

or social media feeds. They recognize images like faces to unlock phones

05:11

or use filters, and the handwriting on remote deposit checks.

05:14

They recognize speech in transcription,

05:16

voice assistants, and voice-enabled

05:18

TV remotes. And they predict text in autocomplete and autocorrect.

05:24

But AI is seeping into more than that.

05:26

There has been this tendency over the last ten plus years

05:30

where people have started

05:33

putting AI into absolutely everything.

05:36

Machine learning algorithms are already

05:38

being used to decide which political ads we see, which jobs we qualify for,

05:42

and whether we qualify for loans or government benefits,

05:45

and often carry the same biases as the human decisions that preceded them.

05:49

Are you actually automating the poor decision

05:53

making that happened in the past and just bringing it into the future?

05:56

If you're going to use historical data to predict

05:59

what's going to happen in the future,

06:00

you're just going to end up with a future that looks like the past.

06:03

And that's part of the reason why it matters to close that gap

06:06

between those who knowingly interact with the AI every day

06:10

and those who don't quite know it yet.

06:12

Awareness needs to grow for folks to be able to participate

06:15

in some of these conversations about the moral and ethical boundaries,

06:18

what air should be used for, and what it shouldn't be used for.

10:44

Over the past decade,

10:45

we've become surrounded by AI systems that perceive our worlds

10:49

that support our decisions, and that mimic our ability

10:52

to create.

11:02

Over the past

11:03

decade, we've become surrounded by AI systems that perceive our worlds,

11:07

that support our decisions, and that mimic our ability to create.

11:16

But over the past decade, we've become surrounded by AI systems

11:20

that perceive our worlds, that support our decisions

11:23

and that mimic our ability to create.

11:26

Over the past decade, we've become surrounded by AI systems

11:29

that perceive our worlds,

11:30

that support our decisions, and that mimic our ability to create.

11:34

Whether we're aware of it or not is another story.

11:38

But when a Pew survey asked Americans to identify whether these technologies use

11:43

AI, they only got it right about 60% of the time.

11:48

But when a Pew survey asked

11:49

Americans to identify whether these technologies use AI,

11:53

they only got it right about 60% of the time.

12:00

But when but when a.

12:03

But when a Pew survey

12:06

but when a Pew survey asked.

12:10

But when a pew.

12:11

But when a Pew survey asked.

12:13

But when a Pew survey asked Americans

12:15

to identify whether these technologies use AI,

12:18

they only got it right about 60% of the time.

12:22

But when a Pew survey asked Americans

12:29

but when a Pew survey asked

12:31

Americans to identify whether these technologies use AI,

12:34

they only got it right about 60% of the time.

12:42

We'll hear more about that

12:43

after a word from this video sponsor.

12:56

We'll hear

12:57

we're going to hear more about that after.

13:00

We'll hear more about that after a word from this video's sponsor.

13:04

We'll hear more about that.

13:06

We'll hear more about that after a word from this video sponsor

13:13

now, machine learning.

13:16

Now machine learning and deep learning.

13:18

Now, machine learning and deep learning models Power, recommendations

13:21

for shows, recommendations.

13:28

Recommendations for shows.

13:30

Recommendations

13:34

We'll hear more about.

13:35

We'll hear more about that after a word from this video sponsor.

13:39

We'll hear more about that.

13:41

We'll hear more about that after a word from this video

13:44

sponsor.

22:12

Okay,

22:18

great.

22:28

Hi, I'm

22:29

Karen Howe and I am a reporter that's been covering

22:32

artificial intelligence for over five years.

22:34

And I am currently also working on a book about opening

22:37

AI for Penguin Press

23:19

and totally,

23:29

yeah,

23:48

yeah,

23:51

yeah.

23:52

Artificial Intelligence is totally this giant umbrella tent term that now

23:57

it's become a kitchen sink of everything but the origins of the term.

24:02

And this sort of helps to understand why it's so broad.

