back to indexAyanna Howard: Human-Robot Interaction & Ethics of Safety-Critical Systems | Lex Fridman Podcast #66
link |
The following is a conversation with Ayanna Howard.
link |
She's a roboticist, Professor Georgia Tech,
link |
and Director of the Human Automation Systems Lab.
link |
With research interests in human robot interaction,
link |
assistive robots in the home, therapy gaming apps,
link |
and remote robotic exploration of extreme environments.
link |
Like me, in her work, she cares a lot
link |
about both robots and human beings.
link |
And so I really enjoyed this conversation.
link |
This is the Artificial Intelligence Podcast.
link |
If you enjoy it, subscribe on YouTube,
link |
give it five stars on Apple Podcast,
link |
follow on Spotify, support it on Patreon,
link |
or simply connect with me on Twitter.
link |
Alex Friedman, spelled F R I D M A N.
link |
I recently started doing ads at the end of the introduction.
link |
I'll do one or two minutes after introducing the episode
link |
and never any ads in the middle
link |
that can break the flow of the conversation.
link |
I hope that works for you
link |
and doesn't hurt the listening experience.
link |
This show is presented by Cash App,
link |
the number one finance app in the App Store.
link |
I personally use Cash App to send money to friends,
link |
but you can also use it to buy, sell,
link |
and deposit Bitcoin in just seconds.
link |
Cash App also has a new investing feature.
link |
You can buy fractions of a stock, say $1 worth,
link |
no matter what the stock price is.
link |
Broker services are provided by Cash App Investing,
link |
a subsidiary of Square and member SIPC.
link |
I'm excited to be working with Cash App
link |
to support one of my favorite organizations called First,
link |
best known for their first robotics and Lego competitions.
link |
They educate and inspire hundreds of thousands of students
link |
in over 110 countries
link |
and have a perfect rating and charity navigator,
link |
which means the donated money
link |
is used to maximum effectiveness.
link |
When you get Cash App from the App Store, Google Play,
link |
and use code LEX Podcast, you'll get $10,
link |
and Cash App will also donate $10 the first,
link |
which again, is an organization
link |
that I've personally seen inspire girls and boys
link |
to dream of engineering a better world.
link |
And now, here's my conversation with Ayanna Howard.
link |
What or who is the most amazing robot you've ever met,
link |
or perhaps had the biggest impact on your career?
link |
I haven't met her, but I grew up with her,
link |
but of course, Rosie.
link |
So, and I think it's because also...
link |
Rosie from the Jetsons.
link |
She is all things to all people, right?
link |
Think about it, like anything you wanted,
link |
it was like magic, it happened.
link |
So people not only anthropomorphize,
link |
but project whatever they wish for the robot to be onto.
link |
But also, I mean, think about it,
link |
she was socially engaging.
link |
She, every so often had an attitude, right?
link |
She kept us honest.
link |
She would push back sometimes
link |
when George was doing some weird stuff.
link |
But she cared about people, especially the kids.
link |
She was like the perfect robot.
link |
And you've said that people don't want
link |
the robots to be perfect.
link |
Can you elaborate that?
link |
What do you think that is?
link |
Just like you said,
link |
Rosie pushed back a little bit every once in a while.
link |
Yeah, so I think it's that.
link |
So if you think about robotics in general,
link |
we want them because they enhance our quality of life.
link |
And usually that's linked to something that's functional,
link |
Even if you think of self driving cars,
link |
why is there a fascination?
link |
Because people really do hate to drive.
link |
Like there's the, like Saturday driving
link |
where I can just speed,
link |
but then there was the,
link |
I have to go to work every day
link |
and I'm in traffic for an hour.
link |
I mean, people really hate that.
link |
And so robots are designed to basically enhance
link |
our ability to increase our quality of life.
link |
And so the perfection comes from this aspect of interaction.
link |
If I think about how we drive,
link |
if we drove perfectly, we would never get anywhere, right?
link |
So think about how many times you had to run past the light
link |
because you see the car behind you
link |
is about to crash into you.
link |
Or that little kid kind of runs into the street
link |
and so you have to cross on the other side
link |
because there's no cars, right?
link |
Like if you think about it,
link |
we are not perfect drivers.
link |
Some of it is because it's our world.
link |
And so if you have a robot that is perfect
link |
in that sense of the word,
link |
they wouldn't really be able to function with us.
link |
Can you linger a little bit on the word perfection?
link |
So from the robotics perspective,
link |
what does that word mean?
link |
And how is sort of the optimal behaviors
link |
you're describing different than what we think is perfection?
link |
Yeah, so perfection, if you think about it
link |
in the more theoretical point of view,
link |
it's really tied to accuracy, right?
link |
So if I have a function,
link |
can I complete it at 100% accuracy with zero errors?
link |
And so that's kind of if you think about perfection
link |
in the sense of the word.
link |
And in a self driving car realm,
link |
do you think from a robotics perspective,
link |
we kind of think that perfection means
link |
following the rules perfectly,
link |
sort of defining, staying in the lane, changing lanes.
link |
When there's a green light, you go,
link |
when there's a red light, you stop.
link |
and be able to perfectly see all the entities in the scene.
link |
That's the limit of what we think of as perfection.
link |
And I think that's where the problem comes,
link |
is that when people think about perfection for robotics,
link |
the ones that are the most successful
link |
are the ones that are, quote unquote, perfect.
link |
Like I said, Rosie is perfect,
link |
but she actually wasn't perfect in terms of accuracy,
link |
but she was perfect in terms of how she interacted
link |
and how she adapted.
link |
And I think that's some of the disconnect,
link |
is that we really want perfection
link |
with respect to its ability to adapt to us.
link |
We don't really want perfection with respect to 100% accuracy
link |
with respect to the rules that we just made up anyway, right?
link |
And so I think there's this disconnect sometimes
link |
between what we really want and what happens.
link |
And we see this all the time, like in my research, right?
link |
Like the optimal, quote unquote, optimal interactions
link |
are when the robot is adapting based on the person,
link |
not 100% following what's optimal based on the rules.
link |
Just a link around autonomous vehicles for a second.
link |
Just your thoughts maybe off the top of the head
link |
is how hard is that problem,
link |
do you think based on what we just talked about?
link |
You know, there's a lot of folks
link |
in the automotive industry, they're very confident
link |
from Elon Musk to Waymo to all these companies.
link |
How hard is it to solve that last piece?
link |
The gap between the perfection
link |
and the human definition
link |
of how you actually function in this world.
link |
Yeah, so this is a moving target.
link |
So I remember when all the big companies
link |
started to heavily invest in this.
link |
And there was a number of even roboticists
link |
as well as folks who were putting in the VCs
link |
and corporations, Elon Musk being one of them,
link |
that said, self driving cars on the road with people
link |
within five years, that was a little while ago.
link |
And now people are saying five years, 10 years, 20 years,
link |
some are saying never, right?
link |
I think if you look at some of the things
link |
that are being successful is these
link |
basically fixed environments
link |
where you still have some anomalies, right?
link |
You still have people walking, you still have stores,
link |
but you don't have other drivers, right?
link |
Like other human drivers or is a dedicated space
link |
Because if you think about robotics in general,
link |
where has always been successful?
link |
I mean, you can say manufacturing,
link |
like way back in the day, right?
link |
It was a fixed environment, humans were not part
link |
of the equation, we're a lot better than that.
link |
But like when we can carve out scenarios
link |
that are closer to that space,
link |
then I think that it's where we are.
link |
So a closed campus where you don't have self driving cars
link |
and maybe some protection so that the students
link |
don't jet in front just because they wanna see what happens.
link |
Like having a little bit, I think that's where
link |
we're gonna see the most success in the near future.
link |
And be slow moving.
link |
Right, not 55, 60, 70 miles an hour,
link |
but the speed of a golf cart, right?
link |
So that said, the most successful
link |
in the automotive industry robots operating today
link |
in the hands of real people are ones
link |
that are traveling over 55 miles an hour.
link |
And in uncle's trains environment,
link |
which is Tesla vehicles, so the Tesla autopilot.
link |
So I would love to hear sort of your,
link |
just thoughts of two things.
link |
So one, I don't know if you've gotten to see
link |
you've heard about something called smart summon.
link |
It would Tesla system autopilot system
link |
where the car drives zero occupancy, no driver
link |
in the parking lot slowly sort of tries to navigate
link |
the parking lot to find itself to you.
link |
And there's some incredible amounts of videos
link |
and just hilarity that happens.
link |
Is it awkwardly tries to navigate this environment?
link |
But it's a beautiful nonverbal communication
link |
between machine and human that I think is a from,
link |
it's like, it's some of the work that you do
link |
in this kind of interesting human robot interaction space.
link |
So what are your thoughts in general about it?
link |
So I do have that feature.
link |
Do you drive a Tesla?
link |
I do, mainly because I'm a gadget freak, right?
link |
So I say it's a gadget that happens to have some wheels.
link |
And yeah, I've seen some of the videos.
link |
But what's your experience like?
link |
I mean, you're a human robot interaction roboticist,
link |
you're a legit sort of expert in the field.
link |
So what does it feel for a machine to come to you?
link |
It's one of these very fascinating things,
link |
but also I am hyper, hyper alert, right?
link |
Like I'm hyper alert, like my, but my thumb is like,
link |
oh, okay, I'm ready to take over.
link |
Even when I'm in my car or I'm doing things
link |
like automated backing into,
link |
so there's like a feature where you can do this
link |
automating backing into our parking space
link |
or bring the car out of your garage
link |
or even, you know, pseudo autopilot on the freeway, right?
link |
I am hyper sensitive.
link |
I can feel like as I'm navigating, I'm like,
link |
yeah, that's an error right there.
link |
Like I am very aware of it,
link |
but I'm also fascinated by it and it does get better.
link |
Like I look and see it's learning
link |
from all of these people who are cutting it on.
link |
Like every time I come on, it's getting better, right?
link |
And so I think that's what's amazing about it is that.
link |
This nice dance of you're still hyper vigilant.
link |
So you're still not trusting it at all.
link |
And yet you're using it on the highway if I were to,
link |
like what, as a roboticist,
link |
we'll talk about trust a little bit.
link |
How do you explain that?
link |
Is it the gadget freak part?
