¿Cómo están afectando los algoritmos de Inteligencia Artificial a nuestros cerebros, opiniones, libertades y profesiones? ¿Nos curaremos de todo o terminaremos siendo dependientes de las máquinas? En nuestras manos está el delegar más decisiones en las máquinas, o no.
Category
😹
FunTranscript
00:00:00Artificial intelligence is here and we like it now.
00:00:05Artificial intelligence.
00:00:06Machines learning, robots.
00:00:08It's exciting and terrifying at the same time.
00:00:10Some of the most prominent names in science and technology
00:00:13warn about the dangers of the use of artificial intelligence
00:00:16for the manufacture of weapons.
00:00:19I think the development of artificial intelligence
00:00:22could be the end of the human race.
00:00:25For me, it's what's most frightening.
00:00:27We need legislators to say,
00:00:29put it on hold and make sure it's safe.
00:00:35Artificial intelligence has the potential
00:00:38to solve difficult problems of the present and the future.
00:00:42Artificial intelligence will have an impact on everything.
00:00:45It won't leave anything untouched.
00:00:47It's a renaissance, a golden age.
00:00:50Robots allow people to do more than they could on their own.
00:00:54What do we need artificial intelligence for?
00:00:57To solve problems and to be more prosperous and healthier.
00:01:01That has to be the purpose.
00:01:03With artificial intelligence, computers begin to learn,
00:01:07to think, to rewrite, to design their own programs.
00:01:11Artificial intelligence can make us hear better, see better.
00:01:14It gives us capabilities that we didn't have before.
00:01:17You can pair a human doctor
00:01:19with an artificial intelligence system
00:01:21to create new drugs that humans in the labs
00:01:24didn't have before.
00:01:26We could see an incredible life expansion.
00:01:30This all sounds like Star Trek, but believe me, it's going to happen.
00:01:35Will we be smart enough to make the right decisions
00:01:39about what we want as a society?
00:01:41Scientists are creating artificial intelligence today
00:01:45and setting the course for a new tomorrow.
00:01:47It depends on us how it affects our lives now
00:01:50and in the future.
00:01:52If we do things right with artificial intelligence,
00:01:55we can help life flourish like never before
00:01:58and get the best out of us.
00:02:00Do you think that intelligence is something that can only exist
00:02:03in biological organisms like humans?
00:02:06Artificial intelligence has been part of our lives
00:02:09without us realizing it.
00:02:11Can we get intelligent behavior from machines
00:02:15that are able to learn, able to reason?
00:02:19There is no definition.
00:02:22This is artificial intelligence.
00:02:24It's a series of practices and pieces that people put together.
00:02:27That's what artificial intelligence is.
00:02:29I think we should define it as the ability
00:02:32to fulfill complex goals.
00:02:34So artificial intelligence is simply intelligence
00:02:37that is not biological.
00:02:39Artificial intelligence has many definitions,
00:02:42but when machines manifest the ability to see,
00:02:45hear, understand and learn like us,
00:02:48they are considered to be artificially intelligent.
00:02:52I was expecting you to guess that I'm a robot
00:02:54because of the cables that come out of my body.
00:02:57And understanding language is one of the biggest challenges
00:03:00when it comes to creating intelligent machines.
00:03:03Hi, Hugo.
00:03:05Hi, Cynthia.
00:03:09Understanding language from the Star Trek computer.
00:03:13Next door on the right.
00:03:16You're welcome, Commander Riker.
00:03:26What's your name?
00:03:27Do you have a name?
00:03:28Yes.
00:03:29Samantha.
00:03:31Yes?
00:03:32Where did you get that name?
00:03:33I gave it to myself.
00:03:35However, getting machines to understand our human language
00:03:39is one of the most difficult tasks in artificial intelligence today.
00:03:43Languages are subtle and complicated
00:03:45and have difficulties for computers that use fixed rules.
00:03:49It can also be misinterpreted.
00:03:51Let's take the famous quote from Groucho Marx as an example.
00:03:54One morning he killed an elephant in his pajamas.
00:03:56How did he get into his pajamas?
00:03:58That's what I don't know.
00:04:00A computer has a hard time understanding
00:04:03what it's really saying.
00:04:05What we're trying to solve
00:04:07through the field of natural language processing
00:04:10is to start deciphering what it means
00:04:12when we use these words in sentences.
00:04:14How do sentences relate to each other?
00:04:17Is this word close to this other word?
00:04:19And why?
00:04:20So it's very easy for us,
00:04:22but very difficult for computers.
00:04:26Debating with a machine.
00:04:29The history of debate can be traced back to the ancient Greeks.
00:04:34It was often used to solve complex ethical problems.
00:04:38In its simplest form,
00:04:40a moderator proposes a proposal.
00:04:43Each side exposes an initial argument,
00:04:46a reply, and a conclusion.
00:04:48Everything must be done in a specific time.
00:04:51The participant in a debate
00:04:53must have critical thinking skills
00:04:56and extensive knowledge of a variety of topics.
00:04:59In a good debate,
00:05:01humor should be used to create a convincing argument.
00:05:04I'm not going to turn age into a topic of this campaign.
00:05:07I'm not going to exploit, for political purposes,
00:05:11the inexperience of my opponent's youth.
00:05:17The range of potential construction
00:05:19of debate arguments is simply infinite.
00:05:22There are so many decisions to be made
00:05:25to construct an argument that is original
00:05:29and that is able to respond to what an opponent presents
00:05:34and, in the end, to be able to infer all the arguments he made as a conclusion.
00:05:39All this, by peeling the layers,
00:05:41goes beyond the limits of what computers can do
00:05:44on each of the fronts.
00:05:47A computer can read hundreds of millions of articles in seconds,
00:05:51but it needs to be able to identify
00:05:54exactly a thousand or two thousand relevant words
00:05:57to debate with humans.
00:05:59For a computer, it is very difficult to understand the question,
00:06:02and for humans, it is exactly the opposite.
00:06:05We understand the question very quickly,
00:06:07but we do not have such effective arguments
00:06:09or enough memory to present relevant answers.
00:06:13As people, we are used to understanding the language
00:06:16without even realizing how difficult it is.
00:06:19For us, it is very clear how to build an argument,
00:06:22how to connect things together.
00:06:25When we try to teach a computer to understand the language,
00:06:28what we have are building blocks.
00:06:30We can teach a computer system to understand parts of the speech.
00:06:34We can teach it to identify concepts,
00:06:37and from there, perhaps identify similarities between concepts
00:06:41and then similarities between sentences.
00:06:43But you have to tell the system in such a way
00:06:45that it can intelligently agglutinate things.
00:06:55The debates you are going to see next
00:06:57have never been shown on television.
00:07:00I hope you enjoy watching live
00:07:02the debates between a totally automatic system
00:07:05and some expert human participants, so have fun.
00:07:11Our proposal for the first round, as you see behind me,
00:07:14is whether or not we should legalize sports betting.
00:07:17The Debater system has not previously seen any of these issues
00:07:21nor has it been part of its training,
00:07:23and the only contribution is that issue.
00:07:27On the side of supporting legalization,
00:07:29we have our Debater, IBM,
00:07:32and on the opposing side we have the charming Jaya Aichler.
00:07:37Greetings, Jaya. It's a pleasure to see you again.
00:07:40I hope you do better today than the last time
00:07:42I had to face your convincing arguments
00:07:45or at least show me some mercy.
00:07:51When you give a Debater a topic,
00:07:53the first thing it does is try to understand the meaning of that topic.
00:07:57It scans millions of articles
00:07:59to identify the arguments it can use
00:08:02to develop the defense of its four-minute stance.
00:08:05Debater does this using a single natural language processing,
00:08:10automatic learning,
00:08:12reasoning techniques to understand the underlying themes
00:08:16and present its arguments effectively.
