WIRED25: Kai-Fu Lee and Fei Fei Li On What's Next for Artificial Intelligence

(calm alternative music)

Thank you both for being here.

I'm gonna go right into this because we're talking about AI,

which I have to say, I am sure everyone

in this audience understands is

one of the biggest topics in the world,

and we don't have that much time.

And their names are up on the screen, so.

(Everyone laughing)

I wanted to start by asking Kai-Fu Lee,

you run a large VC company that focuses

you know, in large part on business that are AI based.

[Kai-Fu] Right.

And you nominated Dr. Li, Fei-Fei Li

to be a person who's going to shape the future.

Why is that?

Well. I think today, there's a lot of talk about AI.

Many by non-experts, people who have had

history, philosophy, physics backgrounds,

and they're not necessarily speaking about it

based on the true expertise.

And Fei-Fei's one of the true top experts in AI,

but more than that, she has a real conscientious and heart,

and is a great spokesperson,

and talks about AI, not as a dystopia,

but with a lot of warmth and the future opportunities

that can make us all better with AI technologies.

Right, right, so Dr. Li,

you do talk about human-based AI,

and I'd love for you to explain what that means.

You've been speaking to Congress, and writing Op-eds,

and been very vocal about it on a high level,

and it'd be great to get a good sense

and a good understanding of what

human-based AI means to you.

Sure, Maria, and also I want to start

by thanking Wired and thanking Kai-Fu for the nomination.

Kai-Fu is another Dr. Li on stage. (laughs)

I was reading his AI papers when I was a student,

so quite an honor to be here.

AI is one of the talks of the town right now.

Believe it or not, it's still a very young field.

It's only had a little more than 60 years of history,

and almost 20 years ago when I started

as a graduate student in AI,

it was my own intellectual curiosity.

Little would I have expected,

it has grown into such a field of major driving force

of the industrial transformation we're seeing.

As the excitement increases, over the past decade,

seeing the impact of AI is potentially making,

as a technologist, I've been thinking a lot about

what is the future of AI,

especially as my job as a professor.

And I not only work on the technology itself,

but I work with the next generation AI leaders.

And that's when my colleagues and I

are coming to this conclusion that

what what's really important is to put humanity

back into the center of this technology.

And to unpack a little bit the three critical ingredients of

human-centered AI, our thinking is first is:

as much as AI is already showing it's power,

it's still amazing technology.

Where is that going as a technology?

It should be more human-inspired.

Should be a cognitively,

inspired by cognition, neuroscience, and human behavior.

Can you give us another example of

what that might look like, what that might mean?

Yes, for example, that computer vision,

which is the subfield of AI I work in,

there's a lot of excitement of looking at

a picture and recognizing objects in a picture.

For example, there's a golden retriever,

or a cat, or a chair, but what about the nuance

that really mattered to humans?

When I take a look at a scene,

the interactions for example, the scene.

Our interactions, our emotional dynamics,

our relationship between us, the audience,

these kind of more human-centered understanding

of the world by a technology would be

critical in helping humans down the road.

We talked briefly about myths that you wanted to dispel.

Because I could feel how people would still be afraid.

And you had mentioned how there's these ideas out there

that AI can be a very scary thing,

and I could hear how people still might think that

even when they see, if you understood

seeing the emotional component of

seeing a cute picture of a golden retriever.

But are there myths that you both would like to dispel

that would help people understand how this is going to work,

and how having that understanding

would lead to a better place?

Well, I think there's so many myths out there.

One myth is because AI is so good at a single task,

and it does so with complex mathematics,

and it exceeds human capabilities,

there is a myth that there is singularity coming,

and when they will wake up,

and we'll all be enslaved or forced to

plug our brains to the AI, I want to ensure everyone.

[Maria] Deep breaths.

We are investors in AI.

Ai is wonderful.

AI is a great pattern recognition engine

in particular domains.

We're commercializing it.

It's going to help us make money, save money.

But it is nowhere close to displacing humans,

and also I totally agree with Fei-Fei that

AI should be used as a tool to help doctors treat patients,

help scientists find new drugs,

and there's a good opportunity for symbiosis.

That brings me to two steps.

One is, you both have spoken about the benefits of AI,

and it would be good to talk a little bit more

about where you see AI being

incredibly helpful in the future.

Other people on this stage have mentioned

really wonderful uses of Artificial Intelligence,

and Jenny Lay-Flurrie just mentioned that in disability.

So, one of the other theme of human centered AI

is about enhancing and augmenting humans

rather than replacing humans.

This word, this very word replacement

having been a central topic of AI during this AI time.

And as a technologist, we recognize that

this is a technology that could play

a huge role in enhancing and augmenting.

As an example, at Stanford,

I've been working with Stanford Medical School

experts and doctors in how the technology of AI

can help the work flow of the clinical health-care delivery.

