Placeholder Image

Subtitles section Play video

  • [MUSIC PLAYING]

  • DIANE GREENE: Hello.

  • FEI-FEI LI: Hi.

  • DIANE GREENE: Who's interested in AI?

  • [CHEERING]

  • Me too.

  • Me three.

  • OK.

  • So I'm the moderator today.

  • I'm Diane Greene, and I'm running Google Cloud

  • and on the Alphabet board.

  • And I'm going to briefly introduce

  • our really amazing guests we have here.

  • I also live on the Stanford campus,

  • so I've known one of our guests for a long time,

  • because she's a neighbor.

  • So let me just introduce them.

  • First is Dr. Fei-Fei Li, and she is the Chief Scientist

  • for Google Cloud.

  • She also runs AI Lab at Stanford University, the Vision Lab,

  • and then she also founded SAILORS,

  • which is now AI4ALL, which you'll

  • hear about a little bit later.

  • And is there anything you want to add to that, Fei-Fei?

  • FEI-FEI LI: I'm your neighbor.

  • [LAUGHTER]

  • That's the best.

  • DIANE GREENE: And so now we have Greg Corrado.

  • And actually there's one amazing coincidence.

  • Both Fei-Fei and Greg were undergraduate physics majors

  • at Princeton together at the same time.

  • And didn't really know each other that well

  • in the 18-person class.

  • FEI-FEI LI: We were studying too hard.

  • GREG CORRADO: No, it was kind of surprising to go

  • to undergrad together, neither of us in computer science,

  • and then rejoin later only once we were here at Google.

  • DIANE GREENE: All paths lead to AI and neural networks

  • and so forth.

  • But anyhow, so Greg is the Principal Scientist

  • in the Google Brain Group.

  • He co-founded it.

  • And more recently, he's been doing

  • a lot of amazing work in health with neural networks

  • and machine learning.

  • He has a PhD in neuroscience from Stanford.

  • And so he came into AI in a very interesting way.

  • And maybe he'll talk about the similarities between the brain

  • and what's going on in AI.

  • Would you like to add anything else?

  • GREG CORRADO: No, sounds good.

  • DIANE GREENE: OK.

  • So I thought since both of them have been involved in the AI

  • field for a while and it's recently

  • become a really big deal, but it'd

  • be nice to get a little perspective on the history,

  • yours in Vision and yours in neuroscience, about AI

  • and how it was so natural for it to evolve to where it is now

  • and what you're doing.

  • And start with Fei-Fei.

  • FEI-FEI LI: I guess I'll start.

  • So first of all, AI is a very nascent field

  • in the history of science of human civilization.

  • This is a field of only 60 years of age.

  • And it started with a very, very simple but fundamental quest--

  • is can machines think?

  • And we all know thinkers and thought leaders

  • like Alan Turing challenged humanity with that question.

  • Can machines think?

  • So about 60 years ago, a group of very pioneering scientists,

  • computer scientists like Marvin Minsky, John McCarthy,

  • started really this field.

  • In fact, John McCarthy, who founded Stanford's AI lab,

  • coined the very word artificial intelligence.

  • So where do we begin to build machines that think?

  • Humanity is best at looking inward in ourselves

  • and try to draw inspiration from who we are.

  • So we started thinking about building machines that

  • resemble human thinking.

  • And when you think about human intelligence,

  • you start thinking about different aspects like ability

  • to reason and ability to see and ability

  • to hear, to speak, to move around, make decisions,

  • manipulate.

  • So AI started from that very core, foundational dream

  • 60 years ago, started to proliferate

  • as a field of multiple subfield, which includes robotics,

  • computer vision, natural language processing,

  • speech recognition.

  • And then a very important development

  • happened around the '80s and '90s,

  • which is a sister field called machine learning started

  • to blossom.

  • And that's a field combining statistical learnings,

  • statistics, with computer science.

  • And combining the quest of machine intelligence, which

  • is what AI was born out of, with the tools and capabilities

  • of machine learning.

  • AI as a field went through an extremely

  • fruitful, productive, blossoming period of time.

  • And fast-forward to the second decade of 21st century.

  • The latest machine learning booming that we are observing

  • is called deep learning, which has

  • a deep root in neuroscience, which I'll let you talk about.

  • And so combining deep learning as

  • a powerful statistical machine learning tool

  • with the quest of making machines more intelligent.

