0:02

Hello, Hello, Welcome to Smart Talks with

0:04

IBM, a podcast from Pushkin

0:06

Industries, iHeartRadio and

0:08

IBM. I'm Malcolm Glabwell. This

0:11

season, we're continuing our conversation with

0:13

new creators visionaries

0:16

who are creatively applying technology

0:18

in business to drive change, but

0:20

with a focus on the transformative

0:22

power of artificial intelligence and

0:25

what it means to leverage AI

0:27

as a game changing multiplier for your

0:29

business. Our guest today

0:32

is doctor David Cox, VP

0:34

of AI Models at IBM

0:36

Research and IBM Director

0:38

of the MIT IBM Watson

0:41

AI Lab, a first of its kind

0:43

industry academic collaboration

0:46

between IBM and MIT focused

0:48

on the fundamental research of artificial

0:51

intelligence. Over the course

0:53

of decades, David Cox watched

0:56

as the AI revolution steadily

0:58

grew from the simmering ideas

1:00

of a few academics and technologists

1:02

into the industrial boom we are experiencing

1:05

today. Having dedicated

1:07

his life to push in the field of AI

1:09

towards new horizons, David has

1:11

both contributed to and presided

1:14

over many of the major breakthroughs

1:16

in artificial intelligence. In

1:19

today's episode, you'll hear David

1:21

explain some of the conceptual

1:23

underpinnings of the current AI

1:25

landscape, things like foundation models

1:28

in surprisingly comprehensible terms,

1:30

am I add. We'll also get into some

1:33

of the amazing practical applications

1:35

for AI in business, as well as what implications

1:37

AI will have for the future of work

1:40

and design. David spoke with Jacob

1:42

Goldstein, host of the Pushkin podcast

1:45

What's Your Problem. A veteran

1:47

business journalist, Jacob has reported

1:49

for The Wall Street Journal, the Miami Herald,

1:52

and was a longtime host of the NPR

1:54

program Planet Money.

1:57

Okay, let's get to the interview.

2:05

Tell me about your job at IBM.

2:08

SO. I wear two hats at IBM.

2:10

SO one, I'm the IBM Director of the MT

2:13

IBM Watson the Lab. SO

2:15

that's a joint lab between IBM and

2:17

MIT where we try and invent us

2:19

next in AI. It's been running for about five years,

2:22

and then more recently I started as the vice

2:24

president for AI Models, and I'm

2:26

in charge of building IBMS

2:28

foundation models, you know, building

2:31

these these big models, generative models that allow

2:33

us to have all kinds of new exciting capabilities in

2:35

a so. So I want to talk

2:37

to you a lot about foundation models,

2:39

about generative AI. But before we get

2:41

to that let's just spend a minute on the on

2:43

the IBM MIT collaboration.

2:47

Where where did that partnership start? How

2:49

did it originate? Yeah,

2:51

So, actually it turns out that M, I, T and

2:53

IBM have been collaborating for

2:56

a very long time in the area of AI. In

2:58

fact, that term artificial

3:00

intelligence was coined in a nineteen

3:03

fifty six workshop that was held

3:05

at Dartmouth, but it was actually organized by an IBM

3:07

or Nathaniel Rochester, who led

3:09

the development of the IBM seven oh one. So

3:12

we've really been together in AI since

3:14

the beginning, and as

3:16

AI kept accelerating more and

3:18

more and more, I think

3:20

there was a really interesting decision to let's

3:23

make this a formal partnership. So IBM

3:25

in twenty seventeen and now so to be committing close to a quarter

3:27

billion dollars over ten years

3:29

to have this joint lab with MT,

3:32

and we we located ourselves right on the

3:34

campus and we've been developing very very deep

3:36

relationships where we can really get to

3:38

know each other, work shoulder to shoulder, conceiving

3:41

what we should work on next, and then executing the projects.

3:44

And it's really, you know, very

3:46

few entities like this exist

3:48

between academia industry. It's been really

3:51

fun of the last five years to be a part

3:53

of it. And what do you think are some of

3:55

the most important outcomes of this collaboration

3:57

between IBM and MIT. Yeah,

4:00

so we're really kind of the tip

4:02

of the spear for for IBM's

4:05

BI strategy. So we're we're really

4:07

looking what, you know, what's coming ahead. And

4:10

you know, in areas like Foundation models, you know, as

4:12

the field changes and i T

4:14

people are interested in working on you know, faculty,

4:17

students and staff are interested in working on what's the latest

4:19

thing, what's the next thing. We at IBM Research are

4:22

very much interested in the same So we can kind

4:24

of put out feelers, you know, interesting things

4:27

that we're seeing in our research, interesting

4:29

things we're hearing in the field. We can go and chase those opportunities.