00:08:20I would suggest that we legalize sports betting.
00:08:23Are we allowed to decide how people should act?
00:08:26Or are we allowed to decide on the actions they decide to take?
00:08:30The purpose of legislation is not to define morality or teach ethics.
00:08:36The Debater questions whether we can decide how other people should behave,
00:08:40and the answer is yes.
00:08:42First of all, when the ability of the person to make decisions has been impaired,
00:08:46and second of all, when there is damage to others.
00:08:50The Debater has to listen to the opposite for four minutes,
00:08:55not a single sentence.
00:08:58The opponent speaks fast,
00:09:00making complex arguments raise ethical considerations,
00:09:05and the system has to understand the essence of all that
00:09:09and present an adequate response.
00:09:12After presenting its initial argument,
00:09:14the Debater listens to the counterargument and elaborates its reply
00:09:18based on what it has heard.
00:09:20Haye has suggested that we have the obligation to protect our citizens.
00:09:24I would like to offer a different vision.
00:09:27We oppose using the excuse of protecting citizens
00:09:30to justify the limitation of freedom and life choices.
00:09:33Getting involved in the decision-making of another person
00:09:36when it comes to their own personal situation
00:09:39infringes the rights of that person.
00:09:41When we talk about bets,
00:09:43the only time you make a rational decision is with the first bet.
00:09:47Then people keep betting,
00:09:49but they no longer make rational decisions because of an addiction problem,
00:09:52and that's where the government has to say,
00:09:54this is dangerous for you.
00:09:56I think we all agree that sports betting has negative consequences.
00:10:00It often involves exploiting the weak,
00:10:02but the best way to help the weak is not to tell them what they don't have to do,
00:10:06but to change the environment and the conditions that surround them.
00:10:09Therefore, I think the proposal should remain.
00:10:12We should legalize sports betting.
00:10:18At the end of the debate, the public votes to decide the winner.
00:10:25So, IBM's debater has achieved a support of 10%,
00:10:30and Haye has received a support of 12%.
00:10:34It's almost a tie.
00:10:39I think the computer is getting better every time I debate against it.
00:10:42The first time I felt it was much better than him,
00:10:45but he's improving his learning
00:10:48of how to use his position to convince people that he's right.
00:10:55I've been debating for 16 years,
00:10:57since I was in high school.
00:10:59I train participants in debates,
00:11:01but when you see that a machine does the same job and is learning
00:11:04and doing it better and better, I think it's magical, it's amazing.
00:11:07Although it's a secondary level debate,
00:11:09the debater intends to impact industries far beyond the debate podium.
00:11:14We've moved from search to research,
00:11:16because when you search, you want to get relevant documents for your search topic.
00:11:21And we try to help in that research topic,
00:11:23in the sense of going deeper into one more layer.
00:11:26So when you search on a topic, you get a list,
00:11:29and when you search on a topic, in the sense that we're talking about,
00:11:32you get arguments in favor and against that topic.
00:11:35This can be applied in any decision-making.
00:11:38We can imagine all kinds of business decisions,
00:11:41or a lawyer trying to come up with an argument.
00:11:44In any kind of decision in which we want to have arguments in favor and against,
00:11:48we could benefit from the debate technologies.
00:11:51This technology focuses on issues that do not have a clear objective answer.
00:11:57Most of the issues that we find ourselves in in life are issues like that.
00:12:02So I think we're touching on something that is fundamental to our daily lives,
00:12:07and so we anticipate that it's going to have many different applications.
00:12:11An artificial intelligence that can argue
00:12:14could change the way we make decisions as a society.
00:12:18But what happens when we have a fraction of a second to make a decision?
00:12:22There are lives at stake, and artificial intelligence takes the lead.
00:12:33Artificial intelligence on the road.
00:12:36Artificial intelligence was part of science fiction,
00:12:40but it's already here.
00:12:42It's real, and it seems to be everywhere.
00:12:47It's used to translate the complex language of dolphins.
00:12:51Recently, a real Robocop has been added to the Dubai police,
00:12:56and maybe the robot Forge Pizza will deliver his next order.
00:13:00Here it is.
00:13:01However, few innovations have been expected with more impatience than the driverless car.
00:13:08FB2 to tower. We're heading to Chicago, please mark the itinerary.
00:13:13Received, FB2.
00:13:15It's in automatic mode.
00:13:17Hands off the wheel.
00:13:19Does it matter that it smokes?
00:13:20Not with this wonderful air conditioning.
00:13:22The cars that drive themselves have been an obsession of our collective imagination for decades.
00:13:28From the Pontiac, the fantastic car, to the Audi of Joe Robot.
00:13:33What are you doing?
00:13:34Driving.
00:13:35Manually?
00:13:36We're close to the reality of the driverless car.
00:13:40One of the great applications of artificial intelligence that we all know is the driverless car.
00:13:46It's something that will have an immense impact,
00:13:49but it's something that hasn't happened yet,
00:13:52despite seeing Google test vehicles on the road,
00:13:55or hearing about Tesla's autonomous mode.
00:13:58The first driverless car dates from 1987,
00:14:02so they've been driving themselves for a long time,
00:14:05but you need a lot of things to make them work in the real world,
00:14:09and how to get from one place to another,
00:14:12and I think the most immediate question is,
00:14:15how are they going to operate these driverless cars in the world that we live in?
00:14:19The reason why we haven't been able to do it yet,
00:14:22is because driving is difficult.
00:14:24Driving under normal conditions in which nothing happens is a solved problem,
00:14:29but facing difficult conditions,
00:14:32the driverless car,
00:14:34it's not a problem,
00:14:36it's a challenge.
00:14:39In the United States,
00:14:4135,000 people die in car accidents every year.
00:14:44In short, I think it's important that stupid people like me,
00:14:48stop controlling tons of metal that move at high speeds.
00:14:52The AI in the real world,
00:14:54it's not a problem,
00:14:56it's a challenge,
00:14:58but it's not a problem,
00:15:00it's not a problem,
00:15:02it's a challenge.
00:15:05The AI in the real world,
00:15:08it doesn't matter if it fails.
00:15:11If you look at a vacuum cleaner,
00:15:13it often makes mistakes.
00:15:15It can hit the same thing three times in a row.
00:15:18That momentary failure doesn't matter in that long-term task,
00:15:22but it does matter if our car hits the same wall three times.
00:15:26It's not going to work.
00:15:28It has to work perfectly when we're inside a large mass.
00:15:34Autonomous cars have been driven on test tracks,
00:15:38but in recent years,
00:15:40companies like Waymo, Tesla and Cruze have taken them to the roads.
00:15:45These pilot programs have successfully recorded millions of kilometers
00:15:50to improve their performance,
00:15:52but several recent fatal accidents have raised doubts
00:15:55about the reliability of technology.
00:15:58Vehicles without a driver are technically possible,
00:16:01but I think it will take time.
00:16:03The alternative is that we drive,
00:16:06and millions and millions of people die.
00:16:11Safety is the most important aspect of this technology,
00:16:14without a doubt.
00:16:16The most important elements that interest the consumer
00:16:19are very instinctive.
00:16:21Is this car seeing the things that I see?
00:16:23And is this car acting the way that I would act
00:16:26according to the things that it's seeing?
00:16:28Those are simple questions
00:16:30that consumers will raise to feel comfortable
00:16:33before they look at this technology.
00:16:37The artificial intelligence used in cars without a driver
00:16:40helps to answer the question,
00:16:42what do I see?
00:16:43That's called perception,
00:16:44which consists of taking all those images
00:16:46and deciphering what are the objects,
00:16:48what are the lanes, etc.
00:16:50Prediction is to discover how those objects move.
00:16:54In addition, the car has to ask itself
00:16:57based on that and based on where it wants to end,
00:17:00where am I going?
00:17:01That's planning.