  • Whether it's to see or is it to hear or to speak,

  • we're seeing this blossom.

  • And last I just want to say, three critical factors

  • converged around the last decade,

  • which is the 2000s and the beginning of 2010s, which are

  • the three computing factors.

  • One is the advance of hardware that

  • enabled more powerful and capable computing.

  • Second is the emergence of big data,

  • powerful data that can drive the statistical learning

  • algorithms.

  • And I was lucky to be involved myself in some of the effort.

  • And then the third one is the advances of machine learning

  • and deep learning algorithms.

  • So this convergence of three major factors

  • brought us the AI boom that we're seeing today.

  • And Google has been investing in all three areas,

  • honestly, earlier than the curve.

  • Most of the effort started even in early 2000s.

  • And as a company, we're doing a lot of AI work

  • from research to products.

  • GREG CORRADO: And it's been really interesting to watch

  • the divergence in exploration in various academic fields

  • and then the re-convergence as we see ideas that are aligned.

  • So it wasn't, as Fei-Fei says, it wasn't so long

  • ago that fields like cognitive science, neuroscience,

  • artificial intelligence, even things

  • that we don't talk about much more like cybernetics,

  • were really all aligned in a single discipline.

  • And then they've moved apart from each other

  • and explored these ideas independently

  • for a couple of decades.

  • And then with the renaissance in artificial neural networks

  • and deep learning, we're starting

  • to see some re-convergence.

  • So some of these ideas that were popular

  • only in a small community for a couple of decades

  • are now coming back into the mainstream

  • of what artificial intelligence is, what statistical pattern

  • recognition is, and it's really been delightful to see.

  • But it's not just one idea.

  • It's actually multiple ideas that you

  • see that were maintained for a long time in fields

  • like cognitive science that are coming back into the fold.

  • So another example beyond deep learning

  • is actually reinforcement learning.

  • So for the longest time, if you looked at a university

  • catalog of courses and you were looking

  • for any mention of reinforcement learning whatsoever,

  • you were going to find it in a psychology

  • department or a cognitive science department.

  • But today, as we all know, we look

  • at reinforcement learning as a new opportunity,

  • as something that we actually look

  • at for the future of AI that might be something that's

  • important to get machines to really learn

  • in completely dynamic environments,

  • in environments where they have to explore entirely

  • new stimuli.

  • So I've been really excited to see how this convergence has

  • happened back in the direction from those ideas

  • into mainstream computer science.

  • And I think that there's some hope for exchange

  • back in the other direction.

  • So neuroscientists and cognitive scientists

  • today are starting to ask whether we

  • can take the kind of computer vision models

  • that Fei-Fei helped pioneer and use those as hypotheses for how

  • it is that neural systems actually compute, how

  • our own biological brains see.

  • And I think that that's really exciting

  • to see this kind of exchange between disciplines

  • that have been separated for a little while.

  • DIANE GREENE: You know, one little piece of history I think

  • that's also interesting is what you did, Fei-Fei,

  • with ImageNet, which is a nice way of explaining building

  • these neural networks where you labeled all these images

  • and then people could refine their algorithms by--

  • go ahead and explain that just real quickly.

  • FEI-FEI LI: OK, sure.

  • So about 10 years ago, the whole community of computer vision,

  • which is a subfield of AI, was working on a holy grail problem

  • of object recognition, which is you open your eyes,

  • you can see the world full of objects

  • like flowers, chairs, people.

  • And that's a building block of visual intelligence

  • and intelligence in general.

  • And to crack that problem, we were building, as a field,

  • different machine learning models.

  • We're making small progress, but we're hitting a lot of walls.

  • And when my student and I started working on this problem

  • and started thinking deeply about what

  • is missing in the way we're approaching this problem,

  • we recognize this important interplay

  • between data as statistical machine learning models.

  • They really reinforce each other in very deep mathematical ways

  • that we're not going to talk about the details here.

  • That realization was also inspired by human vision.

  • If you look at how children learn,

  • it's a lot of learning through big data

  • experiences and exploration.

  • So combining that, we decided to put together

  • a pretty epic effort of we wanted

  • to label all the images we can get on the internet.

  • And of course, we Google Searched a lot

  • and we downloaded billions of images

  • and used crowdsourcing technology

  • to label all the images, organize them

  • into a data set of 50 million images, organized

  • in 22,000 categories of objects, and put that together,

  • and that's the ImageNet project.