4:32

So when something big comes, like the big

4:34

change that's been happening lately with Foundation

4:36

Models, we're ready to jump on it. That's

4:38

really the purpose, that's that's the lab functioning

4:41

the way it should. We're also really interested

4:43

in how do we advance you

4:45

know AI that can help with climate change

4:48

or you know, build better materials

4:50

and all these kinds of things that are you know, a broader

4:52

aperture sometimes than than what we might

4:55

consider just looking at the product portfolio

4:57

of IBM, and that that gives us again a

4:59

breadth where we can connections that we might

5:01

not have seen otherwise. We can you

5:03

know, think things that help out society and

5:05

also help out our customers. So

5:08

the last whatever six

5:10

months, say, there has been this wild

5:15

rise in the public's interest in

5:17

AI, right clearly coming out of

5:20

these generative AI models that are really accessible,

5:22

you know, certainly chat GPT language

5:25

models like that, as well as models that generate images

5:28

like mid Journey. I mean, can

5:30

you just sort of briefly talk about

5:32

the breakthroughs in AI

5:34

that have made this moment feel so

5:37

exciting, so revolutionary for artificial

5:39

intelligence. Yeah.

5:41

You know, I've been studying AI

5:44

basically my entire adult life. Before

5:47

I came to IBM, I was a professor at Harvard.

5:49

I've been doing this a long time, and I've gotten used

5:51

to being surprised. It sounds like a joke, but it's

5:54

serious, Like I'm getting used

5:56

to being surprised at the acceleration of

5:58

the pace Again. It tracks

6:00

actually a long way back, you know, there's

6:03

lots of things where there was an idea that

6:05

just simmered for a really

6:07

long time. Some of the key

6:09

math behind the

6:12

stuff that we have today, which is amazing. There's

6:14

an algorithm called back propagation, which

6:17

is sort of key to training neural networks that's

6:19

been around, you know, since the eighties in

6:21

wide use. And really

6:23

what happened was it simmered for a

6:25

long time and then enough

6:28

data and enough compute came. So

6:30

we had enough data because you

6:33

know, we all started carrying multiple

6:35

cameras around with us, our mulbile phones have

6:37

all, you know, all these cameras and this we

6:39

put everything on the Internet, and there's all this data

6:42

out there. We call a lucky break that there

6:44

was something called graphics processing unit, which

6:46

turns out to be really useful for doing these kinds

6:48

of algorithms, maybe even more useful than

6:50

it is for doing graphics. They're greater graphics too,

6:53

And things just kept kind

6:55

of adding to the snowball. So we had

6:57

deep learning, which is sort of a a

7:00

rebrand of neural networks

7:02

that I mentioned from from the eighties, and that was

7:04

enabled again by data because we digitalized

7:07

the world and compute because because we

7:09

kept building faster and faster and more powerful computers,

7:12

and then that allowed us to make this this

7:14

big breakthrough. And then, you know, more

7:16

recently, using the same building

7:19

blocks, that inexorable rise

7:21

of more and more and more data met

7:24

the technology called self supervised

7:26

learning, where the key

7:29

difference there in traditional

7:31

deep learning, you know, for classifying images,

7:33

you know, like is this a cat or is this a dog? And

7:35

a picture, those technologies

7:38

require supervision, so you have to

7:40

take what you have and then you have to label

7:42

it. So you have to take a picture of a cat and then you label

7:45

it as a cat. And it turns

7:47

out that you know, that's very powerful, that

7:49

it takes a lot of time to label cats

7:51

and to the label dogs, and there's only

7:53

so many labels that exists in the world. So

7:55

what really changed more recently is

7:58

that we have self supervised learning, where you don't

8:00

have to have the labels. We can just take unannotated

8:02

data. And what that does is allots you use even

8:05

more data. And that's really what drove

8:08

this this latest sort of rage.

8:10

And then and then all of a sudden we start getting

8:13

these these really powerful models,

8:15

and then really this has been simmering

8:18

technologies. Right, this has been

8:20

happening for a while and progressively

8:23

getting more and more powerful. One of

8:26

the things that really happened with

8:28

chat Gypt and technologies like stable

8:31

diffusion and mid Journey was that

8:33

they made it visible to the public.

8:36

You know, if you put it out there, the public can touch

8:38

and feel and they're like, wow, not only is there

8:40

palpable change, and wow, you

8:43

know I can talk to this thing. Wow, this thing can generate

8:45

an image. Not only that, but everyone

8:47

can touch and feel and try. My

8:49

kids can use some

8:51

of these AI our generation technologies.

8:54

And that's really just launched.

8:57

You know. It's like a from held slingshot

8:59

at us into a different regime.

9:01

In terms of the public awareness of these technologies.

9:04

You mentioned earlier in the conversation foundation

9:07

models, and I want to talk a little bit about that.