00:17:02Some other challenges that cars without a driver have
00:17:05is understanding the social context
00:17:07that we give by sitting.
00:17:09For example, if we see some kids on the sidewalk playing soccer,
00:17:13we know that the ball may come out of the sidewalk,
00:17:16but the robot will only see people hitting a circular object.
00:17:19Systems improve by learning,
00:17:21and one of the biggest advances in artificial intelligence
00:17:24is deep learning,
00:17:25so that it recognizes,
00:17:26oh, that ball may come out of the sidewalk,
00:17:29but it needs to see a lot of scenarios
00:17:31and needs to know how to react.
00:17:33And then humans can help to say,
00:17:35that's right or that's wrong,
00:17:37and it will learn over time.
00:17:39It's like a Nobel Prize winning driver
00:17:41and it will learn over time.
00:17:44Thanks to the ability of artificial intelligence
00:17:47to learn from experience
00:17:49and the $80 billion invested in technology,
00:17:53autonomous cars could populate the streets
00:17:55in just over a decade.
00:17:57Researchers estimate that one in four cars
00:18:00will be without a driver in 2030.
00:18:02It is expected that shared vehicle flocks
00:18:05will reduce the number of cars
00:18:07and transform the urban landscape.
00:18:09We're going to be able to redesign our streets
00:18:11so that they can accommodate all kinds of transportation,
00:18:14bicycles, walkers, less cars.
00:18:17When I look at my seven-year-old son,
00:18:19he is not interested in having a car.
00:18:21He is interested in owning an autonomous car.
00:18:25And I'm hoping that in nine years,
00:18:27when those seven years turn into 16,
00:18:30that will happen in circulation.
00:18:33The safety and reliability of cars without a driver
00:18:36depends on how well artificial intelligence learns.
00:18:41How do machines learn?
00:18:44The basis of our human intelligence
00:18:47is the ability to learn.
00:18:49All human advances have been the result
00:18:51of the learning of our past work,
00:18:54of evolving in our understanding
00:18:56and of inventing new solutions.
00:18:59Computers have been trying for decades
00:19:01to make machines learn like humans.
00:19:04It started with a board game.
00:19:06That man is not playing the ladies
00:19:08against a computer, is he?
00:19:10Sure, he plays pretty well.
00:19:12The first great advance of artificial intelligence
00:19:15for the public media was a system
00:19:17that taught itself to play the ladies.
00:19:19It did it better than its programmer,
00:19:21and that was in 1957.
00:19:22By analyzing the data of thousands of games,
00:19:25the system learned the movements
00:19:27that would make it win more likely.
00:19:30In automatic learning,
00:19:32the computer analyzes the data,
00:19:34finds patterns and uses them
00:19:36to find the best route and reach its goal.
00:19:39Deep Blue has started with tower E1?
00:19:42Yes, tower F1 to E1.
00:19:44Tower F1 to E1.
00:19:46The game computers
00:19:48were an effective test bench
00:19:50for the advance of artificial intelligence.
00:19:53Wow!
00:19:54Deep Blue has made Kasparov give up.
00:19:57The leap in processing capacity,
00:20:00the increase of data
00:20:02and the advances of algorithms
00:20:04have contributed to the current revolution
00:20:06of artificial intelligence.
00:20:08Things have gone a million times better,
00:20:10largely due to the greater power
00:20:12of computers that can handle
00:20:14a greater number of data,
00:20:16but also thanks to intelligent mathematicians
00:20:18who have been studying them for many, many years.
00:20:21All these combined factors triggered
00:20:23a machine that learned by itself,
00:20:25which was slow and unreliable,
00:20:27and that little by little was faster and more reliable
00:20:30and accelerated a million times.
00:20:32That old idea didn't work very well
00:20:34for a long period of time,
00:20:36but data and computing
00:20:38have transformed artificial intelligence
00:20:41into something that we can now use
00:20:43for a great number of tasks.
00:20:48Every day we create more data
00:20:50and at a faster rate
00:20:52than at any time in human history.
00:20:54Every minute we watch
00:20:56more than 4 million YouTube videos.
00:20:58We run 3.5 million Google searches
00:21:01and tweet about half a million times.
00:21:05Artificial intelligence takes all that data
00:21:08and learns from it.
00:21:10It does it on a scale
00:21:12that we can't possibly measure.
00:21:14I mean, we couldn't look at a million rows of data
00:21:17and give it a sense of it.
00:21:19All this data offers infinite opportunities
00:21:22for computers to learn
00:21:24and computers to experiment
00:21:26and perfect automatic learning.
00:21:29As we all contribute to this data on the Internet,
00:21:32we are all participants in that experiment.
00:21:35We are swimming in seas of data.
00:21:38Every time we look at a smartphone,
00:21:40every time we click on a web link,
00:21:42all the sensors that surround us today,
00:21:44all of this is collecting information
00:21:46about our lives.
00:21:48Artificial intelligence systems watch us
00:21:50and offer options
00:21:52according to our preferences,
00:21:54like Spotify, which creates a playlist for us,
00:21:56or Facebook, which chooses the weather for us.
00:21:59But they also guide space rockets
00:22:01to land alone for SpaceX
00:22:03and help doctors
00:22:05diagnose cancer
00:22:07from a scanner.
00:22:09Over the last few years,
00:22:11there has been a spectacular advance
00:22:13because there have been a lot of things
00:22:15that experts thought would take years to happen.
00:22:17For example, the artificial intelligence systems
00:22:19that beat the best humans at chess
00:22:21have the necessary intelligence
00:22:23programmed by humans who know how to play chess.
00:22:25What needs to be achieved
00:22:27is that machines learn for themselves.
00:22:31The advances driven by automatic learning
00:22:34are changing some aspects of our lives.
00:22:37But as machines become more intelligent,
00:22:40the question we ask ourselves is,
00:22:44will they take my job away from me?
00:22:52Work with AI.
00:22:54Possibly the robots will take their job away from them.
00:22:56Machines will end up replacing people
00:22:58as labor.
00:23:00Will they end up replacing advances
00:23:02in robotics and artificial intelligence
00:23:04with more and more jobs?
00:23:08In the next 10 years,
00:23:10many jobs will be replaced.
00:23:12People will lose their jobs more and more,
00:23:14whether they are workers or businessmen.
00:23:16Nobody is safe.
00:23:21Historically, machines have been
00:23:23taking over human labor
00:23:25since the first industrial revolution.
00:23:27But the first factories were also created
00:23:29and offered new jobs
00:23:31that did not exist before.
00:23:34This cycle of loss and creation
00:23:36of jobs has been repeated
00:23:38over the decades.
00:23:44The Internet and the digital age
00:23:46have ended with jobs like
00:23:48cartographer, mechanic,
00:23:50travel agent and telephone operator.
00:23:52It has also redefined
00:23:54the music and entertainment industry.
00:23:56It has generated a global economy,
00:23:58the online market and new jobs
00:24:00that we had not heard of
00:24:02such as YouTube star,
00:24:04social network administrator
00:24:06and application developer.
00:24:08In recent decades,
00:24:10we have seen how the manufacturing methods
00:24:12have changed.
00:24:14We adapt, we think about new things
00:24:16that we have to do and we continue.
00:24:18Now we are at the door of another revolution
00:24:20in which advances in artificial intelligence
00:24:22will affect our lives.
00:24:24There will be millions of jobs
00:24:26that we cannot anticipate right now
00:24:28and that will arise
00:24:30due to artificial intelligence.
00:24:32What is happening today
00:24:34is not that machines are replacing humans,
00:24:36but that humans who work with artificial intelligence
00:24:38are replacing humans
00:24:40who work without it.
00:24:42So it is fundamental to think
00:24:44about how we can work better
00:24:46using this technology.
00:24:52In Boston, Massachusetts,
00:24:54Rethink Robotics creates robots
00:24:56with artificial intelligence
00:24:58that are partners of humans in factories.