link |
Like where you just enjoy exploring technology
link |
or is that the right actually balance
link |
between robotics and humans is where you use it,
link |
but don't trust it.
link |
And somehow there's this dance
link |
that ultimately is a positive.
link |
Yeah. So I think I'm,
link |
I just don't necessarily trust technology,
link |
but I'm an early adopter, right?
link |
So when it first comes out, I will use everything,
link |
but I will be very, very cautious of how I use it.
link |
Do you read about it or do you explore it, but just try it?
link |
Do you like crudely, to put it crudely,
link |
do you read the manual or do you learn through exploration?
link |
If I have to read the manual, then I do design,
link |
then it's a bad user interface, it's a failure.
link |
Elon Musk is very confident that you kind of take it
link |
from where it is now to full autonomy.
link |
So from this human robot interaction
link |
where you don't really trust and then you try
link |
and then you catch it when it fails to,
link |
it's going to incrementally improve itself
link |
into full, full way you don't need to participate.
link |
What's your sense of that trajectory?
link |
So the promise there is by the end of next year,
link |
by the end of 2020 is the current promise.
link |
What's your sense about that journey that Tesla's on?
link |
So there's kind of three things going on though.
link |
I think in terms of will people go,
link |
like as a user, as a adopter,
link |
will you trust going to that point?
link |
I think so, right?
link |
Like there are some users and it's because what happens
link |
is when you're hypersensitive at the beginning
link |
and then the technology tends to work,
link |
your apprehensions slowly goes away.
link |
And as people, we tend to swing to the other extreme, right?
link |
Because like, oh, I was like hyper, hyper fearful
link |
or hypersensitive and it was awesome.
link |
And we just tend to swing.
link |
That's just human nature.
link |
And so you will have, I mean,
link |
That's a scary notion because most people
link |
are now extremely untrusting of autopilot.
link |
They use it, but they don't trust it.
link |
And it's a scary notion that there's a certain point
link |
where you allow yourself to look at the smartphone
link |
for like 20 seconds.
link |
And then there'll be this phase shift
link |
where it'll be like 20 seconds, 30 seconds,
link |
one minute, two minutes.
link |
It's a scary proposition.
link |
But that's people, right?
link |
That's just humans.
link |
I mean, I think of even our use of,
link |
I mean, just everything on the internet, right?
link |
Like think about how reliant we are on certain apps
link |
and certain engines, right?
link |
20 years ago, people have been like,
link |
oh yeah, that's stupid.
link |
Like that makes no sense.
link |
Like of course that's false.
link |
Like now it's just like, oh, of course I've been using it.
link |
It's been correct all this time.
link |
Of course, aliens, I didn't think they existed,
link |
but now it says they do, obviously.
link |
100% Earth is flat.
link |
So, okay, but you said three things.
link |
So one is the human.
link |
Okay, so one is the human.
link |
And I think there will be a group of individuals
link |
that will swing, right?
link |
Teenage, I mean, it'll be adults.
link |
There's actually an age demographic
link |
that's optimal for technology adoption.
link |
And you can actually find them.
link |
And they're actually pretty easy to find.
link |
Just based on their habits, based on...
link |
So someone like me who wasn't a roboticist
link |
would probably be the optimal kind of person, right?
link |
Early adopter, okay with technology,
link |
very comfortable and not hypersensitive, right?
link |
I'm just hypersensitive because I designed this stuff.
link |
So there is a target demographic that will swing.
link |
The other one though is you still have these humans
link |
that are on the road.
link |
That one is a harder thing to do.
link |
And as long as we have people that are on the same streets,
link |
that's gonna be the big issue.
link |
And it's just because you can't possibly,
link |
you can't possibly map some of the silliness
link |
of human drivers, right?
link |
Like as an example, when you're next to that car
link |
that has that big sticker called student driver, right?
link |
Like you are like, oh, either I am going to like go around.
link |
Like we are, we know that that person
link |
is just gonna make mistakes that make no sense, right?
link |
How do you map that information?
link |
Or if I am in a car and I look over
link |
and I see two fairly young looking individuals
link |
and there's no student driver bumper
link |
and I see them chatting to each other,
link |
I'm like, oh, that's an issue, right?
link |
So how do you get that kind of information
link |
and that experience into basically an autopilot?
link |
Yeah, and there's millions of cases like that
link |
where we take little hints to establish context.
link |
I mean, you said kind of beautifully poetic human things,
link |
but there's probably subtle things about the environment,
link |
about it being maybe time for commuters
link |
to start going home from work.
link |
And therefore you can make some kind of judgment
link |
about the group behavior of pedestrians,
link |
blah, blah, blah, so on and so on.
link |
Are even cities, right?
link |
Like if you're in Boston, how people cross the street,
link |
like lights are not an issue versus other places
link |
where people will actually wait for the crosswalk.
link |
Seattle or somewhere peaceful.
link |
And but what I've also seen, so just even in Boston
link |
that intersection to intersection is different.
link |
So every intersection has a personality of its own.
link |
So certain neighborhoods of Boston are different.
link |
So we kind of end based on different timing of day
link |
at night, it's all, there's a dynamic to human behavior
link |
that we kind of figure out ourselves.
link |
We're not able to introspect and figure it out,
link |
but somehow our brain learns it.
link |
And so you're saying, is there a shortcut?
link |
Is there a shortcut though for a robot?
link |
Is there something that could be done?
link |
You think that, you know, that's what we humans do.
link |
It's just like bird flight, right?
link |
This example they give for flight.
link |
Do you necessarily need to build a bird that flies
link |
or can you do an airplane?
link |
So is there a shortcut to it?
link |
So I think that the shortcut is,
link |
and I kind of, I talk about it as a fixed space.
link |
Where, so imagine that there's a neighborhood
link |
that's a new smart city or a new neighborhood that says,
link |
you know what, we are going to design this new city
link |
based on supporting self driving cars.
link |
And then doing things, knowing that there's anomalies,
link |
knowing that people are like this, right?
link |
And designing it based on that assumption
link |
that like we're gonna have this,
link |
that would be an example of a shortcut.
link |
So you still have people, but you do very specific things
link |
to try to minimize the noise a little bit.
link |
And the people themselves become accepting
link |
of the notion that there's autonomous cars, right?
link |
Right, like they move into,
link |
so right now you have like a,
link |
you will have a self selection bias, right?
link |
Like individuals will move into this neighborhood
link |
knowing like this is part of like the real estate pitch, right?
link |
And so I think that's a way to do a shortcut.
link |
When it allows you to deploy,
link |
it allows you to collect then data with these variances
link |
and anomalies, cause people are still people,
link |
but it's a safer space and is more of an accepting space.
link |
IE when something in that space might happen
link |
because things do, because you already have
link |
the self selection, like people would be,
link |
I think a little more forgiving than other places.
link |
And you said three things, did we cover all of them?
link |
The third is legal law, liability,
link |
which I don't really want to touch,
link |
but it's still of concern.
link |
And the mishmash with like, with policy as well,
link |
sort of government, all that, that whole.
link |
That big ball of stuff.
link |
So that's, so we're out of time now.
link |
Do you think from a robotics perspective,
link |
you know, if you're kind of honest with what cars do,
link |
they kind of threaten each other's life all the time.
link |
So cars are very, I mean, in order to navigate intersections,
link |
there's an assertiveness, there's a risk taking,
link |
and if you were to reduce it to an objective function,
link |
there's a probability of murder in that function,
link |
meaning you killing another human being,
link |
and you're using that.
link |
First of all, it has to be low enough
link |
to be acceptable to you on an ethical level,
link |
as an individual human being,
link |
but it has to be high enough for people to respect you,
link |
to not sort of take advantage of you completely,
link |
and jaywalk in front of you, and so on.
link |
So, I mean, I don't think there's a right answer here,
link |
but how do we solve that?
link |
How do we solve that from a robotics perspective
link |
when danger and human life is at stake?
link |
Yeah, as they say, cars don't kill people,
link |
people kill people.
link |
Kill people, kill people.
link |
Right, so I think.
link |
And now robotic algorithms would be killing people.
link |
Right, so it will be robotics algorithms that are,
link |
no, it will be robotic algorithms don't kill people,
link |
developers of robotic algorithms kill people, right?
link |
I mean, one of the things is people are still in the loop,
link |
and at least in the near and midterm,
link |
I think people will still be in the loop.
link |
At some point, even if it's the developer,
link |
like we're not necessarily at the stage
link |
where robots are programming autonomous robots
link |
with different behaviors quite yet.
link |
That's a scary notion, sorry, to interrupt,
link |
that a developer has some responsibility
link |
in the death of a human being.
link |
That's a heavy burden.
link |
I mean, I think that's why the whole aspect of ethics
link |
in our community is so, so important, right?
link |
Like, because it's true, if you think about it,
link |
you can basically say,
link |
I'm not going to work on weaponized AI, right?
link |
Like, people can say, that's not what I'm gonna do.
link |
But yet, you are programming algorithms
link |
that might be used in healthcare algorithms
link |
that might decide whether this person
link |
should get this medication or not,
link |
and they don't, and they die.
link |
Okay, so that is your responsibility, right?
link |
And if you're not conscious and aware
link |
that you do have that power when you're coding
link |
and things like that,
link |
I think that's just not a good thing.
link |
Like, we need to think about this responsibility
link |
as we program robots and computing devices
link |
much more than we are.
link |
Yeah, so it's not an option to not think about ethics.
link |
I think it's a majority, I would say, of computer science.
link |
Sort of, it's kind of a hot topic now,
link |
I think about bias and so on,
link |
but it's, and we'll talk about it,
link |
but usually it's kind of,
link |
it's like a very particular group of people
link |
that work on that.
link |
And then, people who do robotics are like,
link |
well, I don't have to think about that.
link |
There's other smart people thinking about it.
link |
It seems that everybody has to think about it.
link |
It's not, you can't escape the ethics,
link |
whether it's bias or just every aspect of ethics
link |
that has to do with human beings.
link |
So think about, I'm gonna age myself,
link |
but I remember when we didn't have like testers, right?
link |
And so what did you do?
link |
As a developer, you had to test your own code, right?
link |
Like you had to go through all the cases
link |
and figure it out and, you know,
link |
and then they realized that, you know,
link |
like we probably need to have testing
link |
because we're not getting all the things.
