Can Empathy Prevent A Robot Uprising?

August 9, 2018

The robot apocalypse is a popular theme in movies, but there is debate over whether or not it's a legitimate concern or a Y2K-esque exaggeration. We take a look at whether integrating emotional intelligence into artificial intelligence may be part of the answer to avoiding the robot uprising.

FULL TRANSCRIPT

MALE_1: Imagine you're a person working in an office

MALE_1: 100 years in the future and you

MALE_1: realize [NOISE] you've run out of paperclips.

MALE_1: Frustrated, you turned to your supersmart

MALE_1: Artificial Intelligence robot assistant and say,

MALE_1: make sure I never run out of paperclips again.

MALE_1: The robot doing as it's told,

MALE_1: [NOISE] goes about learning how to acquire

MALE_1: every single paper they can,

MALE_1: every single paperclip in existence.

MALE_1: Once it's collected all of those,

MALE_1: it teaches itself how to make paperclips,

MALE_1: [NOISE] gathering all the resources,

MALE_1: like iron it needs to do this.

MALE_1: [MUSIC] Eventually it's used up all the iron in

MALE_1: the entire world making

MALE_1: paperclips and starts to figure out what else it can use.

MALE_1: Humans have iron in their blood.

MALE_1: [NOISE] So, why not use them to make paperclips?

MALE_1: Eventually, the entire planet becomes

MALE_1: paperclips and then the entire universe.

MALE_1: Your trusty AI robot has made sure you

MALE_1: never ever run out of paperclips again.

MALE_1: [MUSIC] This is a bizarre scenario but it represents

MALE_1: a real concern for researchers

MALE_1: theorizing about the future of Artificial Intelligence.

MALE_1: If we created supersmart machines,

MALE_1: machines that are way smarter than us,

MALE_1: would they destroy us?

MALE_1: And if this could happen wouldn't it make

MALE_1: sense for us to teach our machines to care about us?

MALE_1: Can we even do that?

MALE_1: [MUSIC] We should probably

MALE_1: take a second to clarify a few things.

MALE_1: Artificial Intelligence is a big wide term

MALE_1: used to describe all types of software machines.

MALE_1: Basically, it means technology that performs

MALE_1: tasks that normally would

MALE_1: require a human brain to execute.

MALE_1: One misconception is that AI is synonymous with robots.

MALE_1: Robots can be a house for AI but it isn't AI itself.

MALE_1: Also, we use the term AI

MALE_1: a lot to talk about technology of the future.

MALE_1: We use AI right now.

MALE_1: We just don't call it that because we've

MALE_1: integrated it so much into our lives.

MALE_1: Computer scientists say we are on

MALE_1: the cusp of an Artificial Intelligence revolution.

MALE_1: Autonomous Robots and computers

MALE_1: that are able to think and make

MALE_1: decisions on their own are no longer a sci-fi fantasy.

MALE_1: AI has the potential to predict

MALE_1: natural disasters and diagnose

MALE_1: cancer better than humans ever could.

MALE_1: Many researchers predict we will have AI with

MALE_1: human level intelligence within the next 50 years.

MALE_1: Humans have dominated the planet

MALE_1: because we are the smartest.

MALE_1: We won the evolutionary game

MALE_1: of roulette because of our brainpower.

MALE_1: But when AI becomes smarter than

MALE_1: us what will happen next?

MALE_1: The robot apocalypse has

MALE_1: certainly become a popular plotline for

MALE_1: Hollywood movies but the rise of the machines and

MALE_1: the extinction of the human race is an actual threat,

MALE_1: according to AI researchers.

MALE_1: Ariel Conn is director of media

MALE_1: and outreach at the Future of Life Institute.

FEMALE_1: [MUSIC] We focus on access interests primarily,

FEMALE_1: so we're looking at especially ways that

FEMALE_1: technology could potentially destroy humanity.

MALE_1: So, are we all destined to die in

MALE_1: the coming robot apocalypse

MALE_1: because of super intelligent AI?

FEMALE_1: I have not heard

FEMALE_1: any AI researcher yet who thinks that is going happen.

FEMALE_1: No one is worried about

FEMALE_1: robots deciding that humans need to die.

FEMALE_1: I think a lot of our worries

FEMALE_1: are this idea of competent that we

FEMALE_1: ask the program to do

FEMALE_1: something and it's so focused on doing that,

FEMALE_1: that it's not as concerned about

FEMALE_1: what sort of damage it does in the process.

MALE_1: [MUSIC] We're nowhere near creating super intelligent AI,

MALE_1: but the concern isn't that

MALE_1: our machines will eventually turn against us,

MALE_1: it's that they won't be able to

MALE_1: consider how their actions impact us.

MALE_1: Meaning, a machine has to be able

MALE_1: to figure out if it should do something,

MALE_1: not just of it can do something.

FEMALE_1: [MUSIC] And also, how can the machines

FEMALE_1: learn this values on their own-

MALE_2: Clues might lie in how we

MALE_2: teach ourselves how to care about each other.