9:09

I mean, can you just tell me, you

9:11

know, what are foundation models

9:13

for AI and why are they a big

9:15

deal? Yeah, So this

9:18

term foundation model was coined

9:20

by a group at Stanford, and

9:23

I think it's actually a really apt term

9:25

because I remember I said, you

9:27

know, one of the big things that unlocked

9:29

this latest excitement was the

9:31

fact that we could use large amounts of unannotated

9:34

data we could we could train a model. We don't have

9:37

to go through the painful effort of labeling

9:39

each and every example. You still

9:41

need to have your model do something you wanted to

9:43

do. You still need to tell it what you want

9:46

to do. You can't just have a model that doesn't

9:48

have any purpose. But what a foundation models

9:50

that provides a foundation, like

9:52

a literal foundation on you can sort of stand

9:54

on the shoulders of giants. You can have one of these massively

9:57

trained models and then do a little bit

9:59

on top. You know, you could use just a few

10:01

examples of what you're looking for and

10:04

you can get what you want from the model. So

10:06

just a little bit on top. Now it gets to the

10:09

results that a huge amount of effort used to have

10:11

to put in, you know, to get from the ground

10:13

up to that level. I

10:15

was trying to think of

10:17

of an analogy for sort

10:19

of foundation models versus what came

10:21

before, and I don't know that I came up with a

10:24

good one, But the best I could do was this. I

10:26

want you to tell me if it's plausible. It's

10:29

like before foundation models, it

10:31

was like you had these sort of single

10:33

use kitchen appliances. You could make a

10:35

waffle iron if you wanted waffles, or you could

10:38

make a toaster if you wanted to make toast.

10:40

But a foundation model is like like an

10:42

oven with a range on top. So it's like this

10:45

machine and you could just cook anything with

10:47

this machine. Yeah, that's that's

10:49

a great analogy. They're they're very versatile.

10:52

The other piece of it, too, is that they dramatically

10:55

lower the effort that it takes

10:57

to do something that you want to do.

11:00

And stand I used to say about

11:02

the old world of AI would say, you know, the problem

11:04

with automation is that it's too labor

11:06

intensive, which sounds like I'm making

11:08

a joke. Indeed, famously, if

11:10

automation does one thing, it substitutes

11:13

machines or computing power for labor.

11:16

Right, So what does that mean to say

11:18

AI is or automation is

11:20

too labor intensive. It sounds like I'm

11:22

making a joke, but I've been actually serious, And what I mean

11:24

is that the effort it took the

11:27

old regime to automate something was very

11:29

very high. So if

11:31

I need to go and curate

11:33

all this data, collect all this data, and then

11:36

carefully label all these examples that

11:38

labeling itself might be incredibly

11:41

expensive in time, and we estimate

11:43

anywhere between eighty to ninety percent of the

11:45

effort it takes to feel an AI solution

11:47

actually is just spent on data,

11:49

so that that has some consequences, which

11:52

is the threshold for

11:55

bothering. You know, if you're going to

11:57

only get a little bit of value back from

12:00

something, are you going to go through this huge effort

12:02

to curate all this data and then

12:05

when it comes time to train the model you need highly

12:07

skilled people are expensive

12:09

or hard to find in the labor market. You

12:12

know, are you really going to do something that's just a tiny, little

12:14

informal thing. Now you're going to do the only

12:16

the highest value things that warrant level

12:20

because you have to essentially build the whole

12:22

machine from scratch, and there

12:24

aren't many things where it's worth that much

12:26

work to build a machine that's only going to do

12:29

one narrow thing that's right,

12:31

and then you tackle the next problem

12:33

and you basically have to start over. And you

12:35

know, there are some nuances here, like for images,

12:38

you can pre train a model on some other task and

12:40

change it around. So there are some examples of

12:42

this, like non recurring cost

12:45

that we have in the old world too, But by and

12:47

large, it's just a lot of effort. It's hard.

12:50

It takes, you know, a large level

12:52

of skill to implement. One

12:55

analogy that I like is, you

12:57

know, think about it as you know, you have a river

12:59

of data, you know, running through your company

13:01

or your institution. Traditional

13:03

AI solutions are kind of like building a dam

13:06

on that river. You know, dams are very

13:08

expensive things to build. They require

13:10

highly specialized skills and

13:12

lots of planning. And you know, you're

13:14

only going to put a dam on a river

13:17

that's big enough that you're gonna get

13:19

enough energy out of it that it was worth your trouble.

13:21

You're gonna get a lot of value out of that dam. If you have

13:23

a river like that, you know, a river of data,

13:26

but it's actually the vast majority

13:28

of the water you know in your kingdom actually

13:30

isn't in that river. It's in puddles

13:33

and creeks and babo brooks, And you

13:36

know, there's a lot of value

13:38

left on the table because it's like, well, I

13:40

can't there's nothing you can do about it. It's just

13:42

that that's too low value.

13:45

So it takes too much effort. So

13:47

I'm just not going to do it. The return on investment just

13:49

isn't there, So you just end up not automating

13:51

things because it's too much of a pain. Now

13:54

what foundation models do is they say, well,

13:56

actually, no, we can train a

13:58

base model, a foundation that you can work on,

14:00

and we don't We don't care. We don't specify what the

14:02

task is ahead of time. We just need to learn

14:04

about the domain of data. So if

14:07

we want to build something that can understand English

14:09

language, there's a ton of English language

14:11

text available out in the world.

14:14

We can now train models on huge

14:17

quantities of it, and then it learned

14:19

the structure, It learned how language

14:22

you know, good part of how language works on all

14:24

that unlabeled data. And then when you roll up

14:26

with your task, you know, I want to solve

14:29

this particular problem. You don't have

14:31

to start from scratch. You're starting from a

14:33

very, very very high place. So

14:35

that just gives you the ability to you know, now

14:38

all of a sudden, everything is accessible.