00:25:02There is a lot of artificial intelligence
00:25:04in the background.
00:25:06It knows that it has to approach in a specific way,
00:25:08it knows how to position its arm
00:25:10and it also has some rules.
00:25:12For example, if I take this,
00:25:14the robot detects that it has not taken something
00:25:16and it knows that it has to try to take it again.
00:25:20We measure the torsions of each
00:25:22of the seven joints
00:25:24and that helps us identify, for example,
00:25:26how much force we are pushing
00:25:28when it carries something,
00:25:30how much force we are pushing
00:25:32when it carries something.
00:25:34We have a camera right here
00:25:36and the head can observe
00:25:38throughout the workspace
00:25:40and know when things change
00:25:42to take different measures.
00:25:44Sawyer learns by demonstration,
00:25:46not by programming,
00:25:48which makes it easier for anyone
00:25:50to train him to do tasks.
00:25:52Sawyer works in some very small factories
00:25:54where less than 10 people work.
00:25:56If you put a traditional robot there,
00:25:58the only thing you can do
00:26:00to optimize the cycle time
00:26:02is to have the pre-oriented clamp.
00:26:04We came up with the idea
00:26:06that instead of programming the robot,
00:26:08we would train it as it does with a person.
00:26:10You show it the task,
00:26:12you show it a few things it has to do
00:26:14and it manages to put it all together
00:26:16and solve whatever it is.
00:26:18What we've seen in some of the small factories
00:26:20is that some people with little technical knowledge
00:26:22have learned how to take advantage of it
00:26:24and have become that robot person
00:26:27Sawyer is changing the way
00:26:29people work with robots
00:26:31in factories all over the world.
00:26:33Even a small seal company
00:26:35in Richmond, Virginia
00:26:37has modernized its production
00:26:39after 40 years.
00:26:45I was in charge of shipping
00:26:47and when the boss told us
00:26:49he had bought a robot,
00:26:51I said, that's awesome,
00:26:53I don't know if I can use it,
00:26:56but I'm going to use it.
00:26:58His first position is this,
00:27:00to take, and from here
00:27:02we would move it to the next location.
00:27:04When you're training Sawyer,
00:27:06you give him some guidelines
00:27:08and his artificial intelligence
00:27:10creates the rest of the movements
00:27:12between those guidelines.
00:27:18We brought it to cut wood
00:27:20because it's the most dangerous task
00:27:22in the center of the work
00:27:24and also to fill the ink,
00:27:26which is very complicated
00:27:28and takes a lot of time.
00:27:30We would cut 500 pieces at a time,
00:27:32but he can use the saw at night
00:27:34and he can cut a thousand mounts
00:27:36before we come back in the morning.
00:27:45I've been working here for over 30 years.
00:27:49For me, it has changed a lot.
00:27:52If we get a big order of 100 mounts,
00:27:54I'd have to cut the 100 mounts,
00:27:56drill the holes,
00:27:58and put the handles.
00:28:01But now Sawyer can do all that.
00:28:05Do I like the task he's done?
00:28:07No, he can keep it.
00:28:14I used to work as a service representative
00:28:16to the client,
00:28:18then I was transferred to the workshop
00:28:20and became a shipment manager.
00:28:22But when Sawyer came to the company,
00:28:24I became a robotic technician.
00:28:27Now when I come to work,
00:28:29I know I'm going to train a robot
00:28:31to do something new
00:28:33that no one else has ever done.
00:28:35It's very exciting.
00:28:37Robots can do the tedious and repetitive things
00:28:39that bore people,
00:28:41and they can enjoy their work
00:28:43and be much more effective
00:28:45as human beings.
00:28:48Although Sawyer imitates our ability
00:28:50to learn through demonstration,
00:28:52other robots show similar agility
00:28:54to that of humans.
00:28:56The mini-robot Spot,
00:28:58from Boston Dynamics,
00:29:00can open and close doors.
00:29:03The Atlas robot can run, jump,
00:29:05and do parkour.
00:29:07And there are other robots
00:29:09that are created to look like us.
00:29:12If we go back thousands of years,
00:29:14for example, to Greek mythology,
00:29:16people already imagined robots.
00:29:19Machines made to reflect our image.
00:29:24There is this desire to emulate a human,
00:29:26to create a robot as similar as possible
00:29:28to a human.
00:29:30Let's dance.
00:29:32I guess in science fiction
00:29:34we refer to them as replicants.
00:29:36One interesting aspect of Blade Runner
00:29:38is an android that is totally indistinguishable
00:29:40from humans.
00:29:42They were designed to copy human beings.
00:29:45There is a whole movement,
00:29:47especially in Japan,
00:29:49to make robots as similar as possible
00:29:51to humans.
00:29:53And I suspect it can be disastrous.
00:29:57We don't want to start seeing ourselves
00:29:59in situations where we feel like
00:30:01machines have somehow the same rights
00:30:03as humans because they look like us.
00:30:07You could have the most sophisticated android
00:30:09that looks like a human,
00:30:11talks like a human,
00:30:14but if it lacks that glimpse of consciousness,
00:30:16it's not a person.
00:30:18Maybe we don't want all our machines
00:30:20to have consciousness.
00:30:22I mean, if we had a toaster,
00:30:24very, very smart, it would be like torture
00:30:26to force it not to do something
00:30:28other than toast bread all day, right?
00:30:30But if we had an attention robot
00:30:32for older people that really
00:30:34had to know the person
00:30:36it was helping and create a bond,
00:30:38I think we would prefer
00:30:40that the robot had consciousness
00:30:42and that all the things it said
00:30:44were really sincere.
00:30:46Only consciousness can provide
00:30:48a foundation for morality,
00:30:50meaning and purpose,
00:30:52all the things that really matter to us.
00:30:54If we can figure out
00:30:56what information process
00:30:58corresponds to real subjective
00:31:00conscious experiences,
00:31:02then we could build machines
00:31:04that feel a real empathy
00:31:06and care for us.
00:31:08We are teaching artificial intelligence
00:31:10to think and reason
00:31:12in revolutionary ways.
00:31:14Learning for life.
00:31:20Everyone learns in a different way
00:31:22and any tool
00:31:24that can be introduced
00:31:26to personalize things
00:31:28that help learning,
00:31:30especially in early ages,
00:31:32in preschool,
00:31:34establishes the foundations
00:31:36for learning for life.
00:31:38We learn between 0 and 8 years
00:31:40more than any other time
00:31:42in our life.
00:31:44Studies have shown
00:31:46that children who start
00:31:48school later are behind
00:31:50and all schools have drastic
00:31:52differences that depend
00:31:54on the state in which we are
00:31:56or the neighborhood in which we live.
00:31:58But learning should not depend
00:32:00on the postal code.
00:32:02How are you? Good morning.
00:32:04Hello. Good morning.
00:32:06I have 23 children.
00:32:08I would love to be able
00:32:10to individualize everything.
00:32:12Good job.
00:32:14But that's very difficult
00:32:16because I am alone
00:32:18and in the kindergarten
00:32:20you don't always have
00:32:22extra help,
00:32:24but the SESAM application
00:32:26has been a blessing.
00:32:28Ed is a vocabulary tutor
00:32:30for the first years of childhood
00:32:32that focuses on preschool students.
00:32:34They use about 30 million
00:32:36more words at 4 years
00:32:38than those with lower incomes
00:32:40and because of that
00:32:42they have a harder time
00:32:44learning throughout their life.
00:32:46Early stages of development
00:32:48determine the success
00:32:50of a child in life,
00:32:52so I think an application
00:32:54like this can close that gap.
00:32:56All right, kids,
00:32:58I'm going to give you the iPads.
00:33:00At the beginning of the application
00:33:02you're going to play
00:33:04all the World Show games
00:33:06that you can before the song ends.