For example, hand hygiene, hospital-inquired infection,

kills three times more people than car accidents in the US.

Mostly due to the lack of good

hand hygiene practiced by clinicians.

And in the Standford Hospital right now,

we are looking at a project by using smart censors

and the deep-learning algorithm behind it

to help clinicians to assist and improve

their hand hygiene practice in

the clinically important situations

when they are interacting with patients.

And this is an example where we give clinicians

the time, the focus back to the patients

but assist them to make better decisions

or practice better work flow,

and this is just one small example of that.

That does lead me to, and you've spoken about jobs.

Because there is some fear, right?

Let me first talk about another example

that shares commonalities in education.

At Sinovation Ventures, we funded about

45 AI companies and about 40 education companies.

And we've been helping them to connect them together.

In China, the top schools and the top cities

have great teachers, and that, in villages,

you still have 40, 50 kids from first to sixth grade

going to one school with one or two teachers

who can't possibly be experienced to be great teachers.

So, we've connected and helped all these

portfolio companies to connect up.

Video conferencing form the city of Beijing,

simultaneously teaching 800 kids in this villages

and in the village schools there's a clicker

with a large video conference.

And also, to help these village teachers,

there are tools that we've provided for them

that can grade homework assignments,

that can do drills for math, (mumbles) their week,

helping with english pronunciation.

Really, AI is coming in to help teachers help students

individually improve their capabilities and their week

and give more time to the students to spend with teachers,

giving teachers more time to one-on-one mentoring.

I was going to ask if you are

then taking the teacher's jobs then.

Well, not at all, because I think there will be

three times as many teachers in the world.

There's not enough teachers in poorer regions in the world,

and teacher's jobs should be a mentor, one on one,

face-to-face, understanding what the student needs,

and giving mentoring and coaching and guidance.

Not the repetitive course, during roll call,

giving assignments, doing tasks.

The AI should do all that.

You've both spoken about

that there will be jobs that are lost,

and there's a balance and an augmentation.

I know you've spoken about that quite a bit.

Is there anything more you'd want to add?

I think we have to separate.

I think it's neither the case that

all of AI will be symbiotic with humans,

nor the case that AI will take human jobs.

In my new book, I talk about different types of situations.

In the creative types, I think AI can amplify the creatives.

In the professional's case, I think AI can do the

analytic thinking with the human adding more to it,

so one plus one equals three.

Like a teacher, or a doctor.

But there are routine jobs that will be taken by AI,

and I think we have to be cognizant of that.

As we look at some of our investments,

some of them are creative,

some of them are augmentative, symbiotic,

but there are some that are displacing jobs.

So part of why would the VC write the book

is to alert people that there will be some jobs

that would be lost due to AI's ability to do them better.

Mostly to the routine jobs.

That brings me to a slight shift,

and I realize we're quickly running out of time,

but Dr. Li, you've talked a lot about diversity,

and you've talked about wanting to make sure

that the tools and the people building

the technology we're going to be using is diverse.

Can you talk a little bit more about that?

When it comes to AI, I feel like humanity

has never created a technology that resembles ourselves

as much as AI technology has.

And in the context of that, who are the people making AI?

Who are the people being influential

to AI technology as well as impact.

We're severely lacking that diversity in that

from the technology world all the way to the policy,

to the policy world.

So, involving diversity has several

really important advantage.

First is actually jobs itself,

we're still lacking computing AI talents

in the world to create better, more AI,

and we're missing huge chunks of humanity by not involving

women and underrepresented minority.

And it's also important for innovation and creativity,

and most of all it's bringing that diverse value

and representation to the creation

and application of this technology

brings that kind of justice and moral compass to this.

And it's critical for AI's future.

You're starting a new Stanford initiative.

Is that what the focus is?

Stanford's in preparation

of a human-centered AI initiative.

This is really bringing together

the top thinkers of humanities

such as social scientists such as economists,

philosophers, law scholars, ethicists,

next together with the technologists, engineers,

to think about the next AI generation of technology,

its social and humanistic impact,

and to investing areas like education, healthcare,

automation and sustainability,

in guiding AI towards a more human-centered

development.

In our last very few seconds,

25 years out, are you optimistic?

Do we have the time now to get to an optimistic view?

Well, I think now we're faced with a lot

of challenges and chaos, but in 25 years,

we're going to look back

and be very very grateful

that AI has liberated us from routine jobs.

So, we can think about why we exist

and what we're passionate about.

Not much more, sorry!

No, no, go ahead.

I'm only optimistic if this generation takes on

the responsibility of human-centered AI.

Do we have time?

We have time, but we have to act now.

We have so many things we could talk

about but we are out of time.

We are out of time. (laughs)

(audience claps)

[Maria] Thank you.