14:40

All the puddles and greeks and babbling books

14:42

and calipons, you know, those are all accessible

14:46

now. And that's that's very exciting,

14:48

But it just changes the equation on what kinds of

14:50

problems you could use AI to solve. And

14:53

so foundation models basically mean

14:56

that automating some new

14:58

task is much less laboring, tensive,

15:00

The sort of marginal effort to do some

15:02

new automation thing is much lower

15:04

because you're building on top of the foundation

15:06

model rather than starting from scratch.

15:09

Absolutely, So that is like

15:12

the exciting good news.

15:15

I do feel like there's a little

15:17

bit of a countervailing idea that's worth talking

15:19

about here, and that is the idea that even

15:21

though there are these foundation models

15:24

that are really powerful, that are relatively

15:26

easy to build on top of, it's still

15:28

the case right that there is not some one

15:31

size fits all foundation model. So

15:34

you know, what does that mean and why is

15:36

that important to think about in this context.

15:39

Yeah, so we believe

15:42

very strongly that there isn't just one model

15:44

to rule them all. There's a number of reasons

15:46

why that could be true. One which

15:49

I think is important and very relevant today

15:51

is how much energy these

15:54

models can consume. So these

15:56

models you can get

15:58

very very large. So one

16:01

thing that we're

16:03

starting to see or starting to believe, is

16:05

that you probably shouldn't use one

16:08

giant sledgehammer model to solve

16:10

every single problem, you know, like we

16:12

should pick the right size model to solve the problem.

16:15

We shouldn't necessarily assume that we need

16:17

the biggest, baddest model for

16:20

every little use case. And we're also

16:22

seeing that, you know, small models that are trained,

16:25

like to specialize on particular

16:27

domains can actually outperform much

16:29

bigger models. So bigger isn't always

16:31

even better, So they're more efficient

16:33

and they do the thing you want them to do better

16:36

as well, that's right. So Stanford,

16:39

for instance, a group of Stanford trained a model is

16:42

a two point seven billion parameter model, which

16:44

isn't terribly big by today's standards. They

16:46

trained it just on the biomedical literature, you

16:48

know, this is the kind of thing that universities do.

16:51

And what they showed was that this model

16:54

was better at answering questions about the biomedical literature

16:56

than some models that were one hundred billion

16:58

parameters, you know, any times larger.

17:01

So it's a little bit like you know, asking

17:04

an expert for help on something versus

17:06

asking the smartest person. You know, the

17:09

smartest person you know, maybe very smart, but

17:11

they're not going to be expertise. And

17:13

then as an added bonus, you know, this is now

17:15

a much smaller model. It's much more efficient

17:17

to run. We are you know, you know, it's cheaper.

17:21

So there's lots of different advantages there. So

17:23

I think we're going to see attention

17:26

in the industry between vendors

17:29

that say, hey, this is the one, you know, big model,

17:31

and then others that say, well, actually, you know,

17:33

there's there's you know, lots of different tools

17:35

we can use that all have this nice quality that

17:37

we outlined at the beginning, and then

17:39

we should really pick the one that makes the most sense for

17:41

the task at hand. So

17:44

there's sustainability basically efficiency.

17:47

Another kind of set of issues that come up

17:49

a lot with AI our bias,

17:51

hallucination. Can you talk a

17:53

little bit about bias and hallucination,

17:56

what they are and how you're working to mitigate

17:58

those problems. Yeah, so there

18:00

are lots of issues still as amazing as these

18:02

technologies are, and they are amazing,

18:04

let's let's be very clear, lots of great

18:07

things we're going to enable with these kinds of technologies.

18:09

Bias isn't a new problem. So

18:12

you know, basically we've

18:14

seen this since the beginning of AI. If

18:17

you train a model on data

18:19

that has a bias in it. The model

18:21

is going to recapitulate that bias and

18:23

it provides its answers. So every

18:26

time, you know, if all the text you have says,

18:29

you know, it's more likely to refer to female nurses

18:31

and male scientists, then you're going to

18:33

get models that you know. For instance, there was

18:35

an example where a machine learning

18:37

based translation system translated from Hungarian

18:40

to English. Hungarian doesn't

18:42

have gendered pronouns. English does, and

18:44

when you ask it to translate to a translate they

18:47

are a nurse to she as a nurse and

18:49

would translate they are a scientist too, he is

18:51

a scientist. And that's not because the people

18:54

who wrote the algorithm were building in bias

18:56

and coding in like, oh, it's gonna be this way.

18:58

It's because the data was like that, you know, we

19:01

have biases in our society and

19:03

they're reflected in our data

19:05

and our text, in our images everywhere.