00:33:08When you play World Show
00:33:10he says a word.
00:33:12Pirouette.
00:33:14And you have to touch
00:33:16the image that matches
00:33:18the word and then he eats it.
00:33:20Delicious.
00:33:22When you get it wrong
00:33:24he spits it out.
00:33:26Try it again.
00:33:28In World Show
00:33:30Watson tracks the answers
00:33:32and assesses the child's knowledge
00:33:34of that particular word.
00:33:36And the words that show up next
00:33:38will be based on how
00:33:40those assessments have been done.
00:33:42So then he adapts words
00:33:44for each child.
00:33:46Delicious.
00:33:48At first glance
00:33:50it may look like a children's game
00:33:52but artificial intelligence
00:33:54collects data and algorithms
00:33:56that assess the child's vocabulary
00:33:58based on his answers.
00:34:00He predicts and decides
00:34:02what words to focus on
00:34:04because each child has a particular
00:34:06learning rhythm.
00:34:08Splash the cake when I say half.
00:34:10Ready? Half.
00:34:12This vocabulary application
00:34:14includes adaptive games
00:34:16and Sesame Street videos
00:34:18that teach new words
00:34:20and reinforce learning.
00:34:22Hi, I'm Mike from Sesame Street
00:34:24and I'm looking for the word on the street.
00:34:26I'm looking for the word on the street.
00:34:28I'm looking for the word on the street.
00:34:30I'm looking for the word on the street.
00:34:32I'm looking for the word on the street.
00:34:34I'm looking for the word on the street.
00:34:36I'm looking for the word on the street.
00:34:38I'm looking for the word on the street.
00:34:40I'm looking for the word on the street.
00:34:42I'm looking for the word on the street.
00:34:44I'm looking for the word on the street.
00:34:46I'm looking for the word on the street.
00:34:48I'm looking for the word on the street.
00:34:50I'm looking for the word on the street.
00:34:52I'm looking for the word on the street.
00:34:54I'm looking for the word on the street.
00:34:56I'm looking for the word on the street.
00:34:58I'm looking for the word on the street.
00:35:00I'm looking for the word on the street.
00:35:02I'm looking for the word on the street.
00:35:04I'm looking for the word on the street.
00:35:06I'm looking for the word on the street.
00:35:08I'm looking for the word on the street.
00:35:10I'm looking for the word on the street.
00:35:12I'm looking for the word on the street.
00:35:14I'm looking for the word on the street.
00:35:16I'm looking for the word on the street.
00:35:18I'm looking for the word on the street.
00:35:20I'm looking for the word on the street.
00:35:22Artificial intelligence is here.
00:35:24Whether we like it or not.
00:35:26Even the kids told me,
00:35:28could you invent a robot that did something for you?
00:35:30And I thought, wow,
00:35:32it's wonderful that 5 and 6 year olds think like that.
00:35:34Could you make a robot
00:35:36that would help you do something,
00:35:38like your homework,
00:35:40if you're not sure about the answer?
00:35:44Artificial intelligence games
00:35:46focus on little kids
00:35:48during the years that are fundamental
00:35:50for the development of the brain.
00:35:52But how could it be adapted
00:35:54to teach us beyond childhood
00:35:56to adulthood?
00:35:58As we reach higher levels
00:36:00and reach adulthood,
00:36:02everything is personalized
00:36:04and artificial intelligence
00:36:06will be able to teach us a way of learning
00:36:08for life.
00:36:10This technology has been developed
00:36:12and tested in 800 students
00:36:14from 12 universities.
00:36:16It's well known in research
00:36:18that individual teaching encourages learning.
00:36:20There's a study that says
00:36:22that 97% of students
00:36:24who received individual teaching
00:36:26were among the 20% of the best in their class.
00:36:28Despite the value of traditional tutoring,
00:36:30it's a difficult approach
00:36:32to implement for all students
00:36:34when and where they need it.
00:36:36Everyone has a different background,
00:36:38different challenges,
00:36:40but this technology will personalize
00:36:42education for each individual student.
00:36:44The Watson Tutor
00:36:46is a virtual individual tutor
00:36:48who talks to the students.
00:36:50A student said that Watson
00:36:52was his classmate,
00:36:54and that underlines the fact
00:36:56that they have that feeling
00:36:58of having someone there
00:37:00while they're studying.
00:37:02So, depending on what the student says,
00:37:04whether or not his answer is correct,
00:37:06completely wrong,
00:37:08or maybe partially correct,
00:37:10the tutor starts to guide him
00:37:12back to demonstrating his knowledge.
00:37:14I had the feeling
00:37:16of having a professor
00:37:18to communicate with.
00:37:20To me, it looks like a person.
00:37:22Really, it does.
00:37:24From the instructor's point of view,
00:37:26we're actually able to see
00:37:28what questions students ask.
00:37:30So, if we start to see
00:37:32that there are a lot of students
00:37:34who really don't understand a concept,
00:37:36we can adjust our classes
00:37:38to focus more on that concept
00:37:40and maybe find a different way
00:37:42to communicate with students.
00:37:44What is this?
00:37:46A blueberry.
00:37:48Before you used the program,
00:37:50it took a long time to study.
00:37:52What is that?
00:37:54Jungles.
00:37:56You got it right.
00:37:58You got it right.
00:38:00Five out of five.
00:38:02You did very well.
00:38:04This is something that you can
00:38:06carry along with you
00:38:08throughout your life.
00:38:10We're actually in the first stages
00:38:12of the process of building
00:38:14the maximum learning tool.
00:38:41From the man of the six million dollars
00:38:44to Iron Man,
00:38:46we've been fascinated
00:38:48by the idea of merging with machines.
00:38:50As artificial intelligence advances,
00:38:52the line between our bodies
00:38:54and technology
00:38:56disappears every day.
00:38:58Thanks to human improvement,
00:39:00we can work to give ourselves
00:39:02new capabilities that we never even dreamed of.
00:39:04An evolution that we never even dreamed
00:39:06that we would have.
00:39:08We can't just think,
00:39:10oh, we want to have more intelligence
00:39:12or more memory,
00:39:14but what about the ability to see better,
00:39:16to hear better,
00:39:18to experience the world
00:39:20that surrounds us in a different way?
00:39:22Mario Kart has a motor
00:39:24that drives the car for you.
00:39:26You just have to keep the accelerator
00:39:28and the computer takes care of the steering.
00:39:30That way I can actually play with them.
00:39:32Am I still fifth?
00:39:34Seventh.
00:39:36Because Franklin and I
00:39:38are still learning.
00:39:40Yes, we are learning.
00:39:44I was born with a retinal disease
00:39:46called retinitis pigmentosa.
00:39:48What it basically is is that the eye
00:39:50is dying over time.
00:39:52At the beginning,
00:39:54perhaps you lose your peripheral vision
00:39:56and once it starts to accentuate,
00:39:58you get night blindness.
00:40:00That is, when it's dark,
00:40:02you can't see anything.
00:40:04And then you start to lose
00:40:06the perception of light.
00:40:08At the age of 18, I went blind.
00:40:10I couldn't get mad
00:40:12at the fact that I went blind.
00:40:14I had to adapt and do it quickly.
00:40:18And running was a quick way
00:40:20to get it.
00:40:24When I started running,
00:40:26I did it on a football field
00:40:28that was behind my house
00:40:30because I thought,
00:40:32if there's no football field,
00:40:34there's no one around,
00:40:36and I can go around the field.
00:40:38Then I learned to run on foot.
00:40:40It's pretty good
00:40:42if you're willing to run into
00:40:44things like streetlights,
00:40:46you remember where everything is.
00:40:48And then after a few years,
00:40:50I started training for the marathon.
00:40:52Running quickly
00:40:54became more than a hobby for Simon.
00:40:56He started competing
00:40:58in marathons all over the world.