link |
And so from there, what happens is like most developers,
link |
they do, you know, a little bit of testing,
link |
but it's usually like, okay, did my compiler bug out?
link |
Let me look at the warnings.
link |
Okay, is that acceptable or not?
link |
Like that's how you typically think about as a developer
link |
and you're just assume that is going to go
link |
to another process and they're gonna test it out.
link |
But I think we need to go back to those early days
link |
when, you know, you're a developer, you're developing.
link |
There should be like this a, you know, okay,
link |
let me look at the ethical outcomes of this
link |
because there isn't a second like testing ethical testers,
link |
We did it back in the early coding days.
link |
I think that's where we are with respect to ethics.
link |
Like let's go back to what was good practices
link |
and only because we were just developing the field.
link |
Yeah, and it's a really heavy burden.
link |
I've had to feel it recently in the last few months,
link |
but I think it's a good one to feel like I've gotten
link |
a message more than one from people, you know,
link |
I've unfortunately gotten some attention recently
link |
and I've gotten messages that say that I have blood
link |
in my hands because of working on semi autonomous vehicles.
link |
So the idea that you have semi autonomy means people
link |
would become, would lose vigilance and so on.
link |
That's actually be humans as we described.
link |
And because of that, because of this idea
link |
that we're creating automation,
link |
there'll be people be hurt because of it.
link |
And I think that's a beautiful thing.
link |
I mean, it's, you know, there's many nights
link |
where I wasn't able to sleep because of this notion.
link |
You know, you really do think about people that might die
link |
because of this technology.
link |
Of course, you can then start rationalizing and saying,
link |
well, you know what, 40,000 people die
link |
in the United States every year
link |
and we're trying to ultimately try to save lives.
link |
But the reality is your code you've written
link |
might kill somebody and that's an important burden
link |
to carry with you as you design the code.
link |
I don't even think of it as a burden
link |
if we train this concept correctly from the beginning.
link |
And I use, and not to say that coding is like
link |
being a medical doctor, but think about it.
link |
Medical doctors, if they've been in situations
link |
where their patient didn't survive, right?
link |
Do they give up and go away?
link |
No, every time they come in,
link |
they know that there might be a possibility
link |
that this patient might not survive.
link |
And so when they approach every decision,
link |
like that's in the back of their head.
link |
And so why isn't that we aren't teaching,
link |
and those are tools though, right?
link |
They are given some of the tools to address that
link |
so that they don't go crazy.
link |
But we don't give those tools
link |
so that it does feel like a burden
link |
versus something of I have a great gift
link |
and I can do great, awesome good,
link |
but with it comes great responsibility.
link |
I mean, that's what we teach in terms of,
link |
you think about the medical schools, right?
link |
Great gift, great responsibility.
link |
I think if we just change the messaging a little,
link |
great gift being a developer, great responsibility.
link |
And this is how you combine those.
link |
But do you think, I mean, this is really interesting.
link |
It's outside, I actually have no friends
link |
who are sort of surgeons or doctors.
link |
I mean, what does it feel like
link |
to make a mistake in a surgery and somebody to die
link |
Like is that something you could be taught
link |
in medical school sort of how to be accepting of that risk?
link |
So because I do a lot of work with healthcare robotics,
link |
I have not lost a patient, for example.
link |
The first one's always the hardest, right?
link |
But they really teach the value, right?
link |
So they teach responsibility,
link |
but they also teach the value.
link |
Like you're saving 40,000,
link |
but in order to really feel good about that,
link |
when you come to a decision,
link |
you have to be able to say at the end,
link |
I did all that I could possibly do, right?
link |
Versus a, well, I just picked the first widget, right?
link |
Like, so every decision is actually thought through.
link |
It's not a habit, it's not a,
link |
let me just take the best algorithm
link |
that my friend gave me, right?
link |
It's a, is this it?
link |
Have I done my best to do good, right?
link |
And I think burden is the wrong word.
link |
It's a gift, but you have to treat it extremely seriously.
link |
So on a slightly related note, in a recent paper,
link |
The Ugly Truth About Ourselves and Our Robot Creations,
link |
you discuss, you highlight some biases
link |
that may affect the function of various robotic systems.
link |
Can you talk through, if you remember, examples of some?
link |
There's a lot of examples.
link |
What is bias, first of all?
link |
Yeah, so bias is this,
link |
and so bias, which is different than prejudice.
link |
So bias is that we all have these preconceived notions
link |
about particular, everything from particular groups
link |
to habits, to identity, right?
link |
So we have these predispositions.
link |
And so when we address a problem,
link |
we look at a problem and make a decision,
link |
those preconceived notions might affect our outputs,
link |
So there, the bias can be positive and negative,
link |
and then it's prejudice, the negative kind of bias?
link |
Prejudice is the negative, right?
link |
So prejudice is that not only are you aware of your bias,
link |
but you are then taken and have a negative outcome,
link |
even though you are aware, like...
link |
And there could be gray areas too.
link |
There's always gray areas.
link |
That's the challenging aspect of all ethical questions.
link |
So I always like, so there's a funny one.
link |
And in fact, I think it might be in the paper
link |
because I think I talk about self driving cars.
link |
But think about this.
link |
We, for teenagers, right?
link |
Typically, insurance companies charge quite a bit of money
link |
if you have a teenage driver.
link |
So you could say that's an age bias, right?
link |
But no one will, I mean, parents will be grumpy,
link |
but no one really says that that's not fair.
link |
That's interesting.
link |
We don't, that's right.
link |
It's everybody in human factors and safety research almost,
link |
I mean, it's quite ruthlessly critical of teenagers.
link |
And we don't question, is that okay?
link |
Is that okay to be agist in this kind of way?
link |
And it is age, right?
link |
It's definitely age, there's no question about it.
link |
And so this is the gray area, right?
link |
Because you know that teenagers are more likely
link |
to be in accidents and so there's actually some data to it.
link |
But then if you take that same example and you say,
link |
well, I'm going to make the insurance higher
link |
for an area of Boston because there's a lot of accidents.
link |
And then they find out that that's correlated
link |
with socioeconomics, well, then it becomes a problem, right?
link |
Like that is not acceptable,
link |
but yet the teenager, which is age, it's against age is, right?
link |
And the way we figure that out as society
link |
by having conversations, by having discourse,
link |
I mean, throughout history, the definition
link |
of what is ethical or not has changed
link |
and hopefully always for the better.
link |
So in terms of bias or prejudice in robotic,
link |
in algorithms, what examples do you sometimes think about?
link |
So I think about quite a bit the medical domain
link |
just because historically, right?
link |
The health care domain has had these biases,
link |
typically based on gender and ethnicity primarily,
link |
a little on age, but not so much.
link |
Historically, if you think about FDA and drug trials,
link |
it's harder to find a woman that aren't childbearing
link |
and so you may not test on drugs at the same level.
link |
Right, so there's these things.
link |
And so if you think about robotics, right?
link |
Something as simple as, I like to design an exoskeleton, right?
link |
What should the material be?
link |
What should the weight be?
link |
What should the form factor be?
link |
Are you, who are you gonna design it around?
link |
I will say that in the US,
link |
women average height and weight is slightly different
link |
than guys, so who are you gonna choose?
link |
Like, if you're not thinking about it from the beginning
link |
as, okay, when I design this and I look at the algorithms
link |
and I design the control system and the forces
link |
and the torques, if you're not thinking about,
link |
well, you have different types of body structure,
link |
you're gonna design to what you're used to.
link |
Oh, this fits in all the folks in my lab, right?
link |
So think about it from the very beginning as important.
link |
What about sort of algorithms that train on data?
link |
Data kind of thing.
link |
Sadly, our society already has a lot of negative bias.
link |
And so if we collect a lot of data,
link |
even if it's a balanced way,
link |
there's going to contain the same bias
link |
that a society contains.
link |
And so, yeah, is there things there that bother you?
link |
Yeah, so you actually said something.
link |
You had said how we have biases,
link |
but hopefully we learn from them and we become better, right?
link |
And so that's where we are now, right?
link |
So the data that we're collecting is historic.
link |
It's, so it's based on these things.
link |
When we knew it was bad to discriminate,
link |
but that's the data we have
link |
and we're trying to fix it now,
link |
but we're fixing it based on the data
link |
that was used in the first place.
link |
Right, and so the decisions,
link |
and you can look at everything from the whole aspect
link |
of predictive policing, criminal recidivism.
link |
There was a recent paper that had the healthcare algorithms,
link |
which had a kind of a sensational titles.
link |
I'm not pro sensationalism in titles,
link |
but again, you read it, right?
link |
So it makes you read it,
link |
but I'm like really like, ah, you could have...
link |
What's the topic of the sensationalism?
link |
I mean, what's underneath it?
link |
What's, if you could sort of educate me
link |
on what kind of bias creeps into the healthcare space.
link |
Yeah, so... I mean, you already kind of mentioned...
link |
Yeah, so this one was,
link |
the headline was racist AI algorithms.
link |
Okay, like, okay, that's totally a clickbait title.
link |
And so you looked at it,
link |
and so there was data that these researchers had collected.
link |
I believe I want to say it was either science or nature.
link |
It just was just published,
link |
but they didn't have a sensational title.
link |
It was like the media.
link |
And so they had looked at demographics,
link |
I believe, between black and white women, right?
link |
And they showed that there was a discrepancy
link |
in the outcomes, right?
link |
And so, and it was tied to ethnicity, tied to race.
link |
The piece that the researchers did
link |
actually went through the whole analysis, but of course...
link |
I mean, the journals with AI are problematic
link |
across the board, let's say.
link |
And so this is a problem, right?
link |
And so there's this thing about,
link |
oh, AI, it has all these problems,
link |
we're doing it on historical data,
link |
and the outcomes aren't even based on gender
link |
or ethnicity or age.
link |
But I'm always saying, it's like, yes,
link |
we need to do better, right?
link |
We need to do better.
link |
It is our duty to do better,
link |
but the worst AI is still better than us.
link |
Like, you take the best of us,
link |
and we're still worse than the worst AI,
link |
at least in terms of these things.
link |
And that's actually not discussed, right?
link |
And so I think, and that's why the sensational title, right?
link |
And so it's like, so then you can have individuals go like,
link |
oh, we don't need to use this AI.