19:08

And then the models they're just mapping

19:11

from what they've what they've seen in their training data to

19:13

to the result that you're trying to get them to do and

19:16

to give, and then these biases

19:18

come out. So there's a very

19:20

active program of research

19:23

and you know, we we do quite a bit at

19:25

IBM Research and I, but

19:27

also all over the community and industry

19:30

and academia trying to figure out how do we explicitly

19:33

remove these biases, how do we identify them,

19:35

how do you know, how do we build tools that allow

19:37

people to audit their systems to make sure they aren't

19:40

biased. So this is a really important

19:42

thing. And you know, again this was here since

19:44

the beginning, you know, of of

19:47

machine learning in AI, but foundation

19:49

models and large language models in generative AI

19:53

just bring it into sharper even sharper focus

19:55

because there's just so much data and it's sort

19:57

of building in banking and all

19:59

these different biases we have. So

20:01

that's that's that's absolutely a

20:03

problem that these models have. Another

20:06

one that you mentioned was hallucinations. So

20:08

even the most impressive of our models

20:11

will often just make

20:13

stuff up. And you know, the technical term

20:15

that the fields chosen is hallucination.

20:18

To give you an example, I asked chat

20:20

tbt to create a biography

20:22

of David Cox at IBM,

20:24

and you know, it started off really well.

20:26

You know, they identified that I was the director of the MNT

20:29

IBM Watson and said a few words about

20:31

that and then it proceeded to create an

20:33

authoritative but completely fake

20:36

biography of me where I was British,

20:38

I was born in the UK, I

20:41

went to British university, you know, universities

20:44

in the UK. I was professed. It's the authority,

20:46

right, it's the certainty that that is

20:48

weird about it, right, It's it's dead certain

20:51

that you're from the UK, et cetera. Absolutely,

20:54

yeah, that's all kinds of flourishes like

20:56

I want awards in the UK. So yeah,

20:59

it's it's problematic

21:01

because it kind of pokes it a lot of

21:03

weak spots in our human psychology

21:06

where if something sounds coherent,

21:09

we're likely to assume it's true. We're

21:11

not used to interacting with people who eloquently

21:13

and authoritatively, you know, admit

21:16

complete nonsense like yeah, you

21:18

know, you know we get debated about that, but yeah,

21:20

we can debate about that, but yes, it the

21:23

it's sort of blithe confidence throws

21:25

you off when you realize it's completely wrong. Right,

21:28

that's right. And we do have a little bit

21:30

of like a great and powerful laws

21:32

sort of vibe going sometimes

21:34

where we're like, well, you know, the AI is all knowing

21:37

and therefore whatever it says must

21:39

be true. But but these things will make up stuff,

21:42

you know, very aggressively.

21:45

And you know, if everyone could try asking

21:47

it for their their bio, you you'll you'll

21:49

get something that you always get, something that's

21:52

of the right form, that has the right

21:54

tone. But you know, the facts just aren't necessarily

21:56

there. So that's obviously a problem.

21:58

We need to figure out how to close those gaps, fix

22:00

those problems. There's lots of ways

22:02

we could use them much more easily. I'd

22:05

just like to say, faced with the awesome

22:07

potential of what these technologies might do,

22:10

it's a bit encouraging to hear that even

22:12

chat GPT has a weakness

22:14

for inventing flamboyant, if

22:16

fictional versions of people's lives.

22:19

And while entertaining ourselves with chat GPT

22:22

and mid journey is important, the

22:24

way lay people use consumer facing

22:26

chatbots and generative AI

22:29

is just fundamentally different from

22:31

the way an enterprise business uses AI.

22:34

How can we harness the abilities of artificial

22:36

intelligence to help us solve the problems

22:39

we face in business and technology. Let's

22:41

listen on as David and Jacob continue

22:44

their conversation. We've been talking

22:46

in a somewhat abstract way about AI

22:49

in the ways it can be used. Let's

22:51

talk in a little bit more of a specific way.

22:54

Can you just talk about

22:56

some examples of business challenges

22:58

that can be solved with automation

23:01

with this kind of automation we're talking about. Yeah,

23:04

so they're really really disguised the

23:06

limit. There's a whole set

23:08

of different applications that these models

23:10

are a really good at. And basically it's a

23:12

super set of everything we used to use Alive

23:15

for in business. So, you know, the

23:17

simple kinds of things are like hey, if I have text

23:20

and I you know, I have like product reviews,

23:22

and I want to be able to tell if these are positive or negative.

23:24

You know, like let's look at all the negative reviews so we can

23:27

have a human look through them and see what was up.

23:30

Very common business use case. You

23:32

can do it with traditional deep learning based

23:34

AI. So so there's things like

23:36

that that are you know, it's very prosaic sort

23:38

that we were already doing it. We've been doing it for a long time.

23:41

Then you get situations that are

23:44

that we're harder for the old AI, Like if

23:46

I'm I want to compress something

23:49

like I want to I have like say I have a chat

23:51

transcript, Like a customer called in and

23:54

they had a complaint, they call back.

23:56

Okay, now a new and you know, a

23:59

person on the line needs to go read the old transcript

24:01

to catch up. Wouldn't it be better if

24:03

we could just summarize that. It's condense it all

24:06

down quick little paragraph. You know, customer

24:08

call they're upset about this, rather than having to read the

24:10

blow by blow. There's just lots of settings

24:12

like that where summarization is really

24:14

helpful. Hey, you have a meeting and

24:17

I'd like to just automatically, you know,

24:19

have have that meeting or that email or whatever.