00:41:00Most of the blind people
00:41:02are tied to guide runners,
00:41:04but Simon wanted independence.
00:41:06I started thinking about
00:41:08how I could learn to run
00:41:10on open roads
00:41:12and move by myself.
00:41:14That's how I started
00:41:16working with WearWorx.
00:41:18This is like our garage
00:41:20in Silicon Valley,
00:41:22but we're not in Silicon Valley,
00:41:24we're in New Jersey.
00:41:26I built this suit
00:41:28to get a reading of how fast
00:41:30you were punching.
00:41:32It was the first time
00:41:34we worked in haptics.
00:41:36Haptics is the tactile science
00:41:38related to our interactions
00:41:40with technology.
00:41:42It's what happens when
00:41:44we feel our phone vibrate
00:41:46instead of hearing it.
00:41:48Does it work?
00:41:50Yes.
00:41:52Perfect.
00:41:54I think I read the article
00:41:56and I said,
00:41:58why don't we work together?
00:42:00I think we're stuck in a loop.
00:42:02At that point,
00:42:04we thought maybe we could
00:42:06make it go around the apple.
00:42:08I don't really know
00:42:10which way to go.
00:42:12But Simon said,
00:42:14I want to challenge you
00:42:16to see if you can make
00:42:18the marathon run.
00:42:20Simon and the WearWorx team
00:42:22worked for months
00:42:24trying to figure out
00:42:26how to make it go around
00:42:28the apple.
00:42:30This year, all we're trying
00:42:32to do is make it go around
00:42:34the whole community of blind people.
00:42:36Six weeks before the event,
00:42:38the team got together
00:42:40in Central Park
00:42:42to try the prototype.
00:42:44Three, two, go.
00:42:46The Weiband works in such a way
00:42:48that it creates a virtual corridor
00:42:50that's slightly wider than a human
00:42:52being.
00:42:54The Weiband,
00:42:56through haptic information,
00:42:58ensures that you're always
00:43:00within that corridor.
00:43:02There we go.
00:43:04Coming to the right.
00:43:06Good.
00:43:08You know you're in the corridor
00:43:10because you don't get comments.
00:43:12If you don't get haptic vibrations,
00:43:14it means you're going in the right direction.
00:43:16It's all good.
00:43:18As soon as you step out of the corridor,
00:43:20do you feel anything now?
00:43:22Vibrations? No, none.
00:43:24Perfect.
00:43:26These haptic corridors
00:43:28will be created based on millions
00:43:30of data points collected
00:43:32by people using the Weiband.
00:43:34According to the information,
00:43:36it will offer the safest and most adequate route.
00:43:38Unlike the normal GPS,
00:43:40the route is very precise
00:43:42and will let you know if you're
00:43:44deviating just a few meters.
00:43:46We think of Waze for pedestrians.
00:43:48Read what everyone is doing
00:43:50to offer the most effective route.
00:43:52But as a blind person,
00:43:54I want the route with less traffic.
00:43:56So imagine an intelligent
00:43:58navigation device.
00:44:00For the race,
00:44:02Simon is going to include
00:44:04the Weiband,
00:44:06a haptic sensor that detects
00:44:08other people in front of him.
00:44:10It works like a car parking sensor,
00:44:12but instead of a beep,
00:44:14it emits a haptic vibration.
00:44:16So we're less stressed.
00:44:18But trust the technology.
00:44:20So we're less stressed.
00:44:30Attention.
00:44:32First series of corridors.
00:44:34Corrals for the first series
00:44:36of corridors close in less than five minutes.
00:44:38This is really the first time
00:44:40I've had the opportunity
00:44:42to compete alone in a city marathon.
00:44:44It's something
00:44:46I've been waiting for a long time.
00:44:58In New York, there's a lot of people running.
00:45:00I think it's the biggest marathon
00:45:02and there's about 50,000 runners.
00:45:10It's just a few laps.
00:45:12There's fantastic support equipment
00:45:14and a lot of noise
00:45:16throughout the race.
00:45:20So it's very hard
00:45:22for a blind runner.
00:45:26In those moments,
00:45:28he was making progress
00:45:30thanks to technology.
00:45:34Come on!
00:45:38Simon was able to run
00:45:40for 21 kilometers of the 42.
00:45:42But about halfway through,
00:45:44things started to fail.
00:45:52The buildings and the metal bridges
00:45:54deflected the direction of the GPS
00:45:56and the mobile signal
00:45:58of tens of thousands of runners
00:46:00and spectators interfered
00:46:02with the signal of the wayband.
00:46:06In addition, Simon faced another obstacle,
00:46:08time.
00:46:10As the rain fell,
00:46:12the ultrasonic sensor
00:46:14stopped working.
00:46:18At that moment,
00:46:20Simon had to rely on his friends
00:46:22to guide him to the finish line.
00:46:30Slow down, Simon!
00:46:32You're out!
00:46:34Get a little closer to me, Simon!
00:46:38Woohoo!
00:46:40Woohoo!
00:46:42Woohoo!
00:46:44Woohoo!
00:46:46Woohoo!
00:46:48It didn't work very well.
00:46:50But for me,
00:46:52finishing the race
00:46:54means that things are moving forward.
00:46:56Artificial intelligence
00:46:58is about to upgrade
00:47:00all of human capacity.
00:47:02It's either improving our senses
00:47:04or our knowledge.
00:47:06to make a difference between us and the type of system,
00:47:09it should be considered a partner.
00:47:15In any situation like this,
00:47:17you have to take into account what you can learn,
00:47:19what you can improve.
00:47:20And today it has been like that.
00:47:21Actually, I don't run for the times.
00:47:24Why do you run?
00:47:26To see if things are possible.
00:47:33The real motivation to finish the marathon
00:47:35was to move forward, to make technology accessible.
00:47:39Because perhaps it's the way that people like me
00:47:42are willing to be the voice that gives the can
00:47:44so that this technology is accessible.
00:47:48The development of technology requires a long process.
00:47:51The Wright brothers' flying machines
00:47:54were very different from today's Boeing 787.
00:47:57Artificial intelligence is no different.
00:47:59Every step and stumble is a progress.
00:48:02This is a concept test
00:48:04to show what can happen years later.
00:48:09There's no shortage of things that can go wrong in the human body,
00:48:12but through human improvement,
00:48:14we can begin to ease those problems.
00:48:19I go to my master's classes at Sheffield University.
00:48:24My master's is in automatic learning.
00:48:26And I suppose the motivation to get into it
00:48:29is to ensure that, to move forward,
00:48:31that technology is accessible for people like me.
00:48:35In the future, the meaning of being an individual
00:48:38will become that diffuse idea
00:48:40that we are not only limited by our flesh and blood,
00:48:43but that there will be a synergy between us
00:48:45and perhaps biological parts that integrate with synthetic parts.
00:48:48We'll see how to reinvent ourselves.
00:48:51Artificial intelligence is going to improve our bodies and minds
00:48:55in ways that we didn't imagine were possible.
00:48:58At the same time, it will see us, listen to us,
00:49:01and analyze us in such a way
00:49:03that it will challenge our idea of privacy.
00:49:12A sixth synthetic sense.
00:49:14Artificial intelligence undoubtedly presents problems
00:49:17when it comes to our privacy.
00:49:19Only the power it will have at surveillance,
00:49:23whether it be facial recognition,
00:49:25any kind of biosignature that we can exert on the world,
00:49:29it will be able to detect and know that we are us.
00:49:32But more than that, it will be able to take back what we say
00:49:35and know what our motivations are.
00:49:39In our effort to create computers
00:49:41that interpret the world as we do,
00:49:43we have installed cameras,
00:49:45microphones and artificial intelligence
00:49:47so that they understand what they see and what they hear.
00:49:50But what are the limits of their perception?