link |
I'm like, oh, no, no, no, no.
link |
I want the AI instead of the doctors
link |
that provided that data,
link |
because it's still better than that, right?
link |
I think that's really important to linger on.
link |
Is the idea that this AI is racist,
link |
it's like, well, compared to what?
link |
Sort of the, I think we set, unfortunately,
link |
way too high of a bar for AI algorithms.
link |
And in the ethical space,
link |
where perfect is, I would argue, probably impossible.
link |
Then if we set the bar of perfection, essentially,
link |
of it has to be perfectly fair, whatever that means,
link |
is it means we're setting it up for failure.
link |
But that's really important to say what you just said,
link |
which is, well, it's still better than it is.
link |
And one of the things I think
link |
that we don't get enough credit for just in terms of,
link |
as developers, is that you can now poke at it, right?
link |
So it's harder to say, is this hospital,
link |
is this city doing something, right?
link |
Until someone brings in a civil case, right?
link |
Well, with AI, it can process
link |
through all this data and say, hey, yes,
link |
there's an issue here, but here it is, we've identified it.
link |
And then the next step is to fix it.
link |
I mean, that's a nice feedback loop,
link |
versus like waiting for someone to sue someone else
link |
before it's fixed, right?
link |
And so I think that power,
link |
we need to capitalize on a little bit more, right?
link |
Instead of having the sensational titles, have the,
link |
okay, this is a problem, and this is how we're fixing it.
link |
And people are putting money to fix it
link |
because we can make it better.
link |
I look at like facial recognition, how Joy,
link |
she basically called out a couple of companies and said,
link |
hey, and most of them were like, oh, embarrassment.
link |
And the next time it had been fixed, right?
link |
It had been fixed better, right?
link |
And then it was like, oh, here's some more issues.
link |
And I think that conversation then moves that needle
link |
to having much more of fear and unbiased and ethical aspects.
link |
As long as both sides, the developers are willing to say,
link |
okay, I hear you, yes, we are going to improve.
link |
And you have other developers who are like,
link |
hey, AI, it's wrong, but I love it, right?
link |
So speaking of this really nice notion that AI is maybe flawed,
link |
but better than humans.
link |
So just made me think of it,
link |
one example of flawed humans is our political system.
link |
Do you think, or you said judicial as well,
link |
do you have a hope for AI sort of being elected
link |
for president or running our Congress
link |
or being able to be a powerful representative of the people?
link |
So I mentioned, and I truly believe that
link |
this whole world of AI is in partnerships with people.
link |
And so what does that mean?
link |
I don't believe or maybe I just don't,
link |
I don't believe that we should have an AI for president,
link |
but I do believe that a president should use AI
link |
as an advisor, right?
link |
Like if you think about it,
link |
every president has a cabinet of individuals
link |
that have different expertise
link |
that they should listen to, right?
link |
Like that's kind of what we do.
link |
And you put smart people with smart expertise
link |
around certain issues and you listen.
link |
I don't see why AI can't function
link |
as one of those smart individuals giving input.
link |
So maybe there's an AI on healthcare,
link |
maybe there's an AI on education and right?
link |
Like all of these things that a human is processing, right?
link |
Because at the end of the day there's people that are human
link |
that are going to be at the end of the decision.
link |
And I don't think as a world, as a culture, as a society
link |
that we would totally, and this is us,
link |
like this is some fallacy about us,
link |
but we need to see that leader, that person as human.
link |
And most people don't realize that like leaders
link |
have a whole lot of advice, right?
link |
Like when they say something,
link |
it's not that they woke up,
link |
well usually they don't wake up in the morning
link |
and be like, I have a brilliant idea, right?
link |
It's usually a, okay, let me listen.
link |
I have a brilliant idea
link |
but let me get a little bit of feedback on this, like, okay.
link |
And then it's a, yeah, that was an awesome idea
link |
or it's like, yeah, let me go back.
link |
We already talked to a bunch of them,
link |
but are there some possible solutions
link |
to the biases present in our algorithms
link |
beyond what we just talked about?
link |
So I think there's two paths.
link |
One is to figure out how to systematically
link |
do the feedback and correction.
link |
So right now it's ad hoc, right?
link |
It's a researcher identify some outcomes
link |
that are not, don't seem to be fair, right?
link |
They publish it, they write about it
link |
and the, either the developer
link |
or the companies that have adopted the algorithms
link |
may try to fix it, right?
link |
And so it's really ad hoc and it's not systematic.
link |
There's, it's just, it's kind of like, I'm a researcher,
link |
that seems like an interesting problem,
link |
which means that there's a whole lot out there
link |
that's not being looked at, right?
link |
Because it's kind of researcher driven.
link |
And I don't necessarily have a solution,
link |
but that process, I think could be done a little bit better.
link |
One way is I'm going to poke a little bit
link |
at some of the corporations, right?
link |
Like maybe the corporations, when they think about a product,
link |
they should, instead of,
link |
in addition to hiring these, bug, they give these...
link |
Oh yeah, yeah, yeah.
link |
Like awards when you find a bug.
link |
Yeah, security bug, you know, let's put it like,
link |
we will give the, whatever the award is
link |
that we give for the people who find these security holes,
link |
find an ethics hole, right?
link |
Like find an unfairness hole
link |
and we will pay you X for each one you find.
link |
I mean, why can't they do that?
link |
One is a win win, they show that they're concerned about it,
link |
that this is important,
link |
and they don't have to necessarily dedicate
link |
their own internal resources.
link |
And it also means that everyone who has their own bias lens,
link |
like I'm interested in age,
link |
and so I'll find the ones based on age,
link |
and I'm interested in gender, right?
link |
Which means that you get all of these different perspectives.
link |
But you think of it in a data driven way.
link |
So like, sort of, if we look at a company like Twitter,
link |
it's under a lot of fire for discriminating
link |
against certain political beliefs.
link |
And sort of, there's a lot of people,
link |
this is the sad thing,
link |
because I know how hard the problem is,
link |
and I know the Twitter folks are working really hard at it,
link |
even Facebook, that everyone seems to hate
link |
are working really hard at this.
link |
You know, the kind of evidence that people bring
link |
is basically anecdotal evidence.
link |
Well, me or my friend, all we said is X,
link |
and for that we got banned.
link |
And that's kind of a discussion of saying,
link |
well, look, that's usually, first of all,
link |
the whole thing is taken out of context.
link |
So they present sort of anecdotal evidence.
link |
And how are you supposed to, as a company,
link |
in a healthy way have a discourse
link |
about what is and isn't ethical?
link |
What, how do we make algorithms ethical
link |
when people are just blowing everything?
link |
Like, they're outraged about a particular
link |
anecdotal piece of evidence that's very difficult
link |
to sort of contextualize in a big data driven way.
link |
Do you have a hope for companies like Twitter and Facebook?
link |
Yeah, so I think there's a couple of things going on, right?
link |
First off, the, remember this whole aspect
link |
of we are becoming reliant on technology,
link |
we're also becoming reliant on a lot of these,
link |
the apps and the resources that are provided, right?
link |
So some of it is kind of anger, like, I need you, right?
link |
And you're not working for me, right?
link |
Yeah, you're not working for me, they're right.
link |
But I think, and so some of it,
link |
and I wish that there was a little bit
link |
of change of rethinking.
link |
So some of it is like, oh, we'll fix it in house.
link |
No, that's like, okay, I'm a fox
link |
and I'm going to watch these hens
link |
because I think it's a problem that foxes eat hens.
link |
Like use, like be good citizens and say, look,
link |
we have a problem and we are willing to open ourselves up
link |
for others to come in and look at it
link |
and not try to fix it in house.
link |
Because if you fix it in house, there's conflict of interest.
link |
If I find something, I'm probably going to want to fix it
link |
and hopefully the media won't pick it up, right?
link |
And that then caused this distrust
link |
because someone inside is going to be mad at you
link |
and go out and talk about how,
link |
yeah, they can the resume survey
link |
because it rightly be best people.
link |
Like just say, look, we had this issue.
link |
Community, help us fix it.
link |
And we will give you like, you know,
link |
the bug finder fee if you do.
link |
So do you ever hope that the community,
link |
us as a human civilization on the whole is good
link |
and can be trusted to guide the future
link |
of our civilization into positive direction?
link |
So I'm an optimist, right?
link |
And, you know, there were some dark times in history always.
link |
I think now we're in one of those dark times.
link |
And it's not just US, right?
link |
So if it was just US, I'd be like, yeah, it's a US thing.
link |
But we're seeing it like worldwide this polarization.
link |
And so I worry about that.
link |
But I do fundamentally believe that at the end of the day,
link |
people are good, right?
link |
And why do I say that?
link |
Because anytime there's a scenario
link |
where people are in danger, and I will use,
link |
so Atlanta, we had a snowmageddon
link |
and people can laugh about that.
link |
People at the time, so the city closed for, you know,
link |
little snow, but it was ice and the city closed down.
link |
But you had people opening up their homes and saying,
link |
hey, you have nowhere to go.
link |
Come to my house, right?
link |
Hotels were just saying like, sleep on the floor.
link |
Like places like, you know, the grocery stores were like,
link |
There was no like, oh, how much are you gonna pay me?
link |
It was like this, such a community.
link |
And like people who didn't know each other,
link |
strangers were just like, can I give you a ride home?
link |
And that was a point I was like, you know what?
link |
That reveals that the deeper thing
link |
is there's a compassion or love that we all have within us.
link |
It's just that when all of that is taken care of
link |
and get bored, we love drama.
link |
And that's, I think almost like the division is a sign
link |
of the time is being good, is that it's just entertaining
link |
on some unpleasant mammalian level to watch,
link |
to disagree with others.
link |
And Twitter and Facebook are actually taking advantage
link |
of that in a sense because it brings you back
link |
to the platform and their advertises are driven
link |
so they make a lot of money.
link |
So you go back and you flick.
link |
Love doesn't sell quite as well in terms of advertisement.
link |
So you've started your career
link |
at NASA Jet Propulsion Laboratory.
link |
But before I ask a few questions there,
link |
have you happened to have ever seen Space Odyssey,
link |
2001 Space Odyssey?
link |
Okay, do you think HAL 9000?
link |
So we're talking about ethics.
link |
Do you think HAL did the right thing
link |
by taking the priority of the mission
link |
over the lives of the astronauts?