24:21

I'd like to just have a condensed down so I can really quickly

24:24

get to the heart of the matter. These models

24:26

are really good at doing that. They're also

24:28

really good at question answering. So if

24:30

I want to find out what's how many vacation days

24:33

do I have? I can now interact

24:35

in natural language with a system

24:38

that can go and it has access to

24:40

our HR policies, and I can actually have

24:42

a you know, multi turn conversation where

24:44

I can, you know, like I would have with you know,

24:46

somebody, you know, an actual HR

24:49

professional or customer service representative.

24:52

So a big part, you

24:54

know, of what this is doing is it's

24:56

it's putting an interface. You know, when

24:58

we think of computer interfaces, were usually thinking about

25:01

UI user interface elements where I

25:03

click on menus and there's buttons and all

25:05

this stuff. Increasingly, now we

25:07

can just talk, you know, you just

25:10

in words. You can describe what you want, you

25:12

want to ask a question, you

25:14

want to sort of command the system to do something,

25:17

rather than having to learn how to do that clicking buttons,

25:19

which might be inefficient. Now we can just sort of spell

25:21

it out. Interesting, right, the graphical

25:24

user interface that we all sort of default

25:26

to, that's not like the state of

25:28

nature, Right, that's a thing that was invented

25:30

and just came to be the standard way that we interact

25:33

with computers. And so you could imagine, as

25:35

you're saying, like chat essentially

25:38

chatting with the machine could could

25:40

become a sort of standard user interface,

25:43

just like the graphical user interface, did you

25:45

know over the past several decades. Absolutely,

25:48

And I think those kinds of conversational interfaces

25:50

are going to be hugely important

25:53

for increasing our productivity. It's just a lot

25:55

easier if I if I have to learn how to use

25:57

a tool or I have to kind of have awkward,

26:00

you know, interactions from the computer. I can just tell it what

26:02

I want and I can understand it. Could you know, potentially

26:04

even ask questions back to clarify and

26:07

have those kinds of conversations that

26:09

can be extremely powerful. And

26:12

in fact, one area where that's going to I think be absolutely

26:15

game changing is in code. When we write

26:17

code. You know, programming

26:19

languages are a way

26:21

for us to sort of match between

26:24

our very sloppy way of talking and

26:27

the very exact way that you need to command a computer

26:29

to do what you wanted to do. They're cumbersome

26:32

to learn, they can you know, create very complex

26:34

systems that are very hard to reason about. And

26:37

we're already starting to see the ability to just

26:39

write down what you want and AI will

26:41

generate the code for you. And I think we're

26:43

just going to see a huge revolution of like we just

26:45

converse, and we can have a conversation to

26:47

say what we want, and then the computer can

26:50

actually not only do fixed

26:52

actions and do things for us, but it can actually

26:54

even write code to do new things, you know,

26:56

and generate the software itself. Given

26:58

how much software we have, of how much craving

27:01

we have for software, like well, we'll never have enough

27:03

software in our world. You

27:05

know, the ability to have a systems

27:07

as a helper in that, I

27:09

think we're going to see a lot of a lot of value

27:12

there. So if you if

27:14

you think about the different ways AI

27:17

might be applied to business, I mean you've talked about

27:19

a number of the sort of classic use cases.

27:21

What are some of the more out

27:24

there use cases. What are some you know, unique

27:27

ways you could imagine AI being applied

27:29

to business. Yeah,

27:32

there's really disguise the limit. I mean,

27:34

we have one project that I'm kind of a fan of

27:36

where we actually we're working

27:38

with a mechanical engineering professor

27:40

at MIT working on a classic

27:42

problem, how do you build linkage systems

27:45

which are like you imagine bars and joints

27:47

and others, you know, the things

27:49

that are building a thing, building a physical

27:52

machine of some kind of like real

27:54

like metal and

27:57

nineteenth century just old school

27:59

and industrial revolution. Yeah yeah, yeah, but

28:02

you know the little arm that's that's holding

28:04

up my microphone in front of me. Cranes,

28:06

get build your buildings, you know, parts of your engines.

28:08

This is like classical stuff. It turns out that

28:10

you know, humans, if you want to build an advanced

28:13

system, you decide what like curve

28:15

you want to create, and then a

28:17

human together with computer program can build

28:19

a five or six you know bar linkage,

28:22

and then that's kind of where you top out is because it gets

28:24

too complicated to work more than

28:26

that. We built a generative AI system

28:28

that can build twenty bar linkages. Like arbitrarily

28:31

complex. These are machines that are beyond

28:33

the capability of a human to design

28:36

themselves. Another example, we

28:39

have an AI system that can generate electronic

28:41

circuits. You know, we had a project where we're working

28:43

where we were building better power converters which

28:45

allow our computers

28:48

and our devices to be more efficient, save

28:50

energy, you know, less less

28:52

carbono. But I think the world around

28:54

us has always been shaped by technology.

