The case for empathetic AI | Pierre Robinet | TEDxNTU
URL: https://www.youtube.com/watch?v=fgm_NoOT7ok Video ID: fgm_NoOT7ok ============================================================ three years ago i was cleaning my room on a saturday afternoon when i heard the tv switch on in the living room with a very noisy cartoon it was quite unusual especially at 3 p.m in the afternoon considering i am the kind of dad who tried to limit screen's exposition for my young children so i moved to the living room and i saw my two boys two and three years old nicely sitted on the sofa watching paw patrol so i first turned back to my wife and asked her why she had turned on the tv but it was not her so i went back to the living room and i asked my son how i turned on the tv knowing he never did it before and to top it off he was watching a youtube video which represents quite a complex process and he replied i asked google papa clever but it was to my surprise that my son three years old was simply able to order to google ok google can you play paul patrol and get something he would never get from me especially at 3 p.m on the afternoon but also without being polite i'm french and you know how important it is to be polite for french people at that time i asked myself should i teach my kids to be polite to google does google need to learn some manners and ask for my permission some things started to be crystal clear google like many others was at my home as part of our lives in forgetting some fundamental family rules google's intelligence assisting and easing our life functionally entertaining our family may slowly influence our kids behaviors from the youngest age when they start getting their personalities shaped we think we create human-centric ai but we create machines that are not entirely serving on mankind and that generates a much larger impact on our lives than the one we expected and not the most positive one for instance social media platform we are using almost every day sometimes 10 hours a day generates a polarization of all society in influencing and exacerbating all beliefs and they are deliberately designed to make us addicted and such addiction generated by those algorithm has unfortunately been linked with feeling of unhappiness depression and mental health issues in compulsive users such platforms supposed to serve our humanity isolating us are finally not considering the social element the social power for human element and this provokes bad human behavior change and ai is even worsening exploiting our human weaknesses and by relying too much on technology sometimes things can go really wrong like this woman in canada who just drive drove into the lake stupidly and following a gps but the reason i'm actually showing you this news is such a similar story is happening to each of us almost every day let me explain you how such dependence and machine affect our health and or cognition this is where i live in singapore i am living on the 29e of this straight which naturally comes after the 29a 29bcnd further down the road before reaching my home most of the time i order a taxi the gps indicates to the taxi driver to stop at the level of the 29a instead of 29e and i can guarantee you taxi driver will always stop at the level of 29a and never look around if someone is waiting a few meters away i could see on your face it has already happened to you right and half of the time the taxi driver unfortunately missed my street certainly because the gps algorithm is not accurate enough but what worries me is this could happen two or three times before the driver even realized something is wrong have you ever realized a taxi driver can go around in circle in your neighborhood for 20 minutes without finding your street without realizing is lost but also without getting informed by the machine that something is not right such utter dependence on machine creates a definite problem in fact it is affecting all cognition and studies have proven that habitual use of gps could pose long-term health issues such as a higher risk of alzheimer as just described ai is embedded in our daily lives making multiple decision a longer human journey taking some of the most advanced example facial recognition is now used to pay at the cashier or open a gate health diagnostic are entirely proposed by machine intelligence and alone approval would be very soon entirely in the ends of an algorithm and these algorithms are walking days and night 24 hours today and we don't see them these shadow walkers govern a substantial part of our existence influence our behaviors potentially affect our cognition they often constitute a threat to our privacy and this without offering us the possibility of interacting with them with a note explaining their choice or their decision and without asking for our permission are we not missing here a two-ways communication between human and machine a kind of homogeneous interaction protocol to freely define ourselves in relation to it and act better accept them as part of our life what are we really missing human are working hard on ai but ai is not being designed to work hard enough for human look ai system focus essentially on servicing our human functional needs however even if they are trained to detect any addictive human behavior they are not trained to prevent them which can affect our relationship or health and have a broader impact on our society and its organization ai seeks to understand how our human cognition works by creating cognitive processes that emulate those of human beings and yes in doing so ai improve the speed accuracy and efficiency of how people think and work however it seems as well its prolonged used weakens or human cognitive abilities over time ai governed lots of our life's decision whether personally or professionally and is unfortunately very often not able to explain the reason of such decision and properly interact with us to help us better accept its choice and what it is offering current ai design is simply not taking enough into consideration what make us truly human and different from the machine or behaviors or cognition or emotion or social interaction look we rely every day on ai and ai is even not able to recognize when we are wearing a mask while we human continue to recognize ourself amongst each other and this is due to our unique cognition and social interaction abilities we have built these things so it is our responsibility to change it in reconsidering our human uniqueness succeeded in human machine cohabitation is what is the most complicated used to say dominic volton head of research at the french national center for scientific research the for such cohabitation to happen efficiently we need to give more space to human in the design of ai product and services and foster more human machine interaction this will help us to create real human centric ai how to proceed first we need to reassess our ai impact our lives behaviors and cognition whether it is on the short term or in the long term and we need to leverage the power of the data gathered by those ai products and services to do so why do we not have a gps based product or transportation platform frequently proposing to their drivers cognition test for instance the richness of this platform in terms of data may really help in detecting any kind of bad human behavior or addiction detecting social media addiction might also be very simple facebook capturing enough data should be able to send a signal when a very addictive behavior is detected and the strengths of its algorithm and machine learning already very much used to detect fake news may also be used to analyze how such addiction could be linked to any mental health issue as a support to our healthcare professional but this would require to be in the mindset of maybe opening a bit more this knowledge and data set to our physician and maybe slightly changed the business model of this platform highly based on our addiction second we need to redesign ai products and services in reconsidering our human and machine complementarity and a very good example is the mine nai process that has been proposed by the liquid news center for innovative cities at sutd here in singapore their team conducted some applied research on the design of a gps 2.0 envisioning what a future of the gps could look like and instead of a turn-by-turn instruction drivers might be given visual and special cues to help them navigate in the city like turn right at the red building or turn left after the petrol station at the level of the bridge thus protecting the drivers cognitive and spatial sense of the city why do we not ask also google to give us the option to train its algorithm to some of our essential social family roles and eventually allow google to recognize how we may be polite to each others last but not least we need to rescue ai from its lack of interaction with human for ai to be truly accepted it should not be designed as a trojan horse but offer more human interaction and if we want to make a step forward towards social acceptability of ai it is a responsibility to pull them out of the shadow allow us to see them and interact with them let's push any machine in the social interaction space to better communicate with human by interacting with the taxi driver in adjusting you know the car temperature to his mood or playing a song he likes or even asking him a question at the end of the day to know how is he feeling an algorithm would both work out its accessibility and eventually get access to really even data point related to the driver's behaviors and emotional status let's also enhance such interaction with the maximum of transparency i would personally value very much that an algorithm could give me the reason i have been rejected from alone such transparency would help remove the high reluctance we human have towards those shadow walkers and their ambiguity ai is everyone's concern and if we want machine to adapt to all human needs each of us would need to assist them in their adaptive process assess if they are doing well and delivering what we are considering is a human-centric outcome for each of us what if we could rate more how machine fulfill our human and social needs when we give a five stars to a taxi driver we are not always assessing how punctual he was but sometimes how nice to us he was right so let's do systematize such behaviors and guide algorithm in delivering a better and more personal human-centric outcome and reward them when they take care of us as human let's reimagine a world where we will assess more how human are the artificial intelligence let's re-imagine a world designed around our coexistence with machine let's re-imagine a world in reconsidering our human uniqueness and re-empowering any human in guiding machine to deliver the right outcome i am very positive towards such world and i do hope my kids will be very soon rewarded by google with the free perpetual episode but only when they have been polite thank you