MALE_2: Empathy is feeling with someone

MALE_2: instead of feeling for someone.

MALE_2: It makes a large part of what has come to be known

MALE_2: as emotional intelligence, or EQ.

MALE_2: An emotionally intelligent person

MALE_2: may be considered well-adjusted,

MALE_2: considerate, and not impulsive.

MALE_2: An example is understanding that your boss'

MALE_2: angry outburst might not be because they're a jerk,

MALE_2: but might be because of the enormous

MALE_2: amount of stress they're under.

MALE_2: It is tapping into our deep human hardwiring

MALE_2: and understanding the causes and effects of emotions.

MALE_2: Research studies have tied

MALE_2: high IQ to better job performance,

MALE_2: strong leadership and superior problem solving.

MALE_2: Emotional intelligence is something we

MALE_2: try to teach from a very young age.

MALE_2: It's essential to help us learn how to

MALE_2: function in a world with other people.

JESSICA KREISLER: Hi I'm Jess, and I'm an elementary school teacher.

MALE_2: Jess spends a lot of time

MALE_2: teaching her students about their emotions.

JESSICA KREISLER: It's really important that we

JESSICA KREISLER: label emotions and teach them to students.

JESSICA KREISLER: Like a child could be crying

JESSICA KREISLER: that they don't know that that means

JESSICA KREISLER: sad unless you teach them that when someone is crying,

JESSICA KREISLER: that usually means that they're feeling sad.

JESSICA KREISLER: Unless you know what the word

JESSICA KREISLER: to describe the way you're feeling is,

JESSICA KREISLER: that emotion really doesn't mean that much.

MALE_2: Artificial intelligence is still in its infancy,

MALE_2: so teaching it to understand

MALE_2: emotions is a long way off probably.

MALE_2: But we are beginning to teach

MALE_2: our machines how to recognize emotions.

MALE_2: The MIT deep empathy AI scrubs

MALE_2: the photo sharing site Flickr for

MALE_2: pictures of the Syrian civil war,

MALE_2: and asks Internet users to choose

MALE_2: which ones gives them more of an emotional reaction,

MALE_2: so it can understand what images

MALE_2: make us feel more empathetic.

MALE_2: In the next five years, integrating AI with

MALE_2: emotional intelligence is projected

MALE_2: to be a multibillion dollar industry.

MALE_2: David Hanson creates robots that have

MALE_2: facial recognition software that can see

MALE_2: a person smile or frown and mimic that.

MALE_2: Companies have programmed

MALE_2: their customer service chat bots to

MALE_2: recognize angry customers and respond sympathetically.

MALE_2: Google has hired comedians to make

MALE_2: their personal assistant sound friendlier.

MALE_2: We're still nowhere close to getting AI to care about us.

MALE_2: That is a very human thing after all,

MALE_2: but we have already experienced how

MALE_2: they can make us feel something.

MALE_2: AlphaGo is an AI that plays the board game Go.

MALE_2: Go is a twenty five hundred year old game

MALE_2: that is considered more complex than chess.

MALE_2: It's super complicated and mastery

MALE_2: of the game is incredibly difficult.

MALE_2: Eighteen time world champion,

MALE_2: Lee Sedol is one of the best Go players in the world,

MALE_2: and in 2016, we sat down to

MALE_2: play a five game series against AlphaGo.

MALE_2: During the second game,

MALE_2: something happened that left the Go community stunned.

MALE_3: That's a very, that's a very surprising move.

MALE_2: Move 37 has become immortalized in the Go community.

MALE_2: When AlphaGo played a move that was so unexpected,

MALE_2: that it totally threw Sedol off his game.

MALE_4: Lee has left the room.

MALE_3: He left the room after that.

MALE_4: [OVERLAPPING] He left the room after that move.

MALE_2: The Go community was shocked because the move was

MALE_2: something a human would never ever do,

MALE_2: creating a possibility that no one had anticipated.

MALE_2: AlphaGo had shown human players how to view

MALE_2: the ancient game in an entirely new way.

MALE_2: We saw through the eyes of the computer

MALE_2: and felt something deeply human.

MALE_2: All of this stuff about super intelligent AI

MALE_2: is theoretical and really hard to

MALE_2: figure out and when it comes to

MALE_2: the potential destruction of humanity, pretty scary.

MALE_2: But for many, the limitless

MALE_2: possibilities of artificial intelligence,

MALE_2: what it can do for us, how it can help us,

MALE_2: and even what it can teach us,

MALE_2: far outweighs the dangers.

ARIEL CONN: The full reason we're

ARIEL CONN: trying to figure out how to design it

ARIEL CONN: safely is because we think it could be

ARIEL CONN: really good and that it's

ARIEL CONN: worth taking the time to do it right.

MALE_2: We just need to make sure we're creating

MALE_2: it with all of humanity in mind.