link |
Do you think HAL is good or evil?
link |
HAL was misguided.
link |
You're one of the people that would be in charge
link |
of an algorithm like HAL.
link |
So how would you do better?
link |
If you think about what happened was
link |
there was no fail safe, right?
link |
So we, perfection, right?
link |
Like what is that?
link |
I'm gonna make something that I think is perfect.
link |
But if my assumptions are wrong,
link |
it'll be perfect based on the wrong assumptions, right?
link |
That's something that you don't know
link |
until you deploy and then you're like, oh yeah, messed up.
link |
But what that means is that when we design software
link |
such as in Space Odyssey,
link |
when we put things out that there has to be a fail safe.
link |
There has to be the ability that once it's out there,
link |
we can grade it as an F and it fails
link |
and it doesn't continue, right?
link |
There's some way that it can be brought in
link |
and removed and that's aspect.
link |
Because that's what happened with HAL.
link |
It was like assumptions were wrong.
link |
It was perfectly correct based on those assumptions
link |
and there was no way to change it,
link |
change the assumptions at all.
link |
And the change, the fallback would be to humans.
link |
You ultimately think like humans should be,
link |
it's not turtles or AI all the way down.
link |
It's at some point, there's a human
link |
that actually makes this change.
link |
I still think that, and again,
link |
because I do human robot interaction,
link |
I still think the human needs to be part
link |
of the equation at some point.
link |
So what, just looking back,
link |
what are some fascinating things in robotic space
link |
that NASA was working at the time?
link |
Or just in general, what have you gotten to play with
link |
and what are your memories from working at NASA?
link |
Yeah, so one of my first memories was,
link |
they were working on a surgical robot system
link |
that could do eye surgery, right?
link |
And this was back in, oh my gosh,
link |
it must have been, oh, maybe 92, 93, 94.
link |
So it's like almost like a remote operation.
link |
Yeah, it was remote operation.
link |
And in fact, you can even find some old tech reports on it.
link |
So think of it, like now we have DaVinci, right?
link |
Like think of it, but these were like the late 90s, right?
link |
And I remember going into the lab one day
link |
and I was like, what's that, right?
link |
And of course it wasn't pretty, right?
link |
Cause the technology, but it was like functional.
link |
And you had this individual that could use
link |
the version of haptics to actually do the surgery.
link |
And they had this mockup of a human face and like the eyeballs
link |
and you can see this little drill.
link |
And I was like, oh, that is so cool.
link |
That one I vividly remember
link |
because it was so outside of my like possible thoughts
link |
of what could be done.
link |
It's the kind of precision.
link |
And I mean, what's the most amazing of a thing like that?
link |
I think it was the precision.
link |
It was the kind of first time
link |
that I had physically seen this robot machine,
link |
human interface, right?
link |
Versus, cause manufacturing had been,
link |
you saw those kind of big robots, right?
link |
But this was like, oh, this is in a person.
link |
There's a person and a robot like in the same space.
link |
The meeting them in person.
link |
Like for me, it was a magical moment that I can't,
link |
as life transforming that I recently met
link |
Spotmini from Boston Dynamics.
link |
I don't know why, but on the human robot interaction,
link |
for some reason I realized how easy it is
link |
to anthropomorphize.
link |
And it was, I don't know, it was almost like falling in love
link |
with this feeling of meeting.
link |
And I've obviously seen these robots a lot in video and so on,
link |
but meeting in person, just having that one on one time
link |
So have you had a robot like that in your life
link |
that made you maybe fall in love with robotics?
link |
Sort of like meeting in person?
link |
I mean, I loved robotics.
link |
From the beginning.
link |
Yeah, so that was a 12 year old,
link |
like I'm gonna be a roboticist.
link |
Actually was, I called it cybernetics,
link |
but so my motivation was Bionic Woman.
link |
I don't know if you know that.
link |
And so, I mean, that was like a seminal moment,
link |
but I didn't meet like that was TV, right?
link |
Like it wasn't like I was in the same space and I met,
link |
I was like, oh my gosh, you're like real.
link |
Just looking at Bionic Woman, which by the way,
link |
because I read that about you, I watched a bit of it
link |
and it's just so, no offense, terrible.
link |
It's cheesy, look at it now.
link |
I've seen a couple of reruns lately.
link |
But of course at the time, it's probably
link |
a captured imagination.
link |
But this is out of fix.
link |
Especially when you're younger, just capture you.
link |
But which aspect, did you think of it,
link |
you mentioned cybernetics, did you think of it as robotics
link |
or did you think of it as almost constructing
link |
artificial beings?
link |
Like is it the intelligent part
link |
that captured your fascination
link |
or was it the whole thing?
link |
Like even just the limbs and just the.
link |
So for me, it would have, in another world,
link |
I probably would have been more of a biomedical engineer
link |
because what fascinated me was the parts,
link |
like the bionic parts, the limbs, those aspects of it.
link |
Are you especially drawn to humanoid
link |
or human like robots?
link |
I would say human like, not humanoid, right?
link |
And when I say human like, I think it's this aspect
link |
of that interaction, whether it's social
link |
and it's like a dog, right?
link |
Like that's human like, because it understand us,
link |
it interacts with us at that very social level
link |
to, you know, humanoid is a part of that,
link |
but only if they interact with us as if we are human.
link |
But just to linger on NASA for a little bit,
link |
what do you think maybe if you have other memories,
link |
but also what do you think is the future
link |
of robots in space?
link |
We mentioned how, but there's incredible robots
link |
that NASA is working on in general,
link |
thinking about in our, as we venture out,
link |
human civilization ventures out into space.
link |
What do you think the future of robots is there?
link |
Yeah, so I mean, there's the near term.
link |
For example, they just announced the rover
link |
that's going to the moon, which, you know,
link |
that's kind of exciting, but that's like near term.
link |
You know, my favorite, favorite, favorite series
link |
is Star Trek, right?
link |
You know, I really hope and even Star Trek,
link |
like if I calculate the years, I wouldn't be alive,
link |
but I would really, really love to be in that world.
link |
Like even if it's just at the beginning,
link |
like, you know, like voyage, like adventure one.
link |
So basically living in space.
link |
With what robots, what are robots?
link |
The data would have to be, even though that wasn't,
link |
you know, that was like later, but.
link |
So data is a robot that has human like qualities.
link |
Right, without the emotion ship, yeah.
link |
You don't like emotion in your robots.
link |
Well, so data with the emotion ship was kind of a mess, right?
link |
It took a while for that, for him to adapt.
link |
But, and so why was that an issue?
link |
The issue is, is that emotions make us irrational agents.
link |
That's the problem.
link |
And yet he could think through things,
link |
even if it was based on an emotional scenario, right?
link |
Based on pros and cons.
link |
But as soon as you made him emotional,
link |
one of the metrics he used for evaluation
link |
was his own emotions.
link |
Not people around him, right?
link |
We do that as children, right?
link |
So we're very egocentric when we're young.
link |
We are very egocentric.
link |
And so, isn't that just an early version
link |
of the emotion ship then?
link |
I haven't watched much Star Trek.
link |
Except I have also met adults, right?
link |
And so that is a developmental process.
link |
And I'm sure there's a bunch of psychologists
link |
that can go through, like you can have a 60 year old adult
link |
who has the emotional maturity of a 10 year old, right?
link |
And so there's various phases
link |
that people should go through in order to evolve.
link |
And sometimes you don't.
link |
So how much psychology do you think
link |
a topic that's rarely mentioned in robotics,
link |
but how much does psychology come to play
link |
when you're talking about HRI, human robot interaction?
link |
When you have to have robots
link |
that actually interact with you?
link |
So we, like my group, as well as I read a lot
link |
in the cognitive science literature
link |
as well as the psychology literature.
link |
Because they understand a lot about human human relations
link |
and developmental milestones and things like that.
link |
And so we tend to look to see what's been done out there.
link |
Sometimes what we'll do is we'll try to match that to see
link |
is that human human relationship the same as human robot?
link |
Sometimes it is and sometimes it's different.
link |
And then when it's different, we have to,
link |
we try to figure out, okay,
link |
why is it different in this scenario?
link |
But it's the same in the other scenario, right?
link |
And so we try to do that quite a bit.
link |
Would you say that's,
link |
if we're looking at the future of human robot interaction,
link |
would you say the psychology piece is the hardest?
link |
Like if, I mean, it's a funny notion for you as,
link |
I don't know if you consider, yeah.
link |
I mean, one way to ask it,
link |
do you consider yourself a roboticist or a psychologist?
link |
Oh, I consider myself a roboticist
link |
that plays the act of a psychologist.
link |
But if you were to look at yourself sort of,
link |
you know, 20, 30 years from now,
link |
do you see yourself more and more wearing the psychology hat?
link |
Sort of another way to put it is,
link |
are the hard problems in human robot interactions,
link |
fundamentally psychology,
link |
or is it still robotics,
link |
the perception manipulation,
link |
planning all that kind of stuff?
link |
It's actually neither.
link |
The hardest part is the adaptation and the interaction.
link |
So it's the interface, it's the learning.
link |
And so if I think of,
link |
like I've become much more of a roboticist slash AI person
link |
than when I, like originally, again,
link |
I was about the bionics.
link |
I was electrical engineer, I was control theory, right?
link |
Like, and then I started realizing
link |
that my algorithms needed like human data, right?
link |
And so then I was like, okay, what is this human thing?
link |
Right, how do I incorporate human data?
link |
And then I realized that human perception had,
link |
like there was a lot in terms of how we perceive the world.
link |
And so trying to figure out,
link |
how do I model human perception from my,
link |
and so I became a HRI person,
link |
human robot interaction person,
link |
from being a control theory and realizing
link |
that humans actually offered quite a bit.
link |
And then when you do that,
link |
you become more of an artificial intelligence AI.
link |
And so I see myself evolving more in this AI world
link |
under the lens of robotics
link |
having hardware interacting with people.
link |
So you're a world class expert researcher in robotics
link |
and yet others, there's a few, it's a small,
link |
but fierce community of people,
link |
but most of them don't take the journey
link |
into the age of HRI, into the human.
link |
So why did you brave into the interaction with humans?
link |
It seems like a really hard problem.
link |
It's a hard problem and it's very risky as an academic.