28:57

If you look around, you know, just think about how many

28:59

steps and how people and how many designs

29:01

went into the table and the chair and

29:03

the I AMP. It's it's really

29:05

just astonishing. And that's already

29:08

you know the fruit of automation

29:10

and computers and those kinds of tools. But we're gonna see

29:12

that increasingly be product also

29:15

of AI. It's just going to be ever around

29:17

us. Everything we touch is going to have you

29:19

know, helped in some way to get get

29:22

to you by a You know,

29:24

that is a pretty profound transformation that

29:26

you're talking about in business. How

29:28

do you think about the implications of that both

29:30

for the sort of you know, business

29:33

itself and also for employees.

29:37

Yeah, so I think for businesses

29:39

this is gonna of costs, make

29:42

new opportunities to like customers,

29:44

you know, like there's just you

29:46

know, it's sort of all upside right, like for

29:49

the for the workers, I think the story is mostly

29:52

good too. You know, like how many things

29:54

do you do in your day that you'd

29:57

really rather not right? You know, and we're

29:59

used to have I think things we don't like automated

30:01

away, you know, we we didn't

30:04

you know, if you didn't like walking many miles

30:06

to work, then you know, like you can have a car and you

30:08

can drive there. Or we used to have a

30:10

huge traction over ninety percent of the US

30:12

population engaged in agriculture, and then we

30:15

mechanized it. Now very few people work

30:17

in agricultures. A small number of people can do the work

30:19

of a large number of people. And then

30:21

you know, things like email and you know,

30:23

they've led to huge productivity enhancements

30:25

because I don't need to be writing letters and sending

30:27

them in the mail. I can just instantly communicate

30:30

with people. We just become more

30:32

effective, Like our jobs have transformed,

30:36

whether it's a physical job like agriculture,

30:38

or whether it's a knowledge worker job where

30:40

you're sending emails and communicating

30:42

with people and coordinating teams. We've

30:44

just gotten better. And you know, the technology

30:46

has just made us more productive. And this is

30:48

just another example. Now, you know,

30:51

there are people who worry that you know, will

30:53

be so good at that that maybe jobs

30:55

will be displaced, and that's that's

30:58

a legitimate concern. But just like it's

31:01

how an agriculture, you know, it's not like suddenly

31:03

we had ninety percent of the population unemployed.

31:06

You know, people transitioned to

31:08

to other jobs. And the

31:10

other thing that we've found, too, is that our

31:12

appetite for for doing more things

31:15

is as humans is sort of insatiable.

31:17

So even if we can dramatically

31:20

increase how much you know, one human can do,

31:23

that doesn't necessarily mean we're going to do a fixed amount

31:25

of stuff. There's an appetite to have even more,

31:27

so we're gonna can continue to grow grow the

31:29

pie. So I think at least certainly

31:32

in the near term. You know, we're going to see a lot of drudgery

31:34

go away from work. We're going to see people

31:36

will be able to be more effective at

31:38

their jobs. You know, we will see some transformation

31:42

in jobs and what look like. But we've

31:44

seen that before and

31:47

the technology at least has the potential to make

31:49

our lives a lot easier. So

31:52

IBM recently launched Watson

31:55

X, which includes Watson X dot

31:57

AI. Tell me about that, Tell me

31:59

about you know what it is and the new possibilities

32:01

that it opens up. Yeah, So

32:04

Watson Next is obviously a

32:07

bit of a new branding on

32:09

the Watson brand. You know T. J. Watson

32:12

that was the founder of IBM

32:14

and our EI technologies have had

32:16

the Watson brand lots of X

32:19

is a recognition that there's

32:21

something new, there's something that actually has changed

32:23

the game. We've gone from

32:25

this old world of automation

32:28

is too labor intensive to this new world of possibilities

32:31

where it's much easier to use AI.

32:33

And what Watson X does

32:36

it brings together tools for

32:38

businesses to harness that power. So

32:41

Watson next dot AI foundational

32:44

models that our customers can use. It includes

32:47

tools that make it easy to run, easy

32:49

to deploy, easy to experiment.

32:52

There's a Watson x dot data component

32:54

which allows you to sort of organize

32:57

and access your data. So what we're really

32:59

trying to do is give our customers a

33:01

cohesive set of tools

33:03

to harness the value of

33:06

these technologies and at the same time be

33:08

able to manage the risks and other

33:10

things you have to keep an eye on and

33:12

in our prise context. So

33:15

we talk about the guests on this show

33:18

as as new creators, by

33:20

which we mean people who are creatively

33:22

applying technology in business

33:25

to drive change. And I'm

33:27

curious how creativity

33:30

plays a role in the research that you do.

33:33

I honestly, I think the creative

33:36

aspects of this job

33:38

this is what makes this work exciting.