  • And we democratized it to the research world

  • and released the open source.

  • And then starting in 2010, we held

  • an international challenge for the whole AI community

  • called ImageNet Challenge.

  • And one of the teams from Toronto,

  • which is now at Google, won the ImageNet Challenge

  • with the deep learning convolutional neural network

  • model.

  • And that was year 2012.

  • And a lot of people think the combination of ImageNet

  • and the deep learning model in 2012

  • was the onset of what Greg--

  • DIANE GREENE: A way to compare how they were doing.

  • And it was really good.

  • So yeah.

  • And so Greg, you've been doing a lot of brain-inspired research,

  • very interesting research.

  • And I know you've been doing a lot of very impactful research

  • in the health area.

  • Could you tell us a little bit about that?

  • GREG CORRADO: Sure.

  • So I mean, I think the ImageNet example actually

  • sort of sets a playbook for how we

  • can try to approach a problem.

  • The kind of machine learning and AI

  • that is most practical and most useful today

  • is ones where machines learn through imitation.

  • It's an imitation game where if you have examples

  • of a task being performed correctly,

  • the machine can learn to imitate this.

  • And this is called supervised learning.

  • And so what happened in the image recognition

  • case is that by Fei-Fei building an object recognition data set,

  • we could all focus on that problem

  • in a really concrete, tractable way

  • in order to compare different methods.

  • And it turned out that methods like deep learning

  • and artificial neural networks were

  • able to do something really interesting in that space

  • that previous machine learning and artificial intelligence

  • methods had not, which was that they were able to go directly

  • from the data to the predictions and break the problem up

  • into many smaller steps without having being

  • told exactly how to do that.

  • So that's what we were doing before is that we were trying

  • to engineer features or cues, things that we could see

  • in the stimuli that then we would do

  • a little bit of statistical learning on to figure out

  • how to combine these signals.

  • But with artificial neural networks and deep learning,

  • we're actually learning to do those things all together.

  • And this applies not only to computer vision,

  • but it applies to most things that you could

  • imagine a machine imitating.

  • And so the kinds of things that we've

  • done like with Google Smart Reply and now Smart Compose,

  • we're taking that same approach.

  • That if you have a lot of text data, which it turns out

  • the internet is full of, what you can actually do

  • is you can look at the sequence of words

  • so far in a conversation or in an email exchange

  • and try to guess what comes next.

  • DIANE GREENE: I'm going to interrupt here a little bit

  • and get a little more provocative here.

  • GREG CORRADO: All right.

  • DIANE GREENE: So you're talking about neural-inspired machine

  • learning and so forth.

  • And so this artificial intelligence

  • is kind of bringing into question what are we humans?

  • And then there's this thing out there

  • called AGI, Artificial General Intelligence.

  • What do you think's going on here?

  • Are we getting to AGI?

  • GREG CORRADO: I really don't think so.

  • So there's a variety of opinions in the community.

  • But my feeling is that, OK, we've finally

  • gotten artificial neural networks

  • to be able to recognize photos of cats.

  • That's really great.

  • We also now can--

  • DIANE GREENE: Fei-Fei, was that AGI when we recognized a cat?

  • FEI-FEI LI: No.

  • That's not enough to define AGI.

  • GREG CORRADO: So the kind of thing that's working well right

  • now is this sort of pattern recognition,

  • this immediate response where we're able to recognize

  • something kind of reflexively.

  • And we now have, I believe, machines

  • can do pattern recognition every bit as well as humans can.

  • And that's why they can recognize objects

  • in photos, that's why they can do speech recognition,

  • and that's why they can win at a game like Go.

  • But that is only one small sliver, a tiny sliver,

  • of what goes into something like intelligence.

  • Notions of memory and planning and strategy

  • and contingencies, even emotional intelligence, these

  • are things that we haven't even scratched the surface.

  • And so to me, I feel like it's really a leap

  • too far to imagine that having finally cracked pattern

  • recognition, after some decades of trying,

  • that we are therefore on the verge of cracking all

  • of these other problems that go into what constitutes

  • general intelligence.

  • DIANE GREENE: Although we have gone

  • way faster than either of you ever expected us to go,

  • I believe.

  • FEI-FEI LI: Yes and no.

  • Humanity has a tendency to overestimate

  • short-term progress and underestimate

  • long-term progress.