24:06

The origins of the term are from an academic field, the founding

24:10

of an academic field called A.I., and that happened in the 1950s,

24:14

and it was a group of academics in the US that actually had a meeting

24:18

at Dartmouth University to decide that they wanted to create a brand new field.

24:21

They wanted to be the founding fathers of this field, and specifically

24:26

they wanted to try and attain human level intelligence in computers.

24:31

And there were there have been over the decades, many different hypotheses

24:37

from like a scientific perspective about how to do this.

24:40

From what?

24:42

Like one of those hypotheses

24:44

is that we are intelligent because we know things,

24:46

and so we should build intelligent computers by encoding

24:50

all of the rules that we know about the universe into a computer.

24:53

Another theory has been we are intelligent because we can learn very quickly,

24:58

so we should build intelligent computers by building learning machines.

25:02

And so that theory that second one has become

25:06

the dominant paradigm of everything.

25:10

Basically everything that we see today

25:13

and those learning machines is now called machine learning

25:16

and machine learning has continued to advance

25:19

in the last decade and a half or so from just simple machine

25:24

learning like statistical calculations to deep learning, which means

25:29

fancier statistical calculations.

25:32

And there's a whole range of commercial products that have spun out of this

25:35

particular technology

25:37

that have really nothing to do with trying to recreate the human brain.

25:40

But more just companies can make money off of it.

25:42

And so they're going to keep doing that.

25:44

So deep learning technologies include things like voice assistants,

25:49

self-driving cars, facial recognition

25:52

and all the way up today to gravity and stable diffusion.

25:56

All of these count as deep learning.

25:59

And what deep learning ultimately is doing

26:02

is it's these techniques that allow

27:27

to become fairly ubiquitous.

27:28

They almost exist in the background, and it's not terribly apparent to folks

27:32

that what they're the tools or services they're using are being

27:35

powered by this technology.

27:37

And I return to the point we talked about earlier that there are others

27:40

where it's more apparent

27:41

chatbots, for example, we didn't get it into generative AI

27:45

in this example,

27:46

but that's one where it's more front and center

27:48

that that a computer is doing some thinking

27:50

or at least mimicking some thinking in a way that's very directly

27:54

more associated with artificial intelligence, where

27:57

some of the consumer technology has become a little bit more in the background.

28:00

And it's harder for folks to perceive that AI is influencing or helping them

28:03

go about their lives

28:24

and it

28:59

well, one thing that we know

29:01

is a question we ask very simple question how much have you heard or read about?

29:05

A You know, on the one hand, 90% of Americans say, well, I've heard

29:09

at least a little about it, but only a third say they've heard a lot about it.

29:13

Sort of a deep or rich knowledge in this matters, Right. Why?

29:18

Why do we care about awareness? Why do we study it?

29:20

It's really the first step towards a broader public engagement

29:24

with the host of moral and ethical questions that A.I.

29:27

raises for society.

29:28

So we're at a really interesting moment here

29:30

where most Americans are generally aware of artificial intelligence,

29:34

But deep knowledge, intimate knowledge is still fairly modest.

29:38

And it is growing. Absolutely, it's growing.

29:40

But it needs to grow for folks to be able to participate

29:44

in some of these conversations about the moral and ethical boundaries.

29:47

What I should be used for and what it shouldn't be used for.

29:50

So that's part of why we study awareness at the center

29:53

and why we feel it's important.

30:41

Know Yeah,

30:50

well, one thing that we do know is that not everyone brings

30:53

the same level of awareness to understanding

30:56

something like artificial intelligence and a really big factor.

30:59

Probably the biggest factor is level of formal education, right?

31:02

Where college graduates and those with post-graduate degrees,

31:05

they express higher self-reported awareness

31:08

and they score better on our awareness sort of scale that we've developed.

31:12

And that's important.