33:41

You know, I should say, you know, the folks who

33:43

work at my organization are

33:45

doing the creating, and I guess you're

33:48

doing the managing so that they could do the

33:50

creative. I'm helping them

33:53

be their best and I still

33:55

get to get involved in the weeds of

33:57

the research as much as I can. But

34:00

you know, there's something really exciting about

34:03

inventing you know, like one of the nice

34:05

things about doing invention and doing

34:07

research on AI and industries,

34:09

it's usually grounded and a real problem that

34:12

somebody's having. You know, a customer wants

34:14

to solve this problem that's losing

34:16

money or there there could be a new opportunity.

34:18

You identify that problem and then you

34:21

you build something that's never been built

34:23

before to do that. And I think that's

34:26

honestly the adrenaline rush

34:28

that keeps all of us in this field.

34:30

How do you do something that nobody else on

34:33

earth has done before or

34:35

tried before, So that that kind of

34:37

creativity, and there's also creativity

34:39

as well, and identifying what those problems are,

34:41

being able to understand the

34:43

places where

34:46

you know the technology is close enough

34:48

to solving a problem, and doing that matchmaking

34:51

between problems that are now

34:53

solvable, you know, and in AI, where the

34:55

fields moving so fast, this is constantly

34:58

growing horizon of things

35:00

that we might be able to solve. So that matchmaking,

35:03

I think is also a really interesting creative

35:06

problem. So I think I think that's

35:08

that's that's why it's so much fun, and

35:10

it's a fun environment we have here too,

35:12

is it's you know, people drawing on whiteboards

35:15

and writing on pages of math

35:17

and like in a movie,

35:20

like in a movie. Yeah, straight from social

35:22

casting, drawing on the drawing of the window,

35:24

writing on the window, and sharpie. Absolutely.

35:27

So, So let's close

35:30

with the really long view. How

35:33

do you imagine AI and people

35:36

working together twenty years

35:38

from now? Yeah,

35:42

it's really hard to make predictions. The

35:45

vision that I

35:47

like, actually this

35:50

came from an MIT economist named

35:52

David Ottur, which was

35:55

imagine AI almost as

35:57

a natural resource. Yeah,

36:00

we have we know about natural resources work,

36:02

right, Like there's an ore we can dig up out of

36:04

the earth. It comes from kind of springs

36:06

from the earth, or we usually think of

36:08

that in terms of physical stuff. With

36:10

AI, you can almost think of it as like there's a new kind

36:13

of abundance potentially twenty years

36:15

from now where not only can we have

36:17

things we can build or eat or use or burn or

36:19

whatever. Now we have, you know, this

36:21

ability to do things and understand things

36:23

and do intellectual work. And I

36:26

think we can get to a world where

36:28

automating things is just seamless. We're

36:31

surrounded by capability

36:33

to augment ourselves to get

36:36

things done. And you

36:38

could think of that in terms of like, well, that's

36:40

going to displace our jobs because eventually the AI

36:42

system is going to do everything we can do. But

36:44

you could also think of it in terms of, like, wow,

36:46

that's just so much abundance that we now have,

36:49

and really how we use that abundance

36:51

is sort of up to us, you know,

36:53

like when you can writing software is super

36:55

easy and fast and anybody can do it. Just

36:58

think about all the things you can do now, think

37:00

about all the new activities and god, all the

37:02

ways we could use that to enrich our lives.

37:05

That's where I'd like to see us in

37:07

twenty years. You know, we can we

37:09

can do just so much more than

37:11

we were able to do before abundance.

37:14

Great, thank you so

37:17

much for your time. Yeah, it's been

37:19

pleasure. Thanks for inviting me. What

37:22

a far ranging, deep conversation.

37:25

I'm mesmerized by the vision David just described.

37:27

A world where natural conversation between

37:30

mankind and machine can generate

37:32

creative solutions to our most

37:34

complex problems. A world where

37:36

we view AI not as our

37:38

replacements, but as a powerful

37:41

resource we can tap into and

37:43

exponentially boost our innovation

37:46

and productivity. Thanks so much

37:48

to doctor David Cox for joining us

37:50

on smart Talks. We deeply appreciate

37:53

him sharing his huge breadth

37:55

of AI knowledge with us and for explaining

37:57

the transformative potential of Foundation

38:00

models in a way that even I can

38:02

understand. We eagerly await his

38:05

next great breakthroom.

38:07

Smart Talks with IBM is produced by Matt Romano,

38:10

David jah, Nisha, Ben Kat

38:12

and Royston Bserve with Jacob

38:15

Goldstein. We're edited by Lydia

38:17

Jeancott. Our engineers are Jason

38:19

Gambrel, Sarah Bugair and

38:21

Ben Holliday. Theme song by

38:24

Gramoscope. Special thanks

38:26

to Carlie Megliori, Andy Kelly,

38:28

Kathy Callahan and the eight Bar

38:30

and IBM teams, as well as

38:32

the Pushkin marketing team.

38:35

Smart Talks with IBM is a production

38:37

of Pushkin Industries and iHeartMedia.

38:40

To find more Pushkin podcasts, listen

38:42

on the iHeartRadio app, Apple Podcasts,

38:45

or wherever you listen to podcasts.

38:48

I'm Malcolm Gladwell. This is a

38:50

paid advertisement from IBM.