  • So eventually, we will be achieving things that we cannot

  • dream of.

  • But Diane and Greg, I want to just give a simple example

  • to define AGI.

  • So the definition of AGI, again, is an introspective definition

  • of what humans and human intelligence can do.

  • I have a two-year-old daughter who doesn't like napping.

  • And I thought I'm smart enough to scheme

  • to put her in a very complicated sleeping bag that doesn't

  • get herself out of the crib.

  • And just a couple of months ago, I

  • was on the monitor watching this kid, two-year-old,

  • where for the first time I was training her

  • for napping by herself.

  • She was very angry.

  • So she looked around, figured out a weak spot on the crib

  • where she might be able to climb out,

  • figured out how to unzip her complicated sleeping

  • bag that I thought I schemed to try to prevent that,

  • and figured out a way to climb out

  • of a crib that's way taller than who

  • she is and managed to escape safely

  • and without breaking her legs.

  • DIANE GREENE: OK, how about AGI equivalent to my cat

  • or equivalent to a mouse?

  • FEI-FEI LI: If you're shifting the definition, sure.

  • DIANE GREENE: I see, OK.

  • FEI-FEI LI: But even cat, I think

  • there are things that a cat is capable of doing.

  • GREG CORRADO: So I do think that if you

  • look at an organism like a cat from a behavioral level,

  • like how cats behave and how they respond

  • to their environments, I think that you could imagine

  • a world where you have something like a toy that

  • is for entertainment purposes that approximates

  • a cat in a bunch of ways in that the sorts of behaviors

  • that the human observe, you're like, oh, it walks around.

  • It doesn't bump into things.

  • It meows at me every once in a while.

  • I do believe that we can build a system like that.

  • But what you can't do is you can't take that robot

  • and then dump it in the forest and have it figure out

  • what it needs to do in order to survive and make things work.

  • FEI-FEI LI: But it's a goal.

  • It's a healthy goal.

  • DIANE GREENE: It's a healthy goal.

  • And along the way, at least we all three agree

  • that AI's capacity to help us solve all our big problems

  • is going to outweigh any kind of negative,

  • and we're pretty excited about that, I guess.

  • In Cloud, you're kind of doing some cool things with AutoML

  • and so forth.

  • FEI-FEI LI: Yeah, so we talk a lot,

  • Diane, about the belief of building benevolent technology

  • for human use.

  • Our technology reflect our values.

  • So I personally, and I know Greg's whole team is working

  • on bringing AI to people and to the fields that really need it

  • to make a positive difference.

  • So at Cloud, we're very lucky to be working with customers

  • and partners from all kinds of vertical industries,

  • from health care where we collaborate,

  • to agriculture, to sustainability,

  • to entertainment, to retail, to commerce, to finance, where

  • our customers bring some of the toughest problem and their pain

  • points, and we can work with them hand-in-hand

  • to solve some of that.

  • So for example, recently we rolled out AutoML.

  • And that is the recognition of the pain

  • of entering machine learning.

  • It's still a highly technical field.

  • The bar is still high.

  • Not enough people are trained experts

  • in the world of machine learning.

  • But yet our industry already has so much need

  • to tag pictures, understand imageries, just as an example,

  • in vision.

  • So how do we answer that call of need?

  • So we've worked hard and thought about this suite

  • of product called AutoML where the customer--

  • we lower the entry barrier by relieving them

  • from coding machine learning custom models themselves.

  • All they have to do is to give us

  • the kind of-- provide the kind of data and concept they need.

  • Here's an example of a ramen company in Tokyo

  • that has many shops of ramens and they

  • want to build an app that recognize the ramens

  • from different ramen stores.

  • And they give us the pictures of ramens

  • and the concepts of their store.

  • One store, two store, three.

  • And what we do is to use a technique,

  • a machine learning technique that Google and many others

  • have developed called learning to learn, and then

  • build a customized model for the customer that recognize ramens

  • for their different stores.

  • And then the customer can take that model

  • to do what they want.

  • DIANE GREENE: I can write a little C++,

  • maybe some JavaScript.

  • Could I do AutoML?

  • FEI-FEI LI: Absolutely.

  • Absolutely.

  • We're working with teams that they don't have not even C++

  • experience.

  • And we have a drag and drop interface,

  • and you can use AutoML that way.