00:49:53With artificial vision,
00:49:55artificial intelligence can navigate three-dimensional space,
00:49:59allow cars without a driver to see their surroundings
00:50:02and understand the facial expressions of a child.
00:50:05I'm glad to see you again.
00:50:08Artificial vision can also analyze faces
00:50:12to identify individuals through facial recognition.
00:50:16Look at a face.
00:50:18Extract the distinctive features,
00:50:20such as the size and position of the eyes,
00:50:23nose, cheeks and jaw.
00:50:25And then look for other images comparing those features.
00:50:29This is what Facebook does
00:50:31when it asks us if we want to tag ourselves in a photo.
00:50:34Or when Apple's iPhone X unlocks
00:50:37by scanning the owner's face.
00:50:42Even so, where is the balance
00:50:44between comfort, privacy and security?
00:50:47Maybe that's the question
00:50:49the next time you get on a plane.
00:50:51Right now,
00:50:52a dozen American international airports
00:50:55use facial recognition to check in passengers.
00:50:59Trump's government wants
00:51:01facial biometric scans of all Americans
00:51:04who make international trips.
00:51:06They take digital photographs of passengers in cabins
00:51:09that are then compared to the images of their passports.
00:51:12JetBlue claims that technology
00:51:14will eliminate boarding cards and speed up the trip,
00:51:16but privacy advocates say it's going too far.
00:51:20The first thing we have to ask ourselves
00:51:22is if we have that privacy right now.
00:51:24Anyone can take a picture of us anywhere,
00:51:27at any time, and we don't even know it.
00:51:29So if someone wants to take our privacy,
00:51:31they can do it.
00:51:33The question is whether we agree
00:51:35with the reduction of privacy sponsored by the state,
00:51:38its elimination or reduction sponsored by the companies.
00:51:41In his attempt to put a stop to crime
00:51:44and terrorist attacks,
00:51:46countries around the world
00:51:48use facial recognition in places of great affluence
00:51:51like sports events, concerts and train stations.
00:51:55China, unlike others,
00:51:57has achieved great success
00:51:59in capturing suspects
00:52:01through its facial recognition systems.
00:52:04According to the police,
00:52:06this technology allowed the arrest of 375 suspects,
00:52:10including 39 fugitives,
00:52:13just in Guiyang.
00:52:15It is considered a progress,
00:52:17but it also raises ethical questions,
00:52:19especially in a country
00:52:21with 170 million surveillance cameras.
00:52:24Currently, China is building
00:52:26the world's most powerful
00:52:28facial recognition system.
00:52:30The goal is to be able to identify
00:52:32any Chinese citizen
00:52:34through their face in seconds.
00:52:36The Chinese government
00:52:38also uses it to promote good behavior.
00:52:41Four Chinese cities
00:52:43use facial recognition
00:52:45to embarrass reckless pedestrians.
00:52:47They are recorded and compared
00:52:49in the database.
00:52:51Their personal information,
00:52:53including their home,
00:52:55is shown on the next screens.
00:52:57These applications are part of
00:52:59the government's Eagle Eyes program,
00:53:01which connects public and private security cameras
00:53:04to create a gigantic national level
00:53:06surveillance platform.
00:53:08What we know about privacy
00:53:10will become something of the past,
00:53:12a kind of question of order.
00:53:14If we look at the Orwellian stories,
00:53:17in which the Big Brother watches us,
00:53:20in which the state wins us ground
00:53:22with its technology against our will,
00:53:25the irony is that we are
00:53:27voluntarily introducing this technology
00:53:29into our homes.
00:53:31Just by pressing a Google Home
00:53:33or Amazon Echo switch,
00:53:35you can monitor everything
00:53:37that we and our families
00:53:39say.
00:53:41Other facial recognition applications
00:53:43could challenge our ideas
00:53:45about what computers
00:53:47should know about us.
00:53:49Researchers have developed
00:53:51a type of artificial intelligence
00:53:53that can identify individuals
00:53:55who are homosexual or heterosexual.
00:53:57It's crazy.
00:53:59In 2017,
00:54:01a group of researchers
00:54:03from Stanford University
00:54:05claimed that they had developed
00:54:07an artificial intelligence
00:54:09that could distinguish
00:54:11between homosexual and heterosexual people
00:54:13based on their features
00:54:15and facial expressions.
00:54:17The algorithm looked at images
00:54:19from a dating site
00:54:21and determined if men were
00:54:23gay or heterosexual
00:54:25with an accuracy of 81%
00:54:27and for women, 74%.
00:54:29How much information
00:54:31do our faces transmit about us?
00:54:33And how could that information be used?
00:54:35Is it possible for a criminal
00:54:37to observe his face
00:54:39without any information
00:54:41about his background?
00:54:43A company says it's possible.
00:54:45We can analyze thousands of pictures
00:54:47from any source,
00:54:49like a surveillance camera,
00:54:51in just a few seconds.
00:54:53In this case,
00:54:55they selected everyone as thieves.
00:54:57Now we can see
00:54:59that we have an accuracy
00:55:01of 90%.
00:55:03A research company
00:55:05based in Tel Aviv, Israel
00:55:07predicts personality traits
00:55:09based only on the facial image,
00:55:11not on data bases
00:55:13only on our face.
00:55:15We constantly observe people
00:55:17and think if they are nice,
00:55:19if they are not nice,
00:55:21if they are attractive or not.
00:55:23However, machines don't care
00:55:25about our race, gender,
00:55:27political opinion, etc.
00:55:29Our idea is to teach machines
00:55:31the image of a face.
00:55:33FaceOption
00:55:35reads the face and assigns it
00:55:37a probability of being
00:55:39a specific type of criminal.
00:55:41The authorities can interrogate
00:55:43the suspects as a consequence.
00:55:45This is already being used
00:55:47by the Israeli Defense Forces
00:55:49and the National Security of the United States.
00:55:51Unlike facial recognition
00:55:53that compares faces with data bases,
00:55:55FaceOption is not linked to any.
00:55:57It does not compare a face
00:55:59with criminal records.
00:56:01It only looks at a face
00:56:03and determines how suspicious it seems
00:56:05based on previous data.
00:56:07The idea that there is
00:56:09a type of criminal face
00:56:11is very old.
00:56:13The danger is that
00:56:15obviously we cannot control
00:56:17our bone structure
00:56:19or our facial structure.
00:56:21Therefore, it is another way
00:56:23of stereotyping.
00:56:25It's very controversial.
00:56:27It makes us think of
00:56:29the famous science fiction movie
00:56:31Minority Report,
00:56:33in which people could predict
00:56:35who was going to commit a crime
00:56:37and could be arrested
00:56:39before he did.
00:56:41He is arrested for the
00:56:43future assassination of Sarah Marks.
00:56:45What? I have not done anything.
00:56:47Why do we want to use
00:56:49artificial intelligence?
00:56:51In fact, it should be used
00:56:53to improve the ability of humans
00:56:55to use fire to heat
00:56:57our homes in winter
00:56:59and it can also be used
00:57:01to start a fire.
00:57:03So, in my opinion,
00:57:05the question is,
00:57:07what can we do now
00:57:09to ensure that artificial
00:57:11intelligence allows
00:57:13humanity to flourish
00:57:15instead of faltering?
00:57:17A part of what we have done
00:57:19is to transform
00:57:21faulty computers
00:57:23into our counterparts.
00:57:25Put your hands in front like this
00:57:27and turn them to the sides.
00:57:29They usually diagnose it
00:57:31at 60 or 70 years old,
00:57:33to me at 50.
00:57:35And that side.
00:57:41It's clear that it costs more,
00:57:43right? Yes, that's right.
00:57:45Good. You think your body
00:57:47is disappointing you.
00:57:49You observe it and you think,
00:57:51get off the bed.
00:57:53Now walk to the wall.