31:13

It just sort of underscores that

31:15

these conversations are a bit different depending on what circles you're in,

31:19

whether it's you're in a formal education

31:20

setting or a job that requires this technology

31:22

or maybe you're not in those settings.

31:24

So absolutely, there are differences across the public that reflect things

31:28

like formal education, job type and even types of conversation

31:33

or social circles that really matter when it comes to where folks are in terms

31:37

of understanding, interpreting and how they feel about artificial intelligence.

31:44

So, well, look, there's so many big questions

32:22

with artificial intelligence, but access and equity are certainly among them.

32:26

We'll folks who do understand this technology, who can use it well

32:31

and leverage it to their advantage, these prime equity and access questions.

32:35

And there's the conversation around that or the solutions are very complex.

32:40

But that's certainly one thing at play here

32:42

is there's immense power with some of these technologies.

32:45

They may not be equally available or understood by all

32:49

spheres of the public in Is that okay?

32:51

That's a public conversation to have.

32:52

That's part of what our research can do

32:54

is provide a foundation for having that conversation right.

32:58

So, well, I'm in trouble now.

33:09

I've gone too far with Shad to find a way to take back at all.

33:15

I hear the heavy hand of lies and deceit.

33:19

Absolutely.

33:20

I mean, part of what we do

33:22

at the center, an enormous amount of time goes into question or development.

33:25

The ultimate question we end up on really represents a fraction

33:29

of all the versions and items and ideas being discussed.

33:33

And we had to balance considerations here.

33:35

One is we have to make this accessible to folks

33:38

we can't use necessarily highly technical or almost esoteric examples

33:42

which don't resonate well with folks who are giving their time to take a survey.

33:47

So there's many other ways of imagining an awareness index

33:51

and there's more to do on this front, both by us and others.

33:55

But we thought a good first step would be to try and identify

33:58

some applications that are fairly well known in everyday life,

34:03

or at least these of the tools they're powering are fairly widely used.

34:07

We thought that was a good first place to start.

34:09

There may be other ways to go deeper.

34:11

Some of the more complex or technical or industry specific uses,

34:16

that's a great opportunity for future research.

34:18

But we wanted to start with ten folks even recognize where this technology

34:22

is, that play and even some of the most common.

34:24

We use tools online today, like online shopping, like email, like fitness

34:28

trackers.

34:28

Let's start there and see what we can learn.

35:02

Well, that's

35:03

that is the question is as folks become more aware, as expressions of

35:08

I increasingly shape the way we work live, how are attitudes going to change?

35:13

What do we need to understand going forward?

35:16

Right now, there's a fair amount of caution

35:17

among the public about A.I., right?

35:20

Far more say they're concerned and excited about its growing role in life.

35:24

How will that change going forward?

35:26

A little more awareness only reinforce or grow concern or a critic go in the other

35:31

direction, whereas the more aware folks get, the more positive they get.

35:35

We've seen examples of both in our past research there.

35:38

There is it's a story that's yet to be written, right?

35:41

But it's one that's important to follow.

35:43

Early indications are even in the past year or two,

35:46

that the public is actually more concerned, not less, as they become

35:50

a bit more aware of the ways in which air is involved in their own life.

35:54

Spaces like health and medicine, spaces like jobs and hiring,

35:58

perhaps some potential uses in law enforcement right there.

36:01

There's some concern out there among Americans

36:03

about how this is all going to play out

36:26

in the.

36:53

Yeah, absolutely.

36:55

And they're really sort of multiple prongs to our research plan here at the center,

37:00

which is for some of the most controversial uses of I like A.I.

37:04

and Health and medicine,

37:05

we focus a little bit less on awareness and more about attitudes.

37:09

How would you feel?

37:09

Would you feel comfortable with air and your own primary care?

37:13

Would you be open to using AI and something like an image review

37:16

and skin cancer detection?

37:18

So in some of the most you could call the most

37:22

perhaps potentially impactful on people's own lives, we go directly to attitudes.