link |
And I knew that when I started down that journey
link |
that it was very risky as an academic in this world
link |
that was nuanced, it was just developing.
link |
We didn't even have a conference, right?
link |
Because it was the interesting problems.
link |
That was what drove me.
link |
It was the fact that I looked at what interests me
link |
in terms of the application space and the problems.
link |
And that pushed me into trying to figure out
link |
what people were and what humans were
link |
and how to adapt to them.
link |
If those problems weren't so interesting,
link |
I'd probably still be sending rovers to glaciers, right?
link |
But the problems were interesting.
link |
And the other thing was that they were hard, right?
link |
So it's, I like having to go into a room
link |
and being like, I don't know what to do.
link |
And then going back and saying, okay,
link |
I'm gonna figure this out.
link |
I do not, I'm not driven when I go in like,
link |
oh, there are no surprises.
link |
Like I don't find that satisfying.
link |
If that was the case,
link |
I'd go someplace and make a lot more money, right?
link |
I think I stay in academic and choose to do this
link |
because I can go into a room and I'm like, that's hard.
link |
Yeah, I think just from my perspective,
link |
maybe you can correct me on it,
link |
but if I just look at the field of AI broadly,
link |
it seems that human robot interaction has the most,
link |
one of the most number of open problems.
link |
Like people, especially relative
link |
to how many people are willing to acknowledge that there are.
link |
This, because most people are just afraid of the humans
link |
so they don't even acknowledge how many open problems there
link |
but it's in terms of difficult problems
link |
to solve exciting spaces,
link |
it seems to be incredible for that.
link |
It is, and it's exciting.
link |
You've mentioned trust before.
link |
What role does trust from interacting with autopilot
link |
to in the medical context,
link |
what role does trust play in the human robot interaction?
link |
So some of the things I study in this domain
link |
is not just trust, but it really is over trust.
link |
How do you think about over trust?
link |
Like first of all, what is trust and what is over trust?
link |
Basically, the way I look at it is
link |
trust is not what you click on a survey,
link |
trust is about your behavior.
link |
So if you interact with the technology
link |
based on the decision or the actions of the technology
link |
as if you trust that decision, then you're trusting, right?
link |
And even in my group, we've done surveys
link |
that on the thing do you trust robots?
link |
Would you follow this robot in a burning building?
link |
Of course not, right?
link |
And then you look at their actions
link |
and you're like, clearly your behavior
link |
does not match what you think, right?
link |
Or what you think you would like to think, right?
link |
And so I'm really concerned about the behavior
link |
because that's really at the end of the day
link |
when you're in the world,
link |
that's what will impact others around you.
link |
It's not whether before you went onto the street,
link |
you clicked on like, I don't trust self driving cars.
link |
Yeah, that from an outsider perspective,
link |
it's always frustrating to me.
link |
Well, I read a lot.
link |
So I'm insider in a certain philosophical sense.
link |
It's frustrating to me how often trust is used in surveys
link |
and how people say make claims
link |
out of any kind of finding they make
link |
while somebody clicking on answer.
link |
You just trust is, yeah, behavior just,
link |
you said it beautifully.
link |
I mean, action, your own behavior is what trust is.
link |
I mean, that everything else is not even close.
link |
It's almost like absurd comedic poetry
link |
that you weave around your actual behavior.
link |
So some people can say they trust,
link |
you know, I trust my wife, husband or not, whatever,
link |
but the actions is what speaks volumes.
link |
If you bug their car, you probably don't trust them.
link |
I trust them, I'm just making sure.
link |
No, no, that's, yeah.
link |
Like even if you think about cars,
link |
I think it's a beautiful case.
link |
I came here at some point,
link |
I'm sure on either Uber or Lyft, right?
link |
I remember when it first came out, right?
link |
I bet if they had had a survey,
link |
would you get in the car with a stranger and pay them?
link |
How many people do you think would have said, like, really?
link |
You know, wait, even worse,
link |
would you get in the car with a stranger
link |
at 1 a.m. in the morning
link |
to have them drop you home as a single female?
link |
Like how many people would say, that's stupid?
link |
And now look at where we are.
link |
I mean, people put kids, right?
link |
Like, oh yeah, my child has to go to school
link |
and I, yeah, I'm gonna put my kid in this car
link |
I mean, it's just fascinating how,
link |
like, what we think we think is not necessarily
link |
matching our behavior.
link |
Yeah, and certainly with robots, with autonomous vehicles,
link |
and all the kinds of robots you work with,
link |
that's, it's, yeah, it's the way you answer it,
link |
especially if you've never interacted
link |
with that robot before.
link |
If you haven't had the experience,
link |
you being able to respond correctly
link |
on a survey is impossible.
link |
But what role does trust play in the interaction,
link |
Like, is it good to, is it good to trust a robot?
link |
What does over trust mean?
link |
Or is it, is it good to kind of how you feel
link |
about autopilot currently, which is like,
link |
from a robotics perspective, is like,
link |
still very cautious?
link |
Yeah, so this is still an open area of research.
link |
But basically what I would like in a perfect world
link |
is that people trust the technology
link |
when it's working 100%,
link |
and people will be hypersensitive
link |
and identify when it's not.
link |
But of course we're not there.
link |
That's the ideal world.
link |
And, but we find is that people swing, right?
link |
They tend to swing, which means that if my first,
link |
and like, we have some papers,
link |
like first impressions is everything, right?
link |
If my first instance with technology with robotics
link |
is positive, it mitigates any risk,
link |
it correlates with like best outcomes.
link |
It means that I'm more likely to either not see it
link |
when it makes some mistakes or faults,
link |
or I'm more likely to forgive it.
link |
And so this is a problem
link |
because technology is not 100% accurate, right?
link |
It's not 100% accurate, although it may be perfect.
link |
How do you get that first moment right, do you think?
link |
There's also an education about the capabilities
link |
and limitations of the system.
link |
Do you have a sense of how you educate people correctly
link |
in that first interaction?
link |
Again, this is an open ended problem.
link |
So one of the study that actually has given me some hope
link |
that I was trying to figure out how to put in robotics.
link |
So there was a research study
link |
that it showed for medical AI systems,
link |
giving information to radiologists about, you know,
link |
here you need to look at these areas on the X ray.
link |
What they found was that when the system provided one choice,
link |
there was this aspect of either no trust or over trust, right?
link |
Like I'm not, I don't believe it at all,
link |
or a yes, yes, yes, yes, and they would miss things, right?
link |
Instead, when the system gave them multiple choices,
link |
like here are the three, even if it knew, like, you know,
link |
it had estimated that the top area you need to look at
link |
was some place on the X ray.
link |
If it gave like one plus others,
link |
the trust was maintained
link |
and the accuracy of the entire population increased.
link |
Right? So basically it was a, you're still trusting the system,
link |
but you're also putting in a little bit of like
link |
your human expertise,
link |
like your human decision processing into the equation.
link |
So it helps to mitigate that over trust risk.
link |
Yeah. So there's a fascinating balance at the strike.
link |
Haven't figured out again.
link |
It's exciting open area research. Exactly.
link |
So what are some exciting applications
link |
of human robot interaction?
link |
You started a company, maybe you can talk about
link |
the exciting efforts there,
link |
but in general also what other space
link |
can robots interact with humans and help?
link |
Yeah. So besides healthcare, because, you know,
link |
that's my bias lens.
link |
My other bias lens is education.
link |
I think that, well, one, we definitely, we,
link |
in the US, you know, we're doing okay with teachers,
link |
but there's a lot of school districts
link |
that don't have enough teachers.
link |
If you think about the teacher student ratio
link |
for at least public education in some districts,
link |
It's like, how can you have learning in that classroom?
link |
Because you just don't have the human capital.
link |
And so if you think about robotics,
link |
bringing that in to classrooms,
link |
as well as the after school space,
link |
where they offset some of this lack of resources
link |
in certain communities, I think that's a good place.
link |
And then turning on the other end is using these systems then
link |
for workforce retraining and dealing with some of the things
link |
that are going to come out later on of job loss,
link |
like thinking about robots and NAI systems
link |
for retraining and workforce development.
link |
I think that's exciting areas that can be pushed even more,
link |
and it would have a huge, huge impact.
link |
What would you say are some of the open problems in education?
link |
Sort of, it's exciting.
link |
So young kids and the older folks or just folks of all ages
link |
who need to be retrained,
link |
who need to sort of open themselves up
link |
to a whole other area of work.
link |
What are the problems to be solved there?
link |
How do you think robots can help?
link |
We have the engagement aspect, right?
link |
So we can figure out the engagement.
link |
What do you mean by engagement?
link |
So identifying whether a person is focused
link |
is like that we can figure out.
link |
What we can figure out,
link |
and there's some positive results in this,
link |
is that personalized adaptation based on any concepts, right?
link |
So imagine I think about I have an agent
link |
and I'm working with a kid learning, I don't know, Algebra 2.
link |
Can that same agent then switch and teach
link |
some type of new coding skill to a displaced mechanic?
link |
What does that actually look like?
link |
Hardware might be the same,
link |
content is different,
link |
two different target demographics of engagement.
link |
How do you do that?
link |
How important do you think personalization
link |
is in human robot interaction?
link |
Not just a mechanic or student,
link |
but literally to the individual human being?
link |
I think personalization is really important,
link |
but a caveat is that I think we'd be okay
link |
if we can personalize to the group, right?
link |
And so if I can label you as along some certain dimensions,
link |
then even though it may not be you specifically,
link |
I can put you in this group.
link |
So the sample size, this is how they best learn,
link |
this is how they best engage.
link |
Even at that level, it's really important.
link |
And it's because, I mean, it's one of the reasons
link |
why educating in large classrooms is so hard, right?
link |
You teach to the median,
link |
but there's these individuals that are struggling,
link |
and then you have highly intelligent individuals,
link |
and those are the ones that are usually kind of left out.
link |
So highly intelligent individuals may be disruptive,
link |
and those who are struggling might be disruptive
link |
because they're both bored.
link |
And if you narrow the definition of the group
link |
or in the size of the group enough,
link |
you'll be able to address their individual needs,
link |
but really the most important group needs, right?
link |
And that's kind of what a lot of successful
link |
recommender systems do, Spotify and so on.