00:57:55Parkinson's disease
00:57:57is a disorder of motion
00:57:59caused by the degeneration
00:58:01of a very specific cell
00:58:03brain cells
00:58:05that are involved in the
00:58:07generation of smooth and controlled
00:58:09movement. Unfortunately,
00:58:11the only treatment we can offer
00:58:13patients is from the 1960s.
00:58:15Jonathan was not yet born.
00:58:22We wanted to identify drugs
00:58:24that were already on the market
00:58:26for certain diseases and that
00:58:28could be used to treat
00:58:30Parkinson's disease.
00:58:32The complicated part of the
00:58:34reusing of the drug is that
00:58:36it is not humanly possible to
00:58:38read all the information about
00:58:40the properties of all those
00:58:42drugs, and that's where artificial
00:58:44intelligence can really help.
00:58:46Parkinson's disease still has
00:58:48no cure, but researchers
00:58:50in artificial intelligence have
00:58:52the ability to change the range
00:58:54and speed at which drugs are
00:58:56discovered. We're starting to
00:58:58see that artificial intelligence
00:59:00is opening up a whole new way
00:59:02of research, and that's something
00:59:04that I don't think would be
00:59:06humanly possible without
00:59:08artificial intelligence. It's
00:59:10just the beginning of the
00:59:12medical applications of
00:59:14artificial intelligence.
00:59:16Artificial intelligence systems
00:59:18can tell us, for example, about
00:59:20our propensity for certain types
00:59:22of cancer, or the effectiveness
00:59:24of a drug in us, as opposed to
00:59:26another patient. That could be
00:59:28really the beginning of a
00:59:30revolution in medicine.
00:59:32Let's go to the tree!
00:59:36You and I, we have a general
00:59:38intelligence. We can do many
00:59:40different things and in many
00:59:42different contexts, and we don't
00:59:44need a computer or remotely
00:59:46close to be able to do them.
00:59:48They're extremely limited in
00:59:50terms of what they can do. They
00:59:52can play chess or go.
00:59:54Deep Blue has responded with
00:59:56the C1 to D3 pin.
00:59:58However, the primary objective
01:00:00is to create what we call
01:00:02Artificial General Intelligence,
01:00:04or AGI, and that would be
01:00:06something very deep, because
01:00:08now we're only in front of one
01:00:10entity or one mind, if you
01:00:12prefer, which in fact could
01:00:14be called Artificial Intelligence
01:00:16when it means singularity.
01:00:18When an Artificial Intelligence
01:00:20is as or more intelligent than
01:00:22we are, and is also aware of
01:00:24itself, that is what people
01:00:26are afraid of. That intelligence
01:00:28is so smart as to surpass
01:00:30our own control.
01:00:34Yes, of course.
01:00:36Singularity is simply an
01:00:38event beyond which nothing
01:00:40can be predicted.
01:00:42That is also an appropriate
01:00:44name for the explosion of
01:00:46intelligence, where in a short
01:00:48time the machines could be not
01:00:50only a little more intelligent
01:00:52than we are, but much more
01:00:54capable. What is an explosion
01:00:56of intelligence? In physics,
01:00:58an explosion is any phenomenon
01:01:00in which the duplication of
01:01:02a certain amount is repeated
01:01:04over and over again. Something
01:01:06small quickly becomes very
01:01:08large. If what happens is that
01:01:10the population continues to
01:01:12duplicate, it is a demographic
01:01:14explosion. And if what continues
01:01:16to happen is that intelligence
01:01:18continues to duplicate, it is
01:01:20an explosion of intelligence.
01:01:22This is the data cell of the
01:01:24processing unit. If each iteration
01:01:26of our computer can design
01:01:28something that is a certain
01:01:30factor better than the
01:01:32predecessor, then it will
01:01:34continue to duplicate, duplicate
01:01:36and duplicate. If that computer
01:01:38is capable of programming
01:01:40artificial intelligence, it will
01:01:42discover ways to rewrite its
01:01:44software to make it much
01:01:46better. And then it will do it
01:01:48many times. Which could do it
01:01:50not only a little better than
01:01:52we are, but much better than
01:01:54we are. I have often thought
01:01:56that artificial superintelligence
01:01:58will probably not be built by
01:02:00a human being or human team.
01:02:02I would not be surprised if we
01:02:04soon had machines that taught
01:02:06artificial intelligence.
01:02:08The turning point of the human race
01:02:10would be the creation of what we
01:02:12could call artificial intelligence
01:02:14on a human level, comparable to
01:02:16the arrival of an alien civilization
01:02:18on Earth.
01:02:20But artificial intelligence
01:02:22is not an alien species.
01:02:24It is something that we design,
01:02:26and in fact we have to be very
01:02:28sure that it is not an alien
01:02:30species with its own objectives,
01:02:32that they are not in tune with
01:02:34their own objectives.
01:02:36Or like, for example, Ex Machina,
01:02:38in which, obviously, the robots
01:02:40were designed in such a way that
01:02:42their designers did not have
01:02:44control over their objectives.
01:02:46The movies make us worry
01:02:48about something wrong.
01:02:50We should not worry about
01:02:52evil, but about competence
01:02:54and inadequate objectives.
01:02:56So if I'm in charge of building
01:02:58a green energy hydroelectric
01:03:00dam and there's an anthill
01:03:02in the middle, that would be
01:03:04harmful to the ants, but that
01:03:06does not mean that I'm an evil
01:03:08enemy of the ants.
01:03:10It's just that my objectives
01:03:12are not in tune with those
01:03:14of the ants, and they are
01:03:16bad for them.
01:03:18We should never put humans
01:03:20in the situation of those ants.
01:03:22When you're facing something
01:03:24that is more intelligent,
01:03:26you can't foresee all the
01:03:28ideas that can happen to you
01:03:30and you can't foresee how
01:03:32the world is going to sit
01:03:34on its seats eating those
01:03:36giant sandwiches.
01:03:38That's not the future we want,
01:03:40and it seems to mean that
01:03:42machines should let us experiment
01:03:44and fail instead of
01:03:46giving us everything on the table.
01:03:48We don't have to be afraid
01:03:50of being in the presence
01:03:52of smarter beings because
01:03:54we've all experienced it as
01:03:56children. Mom and Dad were
01:03:58just as young as we were.
01:04:00So even though, this is
01:04:02just the beginning,
01:04:04If we as a society actually
01:04:06take the risk of creating
01:04:08machines that are smarter
01:04:10than we are, we have to
01:04:12first figure out how to
01:04:14make sure they can understand
01:04:16our objectives and adapt them.
01:04:18Here you go, Levi.
01:04:20The great human quest
01:04:22is to create these technologies.
01:04:24I think we do it to understand
01:04:26of artificial intelligence is the opportunity to create a better future.
01:04:30I think if we plan it and we make an effort to do it,
01:04:33we can use that powerful technology to create a future that is more motivating
01:04:37than our predecessors could have imagined.
01:04:40Thank you for your attention.
01:04:42Artificial intelligence is a tool.
01:04:45Will it help me to be more successful professionally
01:04:48or help me to be a better artist or a better doctor?
01:04:52As a scientist, can I make a quicker discovery?
01:04:55Those are good conclusions.
01:04:57And that should be the right way for artificial intelligence to be good.
01:05:01I think from a social perspective,
01:05:03we're going to see an evolution in which it will be less utopian or dystopian.
01:05:08The debate is always framed as if utopia is the technology
01:05:12that will do all the work for us.
01:05:15And the dystopia is that there will be supreme lords over us
01:05:18and we will lose all human action.
01:05:21But what I tell people is that it's an election
01:05:24and that we should not fall into technological determinism.
01:05:28Technology is simply a tool to achieve goals.
01:05:33We need to decide how we want to use it to solve problems for us.
01:05:52Microsoft Mechanics
01:05:54www.microsoft.com
01:05:56www.microsoft.com