  • GREG CORRADO: Because I really believe that there are so

  • many problems that can be solved using this technique that it's

  • critical that we share as much as possible about how

  • these things work.

  • I don't believe that these technologies should

  • live in walled gardens, but instead we

  • should develop tools that can be used

  • by everyone in the community.

  • And that's part of why we have a very aggressive open source

  • stance to our software packages, particularly in AI.

  • And that includes things like TensorFlow

  • that are available completely freely,

  • and it includes the kinds of services

  • that are available on Cloud to do the kind of compute,

  • storage, and model tuning and serving that you need to use

  • these things in practice.

  • And I think it's amazing that the same tools

  • that my applied machine learning team

  • uses to tackle problems that we're interested

  • in, those same tools are accessible to all of you

  • as well to try to solve the same problems in the same way.

  • And I've been really excited with how great the uptake is

  • and how we're seeing expanding to other languages.

  • Mentioning JavaScript.

  • Quick plug for tensorflow.js is actually really awesome.

  • DIANE GREENE: Oh, and you should probably run it on a TPU.

  • GREG CORRADO: Yes, exactly.

  • DIANE GREENE: It does give a nice boost.

  • So you're building, I mean, with machine learning,

  • we're bringing it to market in so many ways,

  • because we have the tools to build your own models,

  • the TensorFlow.

  • We have the AutoML that brings it to any programmer.

  • And then what's going on with all the APIs,

  • and how is that going to affect every industry,

  • and what do you see going on there?

  • FEI-FEI LI: So Cloud already has a suite

  • of APIs for a lot of our industry partners

  • and customers, from Translate to Speech to Vision.

  • DIANE GREENE: Which are based on models we build.

  • FEI-FEI LI: Yes.

  • For example, Box is a major partner

  • with Google Cloud where they recognize a tremendous need

  • for organizing customers' imagery data to help customers.

  • So they actually use Google's Vision API to do that.

  • And that's a model easily delivered to our customers

  • through our service.

  • DIANE GREENE: Yeah, it's pretty exciting.

  • I mean, Greg, how do you think that's going to play out

  • in the health industry?

  • I know you've been thinking about that.

  • GREG CORRADO: So health care is one of the problems

  • that a bunch of people are working on at Google,

  • and a lot of people are working on outside as well, because I

  • think there's a huge opportunity to use these technologies

  • to expand the availability and the accuracy of health care.

  • And part of that is because doctors today are basically

  • trying to weather an information hurricane in order

  • to provide care.

  • And so I think there are thousands

  • of individual opportunities to make doctors' work more fluid,

  • to build tools to solve problems that they want solved,

  • and to do things that help patients

  • and improve patient care.

  • DIANE GREENE: I mean, I think you were telling me

  • that so many doctors are so unhappy because they

  • have so much drudgery to do.

  • Is this a big breakthrough?

  • GREG CORRADO: Yeah, absolutely.

  • I mean, I believe that there's been a great--

  • when you go to a doctor, you're looking for medical attention.

  • And right now a huge amount of their attention

  • is not actually focused on the practice of medicine,

  • but is focused on a whole bunch of other work

  • that they have to do that doesn't require

  • the kind of insights and care and connection

  • the real practice of medicine does.

  • And so I believe that machine learning and AI

  • is going to come in for health care

  • through assistive technologies that help the doctors do

  • what they want to do better.

  • DIANE GREENE: By understanding what they do in a system.

  • No substitute for the humans.

  • GREG CORRADO: No.

  • FEI-FEI LI: No substitutes.

  • DIANE GREENE: Speaking of human, Fei-Fei,

  • do you want to talk a little bit about why

  • you think this humanistic AI approach is so critical?

  • FEI-FEI LI: Yeah.

  • Thank you.

  • So if we look at the history of AI, we've entered phase two.

  • The first 60 years is AI as more or less a niche technical field

  • where we're still laying down scientific foundations.

  • But starting this point on, AI is

  • one of the biggest drivers of societal changes to come.

  • So how do we think about AI in the next phase?

  • What is the frame of mind that should be driving

  • us has been on top of my mind.

  • And I think deeply about the need for human-centered AI,

  • which in my opinion, includes three elements to complete

  • the human-center AI thinking.

  • The first element is really advancing AI to the next stage.

  • And here we bring our collective background

  • from neuroscience, cognitive science.