link |
It's sad to believe, but as a music listener,
link |
probably in some sort of large group,
link |
it's very sadly predictable.
link |
You have been labeled.
link |
Yeah, I've been labeled and successfully so,
link |
because they're able to recommend stuff.
link |
Yeah, but applying that to education, right?
link |
There's no reason why it can't be done.
link |
Do you have a hope for our education system?
link |
I have more hope for workforce development,
link |
and that's because I'm seeing investments.
link |
Even if you look at VC investments in education,
link |
the majority of it has lately been going to workforce retraining,
link |
right? And so I think that government investments
link |
is increasing. There's like a claim.
link |
And some of this based on fear, right?
link |
Like AI is going to come and take over all these jobs.
link |
What are we going to do with all these nonpaying taxes
link |
that aren't coming to us by our citizens?
link |
And so I think I'm more hopeful for that.
link |
Not so hopeful for early education,
link |
because it's this, it's still a who's going to pay for it,
link |
and you won't see the results for like 16 to 18 years.
link |
It's hard for people to wrap their heads around that.
link |
But on the retraining part, what are your thoughts?
link |
There's a candidate, Andrew Yang, running for president,
link |
saying that sort of AI automation, robots,
link |
universal basic income.
link |
Universal basic income in order to support us
link |
as we kind of automation takes people's jobs
link |
and allows you to explore and find other means.
link |
Like, do you have a concern of society transforming effects
link |
of automation and robots and so on?
link |
I do. I do know that AI robotics will displace workers.
link |
Like, we do know that.
link |
But there'll be other workers that will be defined new jobs.
link |
What I worry about is, that's not what I worry about.
link |
Like, will all the jobs go away?
link |
What I worry about is a type of jobs that will come out, right?
link |
Like, people who graduate from Georgia Tech will be okay, right?
link |
We give them the skills, they will adapt,
link |
even if their current job goes away.
link |
I do worry about those that don't have that quality of an education, right?
link |
Will they have the ability, the background to adapt to those new jobs?
link |
That, I don't know. That I worry about.
link |
Which will create even more polarization in our society, internationally,
link |
and everywhere. I worry about that.
link |
I also worry about not having equal access to all these wonderful things
link |
that AI can do and robotics can do.
link |
I worry about that.
link |
People like me from Georgia Tech, from say MIT, will be okay, right?
link |
But that's such a small part of the population
link |
that we need to think much more globally of having access to the beautiful things,
link |
whether it's AI in healthcare, AI in education, AI in politics, right?
link |
I worry about that.
link |
And that's part of the thing that you were talking about
link |
is people that build the technology had to be thinking about, ethics had to be thinking about access
link |
and all those things, and not just a small subset.
link |
Let me ask some philosophical, slightly romantic questions.
link |
People that listen to this will be like, here he goes again.
link |
Okay. Do you think one day we'll build an AI system
link |
that a person can fall in love with and it would love them back?
link |
Like in a movie, Her, for example.
link |
Although she kind of didn't fall in love with him.
link |
She fell in love with like a million other people, something like that.
link |
You're the jealous type, I see.
link |
We humans are the jealous type.
link |
So I do believe that we can design systems where people would fall in love
link |
with their robot, with their AI partner.
link |
That I do believe.
link |
Because it's actually, and I don't like to use the word manipulate,
link |
but as we see, there are certain individuals that can be manipulated
link |
if you understand the cognitive science about it, right?
link |
So I mean, if you could think of all close relationship and love in general
link |
as a kind of mutual manipulation, that dance, the human dance.
link |
I mean, manipulation is a negative connotation.
link |
And that's why I don't like to use that word particularly.
link |
I guess another way to phrase it is you're getting at it,
link |
it could be algorithmatized or something.
link |
The relationship building part can be.
link |
I mean, just think about it.
link |
We have, and I don't use dating sites, but from what I heard,
link |
there are some individuals that have been dating that have never saw each other, right?
link |
In fact, there's a show, I think, that tries to weed out fake people.
link |
There's a show that comes out, right?
link |
Because people start faking.
link |
What's the difference of that person on the other end being an AI agent, right?
link |
And having a communication and you building a relationship remotely,
link |
there's no reason why that can't happen.
link |
In terms of human robot interaction,
link |
what role, you've kind of mentioned with data, emotion being,
link |
can be problematic if not implemented well, I suppose.
link |
What role does emotion and some other human like things,
link |
the imperfect things come into play here for good human robot interaction
link |
and something like love?
link |
So in this case, and you had asked, can an AI agent love a human back?
link |
I think they can emulate love back, right?
link |
And so what does that actually mean?
link |
It just means that if you think about their programming,
link |
they might put the other person's needs in front of theirs in certain situations, right?
link |
Think about it as return on investment.
link |
Like, was my return on investment as part of that equation,
link |
that person's happiness has some type of algorithm waiting to it.
link |
And the reason why is because I care about them, right?
link |
That's the only reason, right?
link |
But if I care about them and I show that, then my final objective function
link |
is length of time of the engagement, right?
link |
So you can think of how to do this actually quite easily.
link |
But that's not love?
link |
Well, so that's the thing.
link |
I think it emulates love because we don't have a classical definition of love.
link |
And we don't have the ability to look into each other's minds to see the algorithm.
link |
And I guess what I'm getting at is, is it possible that,
link |
especially if that's learned,
link |
especially if there's some mystery and black box nature to the system, how is that, you know,
link |
How is it any different?
link |
How is it any different?
link |
And in terms of sort of, if the system says, I'm conscious, I'm afraid of death,
link |
and it does indicate that it loves you.
link |
Another way to sort of phrase it, I'd be curious to see what you think.
link |
Do you think there'll be a time when robots should have rights?
link |
You've kind of phrased the robot in a very roboticist way.
link |
It's just a really good way.
link |
But saying, okay, well, there's an objective function,
link |
and I could see how you can create a compelling human robot interaction experience
link |
that makes you believe that the robot cares for your needs
link |
and even something like loves you.
link |
But what if the robot says, please don't turn me off?
link |
What if the robot starts making you feel like there's an entity of being a soul there?
link |
Do you think there'll be a future?
link |
Hopefully you won't laugh too much at this,
link |
but where they do ask for rights.
link |
So I can see a future if we don't address it in the near term,
link |
where these agents, as they adapt and learn, could say, hey,
link |
this should be something that's fundamental.
link |
I hopefully think that we would address it before it gets to that point.
link |
You think that's a bad future.
link |
Is that a negative thing where they ask or being discriminated against?
link |
I guess it depends on what role have they attained at that point.
link |
And so if I think about now...
link |
Careful what you say, because the robot's 50 years when I'll be listening to this,
link |
and you'll be on TV saying, this is what roboticists used to believe.
link |
And so this is my, and as I said, I have a biased lens,
link |
and my robot friends will understand that.
link |
But so if you think about it,
link |
and I actually put this in kind of the...
link |
As a roboticist, you don't necessarily think of robots as human with human rights,
link |
but you could think of them either in the category of property,
link |
or you could think of them in the category of animals, right?
link |
And so both of those have different types of rights.
link |
So animals have their own rights as a living being,
link |
but they can't vote, they can't write, they can be euthanized.
link |
But as humans, if we abuse them, we go to jail, right?
link |
So they do have some rights that protect them,
link |
but don't give them the rights of citizenship.
link |
And then if you think about property,
link |
property, the rights are associated with the person, right?
link |
So if someone vandalizes your property,
link |
or steals your property, like there are some rights,
link |
but it's associated with the person who owns that.
link |
If you think about it, back in the day,
link |
and remember we talked about how society has changed,
link |
women were property, right?
link |
They were not thought of as having rights,
link |
they were thought of as property of...
link |
Assaulting a woman meant assaulting the property of somebody else.
link |
And so what I envision is that we will establish
link |
some type of norm at some point, but that it might evolve, right?
link |
If you look at women's rights now,
link |
there are still some countries that don't have,
link |
and the rest of the world is like, why?
link |
That makes no sense, right?
link |
And so I do see a world where we do establish some type of grounding.
link |
It might be based on property rights.
link |
It might be based on animal rights.
link |
And if it evolves that way,
link |
I think we will have this conversation at that time,
link |
because that's the way our society traditionally has evolved.
link |
Just out of curiosity,
link |
Anki, Gibo, Mayfield Robotics,
link |
with the Robot Curie, SciFile Works,
link |
we think Robotics were all these amazing robotics companies
link |
created by incredible roboticists,
link |
and they've all went out of business recently.
link |
Why do you think they didn't last long?
link |
Why is it so hard to run a robotics company,
link |
especially one like these,
link |
which are fundamentally HRI,
link |
human robot interaction robots?
link |
Yeah, each one has a story.
link |
Only one of them I don't understand.
link |
And that was Anki.
link |
That's actually the only one I don't understand.
link |
I don't understand it either.
link |
I mean, from the outside,
link |
I've looked at their sheets,
link |
I've looked at the data that's...
link |
Oh, you mean like business wise?
link |
Yeah, and I look at that data,
link |
and I'm like, they seem to have product market fit.
link |
So that's the only one I don't understand.
link |
The rest of it was product market fit.
link |
What's product market fit?
link |
How do you think about it?
link |
Yeah, so although we think Robotics was getting there,
link |
but I think it's just the timing,
link |
their clock just timed out.
link |
I think if they had been given a couple more years,
link |
they would have been okay.
link |
But the other ones were still fairly early
link |
by the time they got into the market.
link |
And so product market fit is,
link |
I have a product that I want to sell at a certain price.
link |
Are there enough people out there, the market,
link |
that are willing to buy the product at that market price
link |
for me to be a functional, viable, profit bearing company?
link |
So product market fit.
link |
If it costs you $1,000 and everyone wants it
link |
and only is willing to pay a dollar,
link |
you have no product market fit.
link |
Even if you could sell it for, you know,
link |
it's enough for a dollar because you can't...
link |
So how hard is it for robots?
link |
Maybe if you look at iRobot,
link |
the company that makes Roombas vacuum cleaners,
link |
can you comment on did they find the right product,
link |
a market product fit?
link |
Like are people willing to pay for robots?
link |
It's also another kind of question.
link |
So if you think about iRobot and their story, right?
link |
Like when they first, they had enough of a runway, right?
link |
When they first started, they weren't doing vacuum cleaners, right?
link |
They were contracts, primarily government contracts,
link |
Yeah, I mean, that's what they were.
link |
That's how they started, right?
link |
They still do a lot of incredible work there.
link |
But yeah, that was the initial thing
link |
that gave them enough funding to...
link |
The vacuum cleaner is what I've been told
link |
was not like their first rendezvous
link |
in terms of designing a product, right?
link |
And so they were able to survive
link |
until they got to the point that they found a product,
link |
price market, right?
link |
And even with, if you look at the Roomba,
link |
the price point now is different
link |
than when it was first released, right?
link |
It was an early adopter price,
link |
but they found enough people who were willing to fund it.
link |
And I mean, I forgot what their loss profile was
link |
for the first couple of years,
link |
but they became profitable in sufficient time
link |
that they didn't have to close their doors.
link |
So they found the right,
link |
there's still people willing to pay
link |
a large amount of money,
link |
sort of over $1,000 for a vacuum cleaner.
link |
Unfortunately for them,
link |
now that they've proved everything out,
link |
figured it all out, now there's competitors.
link |
Yeah, and so that's the next thing, right?
link |
The competition, and they have quite a number,
link |
even internationally,
link |
like there's some products out there,
link |
you can go to Europe and be like,
link |
oh, I didn't even know this one existed.
link |
So this is the thing though,
link |
like with any market,
link |
I would, this is not a bad time,
link |
although, you know, as a roboticist,
link |
it's kind of depressing,
link |
but I actually think about things like with,
link |
I would say that all of the companies
link |
that are now in the top five or six,
link |
they weren't the first to the stage, right?
link |
Like Google was not the first search engine,
link |
sorry, Alta Vista, right?
link |
Facebook was not the first, sorry, MySpace, right?
link |
Like, think about it,
link |
they were not the first players.
link |
Those first players, like,
link |
they're not in the top five, 10 of Fortune 500 companies, right?
link |
They proved, they started to prove out the market,
link |
they started to get people interested,
link |
they started the buzz,
link |
but they didn't make it to that next level.
link |
But the second batch, right?
link |
The second batch, I think, might make it to the next level.
link |
When do you think the Facebook of...
link |
The Facebook of robotics.
link |
Sorry, I take that phrase back
link |
because people deeply, for some reason,
link |
but it's, I think, exaggerated distrust Facebook
link |
because of the privacy concerns and so on.
link |
And with robotics,
link |
one of the things you have to make sure
link |
is all the things we talked about
link |
is to be transparent and have people deeply trust you
link |
to let a robot into their lives, into their home.
link |
But when do you think the second batch of robots...
link |
Is it five, 10 years, 20 years
link |
that we'll have robots in our homes
link |
and robots in our hearts?
link |
So if I think about...
link |
Because I try to follow the VC kind of space
link |
in terms of robotic investments.
link |
and I don't know if they're going to be successful,
link |
I don't know if this is the second batch,
link |
but there's only one batch that's focused on the first batch, right?
link |
And then there's all these self driving Xs, right?
link |
And so I don't know if they're a first batch of something
link |
or if, like, I don't know quite where they fit in,
link |
but there's a number of companies,
link |
the co robot, I would call them co robots,
link |
that are still getting VC investments.
link |
They, some of them have some of the flavor
link |
of, like, rethink robotics.
link |
Some of them have some of the flavor of, like, Curie.
link |
What's a co robot?
link |
So basically a robot and human working in the same space.
link |
So some of the companies are focused on manufacturing.
link |
So having a robot and human working together in a factory,
link |
some of these co robots are robots
link |
and humans working in the home, working in clinics.
link |
Like there's different versions of these companies
link |
in terms of their products,
link |
but they're all, so rethink robotics would be, like,
link |
one of the first, at least well known,
link |
companies focused on this space.
link |
So I don't know if this is a second batch
link |
or if this is still part of the first batch,
link |
that I don't know.
link |
And then you have all these other companies
link |
in this self driving, you know, space.
link |
And I don't know if that's a first batch
link |
or, again, a second batch.
link |
So there's a lot of mystery about this now.
link |
Of course, it's hard to say that this is the second batch
link |
until it proves out, right?
link |
We need a unicorn.
link |
Why do you think people are so afraid,
link |
at least in popular culture of legged robots
link |
like those worked in Boston Dynamics
link |
or just robotics in general,
link |
if you were to psychoanalyze that fear,
link |
what do you make of it?
link |
And should they be afraid?
link |
So should people be afraid?
link |
I don't think people should be afraid.
link |
But with a caveat.
link |
I don't think people should be afraid
link |
given that most of us in this world
link |
understand that we need to change something, right?
link |
Now, if things don't change, be very afraid.
link |
Which is the dimension of change that's needed?
link |
So changing, thinking about the ramifications,
link |
thinking about like the ethics,
link |
thinking about like the conversation is going on, right?
link |
It's no longer a...
link |
We're going to deploy it and forget that, you know,
link |
this is a car that can kill pedestrians
link |
that are walking across the street, right?
link |
We're not in that stage.
link |
We're putting these roads out.
link |
There are people out there.
link |
A car could be a weapon.
link |
People are now, solutions aren't there yet.
link |
But people are thinking about this
link |
as we need to be ethically responsible
link |
as we send these systems out,
link |
robotics, medical, self driving.
link |
Which is not as often talked about,
link |
but it's really where probably these robots
link |
will have a significant impact as well.
link |
Right, making sure that they can think rationally,
link |
even having the conversations,
link |
who should pull the trigger, right?
link |
But overall, you're saying if we start to think
link |
more and more as a community about these ethical issues,
link |
people should not be afraid.
link |
Yeah, I don't think people should be afraid.
link |
I think that the return on investment,
link |
the impact, positive impact will outweigh
link |
any of the potentially negative impacts.
link |
Do you have worries of existential threats
link |
of robots or AI that some people kind of
link |
talk about and romanticize about
link |
in the next decade, in the next few decades?
link |
Singularity would be an example.
link |
So my concept is that, so remember,
link |
robots, AI is designed by people.
link |
It has our values.
link |
And I always correlate this with
link |
a parent and a child, right?
link |
So think about it.
link |
As a parent, what do we want?
link |
We want our kids to have a better life than us.
link |
We want them to expand.
link |
We want them to experience the world.
link |
And then as we grow older, our kids think
link |
and know they're smarter and better
link |
and more intelligent and have better opportunities
link |
and they may even stop listening to us.
link |
They don't go out and then kill us, right?
link |
It's because it's instilled in them values.
link |
We instilled in them this whole aspect of community.
link |
And yes, even though you're maybe smarter
link |
and have more money and da, da, da,
link |
it's still about this love, caring relationship.
link |
And so that's what I believe.
link |
So even if we've created the Singularity
link |
and some archaic system back in 1980
link |
that suddenly evolves, the fact is it might say,
link |
I am smarter, I am sentient.
link |
These humans are really stupid.
link |
But I think it'll be like, yeah,
link |
but I just can't destroy them.
link |
Yeah, percent of mental value.
link |
Still just to come back
link |
for Thanksgiving dinner every once in a while.
link |
That's so beautifully put.
link |
You've also said that The Matrix may be
link |
one of your more favorite A.I. related movies.
link |
Can you elaborate why?
link |
Yeah, it is one of my favorite movies.
link |
And it's because it represents
link |
kind of all the things I think about.
link |
So there's a symbiotic relationship
link |
between robots and humans, right?
link |
That symbiotic relationship is that they don't destroy us.
link |
They enslave us, right?
link |
But think about it.
link |
Even though they enslaved us,
link |
they needed us to be happy, right?
link |
And in order to be happy,
link |
they had to create this crudy world
link |
that they then had to live in, right?
link |
That's the whole premise.
link |
But then there were humans that had a choice, right?
link |
Like you had a choice to stay in this horrific,
link |
horrific world where it was your fantasied life
link |
with all of the anomalies, perfection, but not accurate.
link |
Or you can choose to be on your own
link |
and have maybe no food for a couple of days,
link |
but you were totally autonomous.
link |
And so I think of that as...
link |
So it's not necessarily us being enslaved,
link |
but I think about us having the symbiotic relationship.
link |
Robots and A.I., even if they become sentient,
link |
they're still part of our society,
link |
and they will suffer just as much as we.
link |
And there will be some kind of equilibrium
link |
that we'll have to find, some symbiotic relationship.
link |
And then you have the ethicists, the robotics folks,
link |
that are like, no, this has got to stop.
link |
I will take the other pill in order to make a difference.
link |
So if you could hang out for a day with a robot,
link |
real or from science fiction, movies, books,
link |
safely and get to pick his or her, their brain,
link |
who would you pick?
link |
I got to say it's data.
link |
I was going to say Rosie,
link |
but I don't, I'm not really interested in her brain.
link |
I'm interested in data's brain.
link |
Data, pre or post emotion chip?
link |
But don't you think it'd be a more interesting conversation,
link |
post emotion chip?
link |
Yeah, it would be drama.
link |
And I, you know, I'm human.
link |
I deal with drama all the time.
link |
But the reason why I want to pick data's brain
link |
is because I could have a conversation with him
link |
and ask, for example, how can we fix this ethics problem?
link |
And he could go through like the rational thinking
link |
and through that, he could also help me think through it as well.
link |
And so that's, there's like these questions,
link |
fundamental questions I think I could ask him.
link |
That he would help me also learn from.
link |
And that fascinates me.
link |
I don't think there's a better place to end it.
link |
Ayanna, thank you so much for talking to us in honor.
link |
Thanks for listening to this conversation.
link |
And thank you to our presenting sponsor, Cash App.
link |
Use code Lex podcast.
link |
You'll get $10 and $10 will go to first a STEM education nonprofit
link |
that inspires hundreds of thousands of young minds
link |
to become future leaders and innovators.
link |
If you enjoy this podcast, subscribe on YouTube,
link |
give it five stars on Apple podcast,
link |
follow on Spotify, support on Patreon,
link |
or simply connect with me on Twitter.
link |
And now let me leave you with some words of wisdom
link |
from Arthur C. Clarke.
link |
Whether we are based on carbon or on silicon
link |
makes no fundamental difference.
link |
We should each be treated with appropriate respect.
link |
Thank you for listening and hope to see you next time.