  • Whether we're getting to AGI tomorrow or in 50 years,

  • there is a need for AI to be a lot more flexible, nuanced,

  • learn faster, and more unsupervised,

  • semi-supervised [INAUDIBLE] learning ways

  • to be able to understand emotion,

  • to be able to communicate with humans.

  • So that is the more human-centered way

  • of advancing AI science.

  • The second part is the human-center AI technology

  • and application is that I love what you're saying that there's

  • no substitute for humans.

  • This technology, like all technology,

  • is to enhance humans, to augment humans, not to replace humans.

  • We'll replace certain tasks.

  • We'll replace humans out of danger or tasks that we cannot

  • perform.

  • But the bottom line is we can use AI to help our doctors,

  • to help our disaster relief workers,

  • to help decision makers.

  • So there is a lot of technology in robotics,

  • in design, in natural language processing that

  • is centered around human-centered AI

  • technology and application.

  • The third element of human-centered AI

  • is really to combine the thinking of AI

  • as a technology as well as the societal impact.

  • We are so nascent in seeing the impact of this technology.

  • But already, like Diane said, that we

  • are seeing the impact in different ways, ways

  • that we might not even predict.

  • So I think it's really important.

  • And it's a responsibility of everyone

  • from academia to industry to government

  • to bring social scientists, philosophers,

  • law scholars, policy makers, ethicists,

  • and historians at the table and to study more deeply about AI's

  • social and humanistic impact.

  • And that is the three elements of human-centered AI.

  • DIANE GREENE: That's pretty wonderful.

  • And I think we at Google here, Alphabet, are working as hard

  • as we can to do humanistic AI.

  • You mentioned what we need to be careful about out there

  • with AI and regulatory.

  • What are some of the barriers to--

  • I think every company in the world

  • has a use for AI in many, many ways.

  • I mean, it's just exploding in all the verticals.

  • But there are some impediments to adoption.

  • For example, in the financial industry

  • they need to have something called explainable AI.

  • And could you just talk about some of the different barriers

  • you see to being able to take advantage of AI?

  • FEI-FEI LI: We should start with health care.

  • GREG CORRADO: Yeah, so I think that there

  • are a bunch of really important things to consider.

  • So one of the things is, of course, we

  • want to have machine learning systems that

  • are designed to fit the needs of the folks that are

  • using them and applying them.

  • And that can often include not just giving me the answer,

  • but telling me something about how that was derived.

  • So some kind of explainability.

  • So in the health care space, for example,

  • we've been working on a bunch of things in medical imaging.

  • And it's not acceptable to just tell the doctor that,

  • oh, something looks fishy in this x-ray

  • or this pathology slide or this retinal scan.

  • You have to tell them, well, what do you think is wrong?

  • But more importantly, you actually

  • have to show them where in the image

  • you think the evidence for that conclusion

  • lies so that they can then look at it

  • and decide whether they concur or they disagree

  • or, oh, well, there was a speck of dust there

  • and that's what the machine is picking up on.

  • And the good news is that these things actually are possible.

  • And I think there's kind of been this unfortunate mythology

  • that AI and deep learning in particular is a black box.

  • And it really isn't.

  • We didn't study how it worked, because for a long time

  • it really didn't work that well.

  • But now that it's working well, there

  • are a lot of tools and techniques

  • that go into examining how these systems work.

  • And I think explainability is a big part of it

  • in terms of making these things available for a bunch

  • of applications.

  • FEI-FEI LI: So in addition to the explainability,

  • I would add bias.

  • I think bias is an issue we need to address in AI.

  • And I see bias, from where I sit, two major kind of bias

  • we need to address.

  • One is the pipeline of AI development,

  • starting from the bias of the data

  • to the outcome of the bias.

  • And we have heard a lot about if the machine learning

  • algorithm is fed with data that does not represent the problem

  • domain in a fair way, we will introduce bias.

  • Whether it's missing a group of people's data

  • or biasing it to a skewed distribution,

  • those are things that would have deep consequences,

  • whether you're in the health care domain or finance

  • or legal decision making.

  • So I think that is a huge issue very nicely that Google

  • is already addressing that.

  • We have a whole team at Google working on bias.

  • DIANE GREENE: Yeah.

  • That's true.

  • FEI-FEI LI: And another bias I think is important

  • is the people who are developing AIs.

  • The human bias and the lack of diversity is also another bias.

  • DIANE GREENE: It's so important.

  • And that kind of brings me to maybe some of our--

  • we're getting close to the end.

  • But where is AI going?

  • I mean, how prevalent is it going to be?

  • I mean, we look at our universities and these machine

  • learning classes have 800 people, 900 people.

  • There is such a demand.

  • Every computer science graduate wants to know it.

  • Where is it going?

  • I mean, will every high school graduating senior

  • be able to customize AI to their own purposes?

  • And what does it look like five, 10 years from now?

  • FEI-FEI LI: So from a technology point of view,

  • I think that because of the tremendous investment

  • in resource, both in the private sector

  • as well in the public sector now,

  • many countries are waking up to investing AI,

  • we're going to see a huge continue development

  • of AI technology.

  • I'm mostly excited either at Cloud

  • or seeing what Greg's team is doing,

  • AI being delivered to the industries that really

  • matter to people's lives and the work quality and productivity.

  • But Diane, I think you're also asking

  • is how are we educating more people in AI?

  • DIANE GREENE: Both making it easier to use

  • and educating them and what's it going to look like?

  • What do you predict?

  • FEI-FEI LI: That's a really tough question,

  • because at the core of today's AI is still calculus.

  • And that's not going to change.

  • GREG CORRADO: So I think that from the kind of tech industry

  • perspective or from the computer science education perspective,

  • I think that we're going to see AI and ML become

  • as essential as networking is.

  • No one really thinks about, oh, well,

  • I'm going to write some software and it's

  • going to be standalone on a box and it's not going

  • to have a TCPI connection.

  • We all know that you're going to have

  • a TCPI connection at the end of the day somewhere.

  • And everyone understands the basics of the networking stack.

  • And that's not just at the level of engineers.

  • That's the level of designers, of executives,

  • of product developers and leaders.

  • And the same thing, I think, is going

  • to happen with machine learning and AI, which

  • is that designers are going to start to understand, how can I

  • make a completely revolutionary kind of product that folds

  • in machine learning the same way that we fold in networking

  • and internet technologies into almost everything we build?

  • So I think we're going to see tremendous uptake

  • and it becoming kind of a pervasive background

  • part of the technologies.

  • But I think in that process the ways

  • that we use AI are going to evolve.

  • So I think right now we're seeing

  • a lot of things where AI and machine learning

  • add some spice, some extra, a little coolness on a feature.

  • And I think that what you're going to see over

  • the next decade is you're going to see more

  • of a core integration into what it means for the product

  • to actually work.

  • And I think that one of the great opportunities

  • there is actually going to be the development

  • of artificial emotional intelligence

  • that allows products to actually have much more natural and much

  • more fluid human interaction.

  • We're beginning to see that in the Assistant now with speech

  • recognition, speech synthesis, understanding

  • dialogues and exchanges.

  • But I think that this is still in its infancy.

  • We're going to get to a point where the products

  • that we build, they interact with humans in the way

  • that the humans find most useful just out of the box.

  • FEI-FEI LI: And I spend a lot of time with high schoolers,

  • because I really believe in the future.

  • We always talk about AI changing the world.

  • And I always say the question is, who is changing AI?

  • And to me, bringing more human mission thinking

  • into technology development and thought leadership

  • is really important.

  • Not only important for the future

  • of our technology and the value we instill in our technology,

  • but also in bringing the diverse group of students

  • and future leaders into the development of AI.

  • So at [? Server ?] at Google, we all work a lot on this issue.

  • And personally, I'm very involved

  • with AI4ALL, which is a nonprofit that

  • educates high schoolers around the country

  • from diverse backgrounds, whether they're

  • girls or students of underrepresented minority

  • groups.

  • And we bring them onto university campus

  • and work with them on AI thinking and AI studies.

  • DIANE GREENE: And at Google, we're

  • just completely committed to bringing all our best

  • technologies to everybody in the world.

  • And we're doing that through the cloud,

  • and we're bringing these tools, we're

  • bringing these APIs and the training

  • and the partnering and the processors.

  • And we're pretty excited to see what all you

  • guys are going to do with it.

  • Thank you very much.

  • GREG CORRADO: Thanks, everybody.

  • [MUSIC PLAYING]

[MUSIC PLAYING]

Subtitles and vocabulary

Click the word to look it up Click the word to find further inforamtion about it