Transcript

Human-AI Collaboration: a Robotic State of Mind | Siri Beerends | TEDxUniversityofGroningen

hey siri are you a machine for anyone who expected a talk from the digital voice assistant i'm sorry you will have to do with the analog version i am not a robot at least not yet i say not yet because in our collaboration with artificial intelligence we become more like robots to understand how and why we become more like robots we first have to go back in time this is an illustration of the mechanical turk it is a chess machine that was built 250 years ago and he could beat the best chess players in the world the inventor was very smart he designed the machine in such a way that you could hide a human chess player inside so the public thought that the machine did all the work by itself i showed you this because now 250 years later we have machines that can really play chess and beat the best players in the world we call this artificial intelligence in short a i but we also call this progress because first we had to hide a human chess player inside a wooden closet and now we have machines that can really play chess but as i will demonstrate it's not really progress artificial intelligence is not necessarily about progress because in our collaboration with artificial intelligence we develop robots that have to behave like humans and humans that have to behave like robots but first i might have to clarify what is artificial intelligence ai is a problematic concept i will tell you more about that later but in general and in the public debate what we mean by ai is computer systems that can independently perform tasks and improve their own performance by learning from data this is very abstract so let's have a look at how ai is implemented in your daily lives you might have noticed that when you searched for a particular product on the internet the product keeps popping up in your social media timeline and in your advertisement or you might have noticed when you type a sentence in your smartphone the next word is predicted this is the work of ai but also governments use ai for example they try to predict all kinds of risks on the basis of big data and ai risks of criminal behavior or burglaries so these are some practical examples there are many popular ted talks about artificial intelligence and most of these talks are about how computers are getting smarter how we can outsource more and more tasks to computers and how this will give us more time to do meaningful things because when computers do more tasks we finally have the time to take care of our loved ones to do some volunteering go to ted talks solve world problems we can do anything when we have more time but i try to prepare an uplifting story about this idea the idea behind artificial intelligence the big promise is that humans will flourish and robots will work but unfortunately i have bad news because in fact it is exactly the other way around what happens is that robots flourish while humans work humans work to train ai systems and humans work to satisfy ai systems and to demonstrate this to show you how robots flourish while humans work i brought two three examples the first example is jira jira has a work from home job he works at a click work platform he trains ai systems in the past 50 days he has been busy with selecting drain pits in sidewalks but he's excited he will be upgraded the next 50 days to selecting helipads on rooftops so now i might hear you think well seems like a pretty boring job but it's not really my problem but actually we are all jira i think that we all recognize this this is a recap chat code you have to select the right images to prove i am not a robot but this is very ironic because when you are doing this you are actually working for a robot you are for example when you are selected it's usually not select all tech billionaires it's like select all cars or select all traffic lights and when you do this you are training an ai system to become better at recognizing objects so you are not the user of artificial intelligence you are the producer of artificial intelligence you are the beta tester the debugger you are the guy you might recognize him again you are the guy who's stuck in here to make sure the machine keeps functioning because we think that our ai systems get smart auto magically but this is not the case our ai systems are powered by an enormous army of human click workers and these click workers are called ghost workers because we don't know that these people are out there and we don't even know that these jobs are out there so ai is not liberating people from repetitive work it is just making it more invisible my second example this is nina and she has a job in a call center and what she really likes about her job is that she can put in her own personalized style of helping people but recently she lost pleasure in her job i'm sorry because she recently found out that an ai system i'm sorry gives all her phone calls an empathy score and her empathy score goes up when she uses the word sorry so now she tries to cram many stories into the conversations to keep her empathy score high and make sure she won't get fired i'm sorry the ai empathy score is not a fiction it is really used in call centers but not only there there are many work settings where employees feel that they have to become more robotic in order to satisfy an ai system workers in hotels warehouses but also in delivery jobs are monitored by algorithms and ai systems that tell them how they must do their job and how quickly so we were promised these super smart ai systems that would improve or replace our work but what many people got instead are ai systems that became their supervisors telling then how they must do their job and how quickly my third example is jimmy and jimmy is nina's little brother and he interacts a lot with grubby the home speaker and nina was a bit shocked recently play peppa pig song now nina was a bit shocked not to find out that grabby collects data and sells it to third parties but she was shocked because she overheard jimmy talking to his friends in the same commanding robotic way as how he talks to the home speaker and this is also not a fiction uh researchers found that we adapt our communication to our digital devices so chatbots digital voice assistants smart smart home speakers they all have an influence on how we communicate with each other also outside ai contexts so these three examples demonstrate that in our collaboration with a i we adapt our behavior to a i instead of the other way around and that is not so strange because we are the most flexible creature of the two it is more likely that we will adapt to the binary logic of the computer instead of the other way around and what these three examples also demonstrate is that artificial intelligence is not just a technology and it is also not as is often suggested just a tool you can use for the good or the bad no ai changes what good and bad means ai changes our moral ideas of what good and bad means and it changes our habits so ai is not just a technology it is an ideology it is an ideology because it is based on the economic principle of commercial data collection collecting as much as data as possible and self-derived data to third parties and this also means that companies and governments try to capture all human processes in data and this is problematic because not everything that matters to us can be properly translated into data not everything can be captured in data so what happens is we now just tend to forget about all aspects that are not captured in data so that's really tragic and another reason why ai is an ideology is because it is based on a very narrow mechanistic understanding of what intelligence means ai redefines intelligence in terms of what computers are capable of so pattern recognition computation power and information processing but intelligence is of course way more than those three things intelligence is also about consciousness empathy willpower common sense intentionality and very important it's also about bodily intelligence our embodied brains are very energy efficient when we have to learn new things we do this in a very efficient way contrary to ai training a single ai model can emit as much fuel as five cars in their lifetimes so artificial intelligence is very energy inefficient and has an enormous climate impact so when you hear this i would say we should not understand ai as a form of intelligence we should understand it more as a form of advanced statistics maybe statistics on steroids would be a more appropriate term because of course machines outsmart us in playing chess and go or any other task in a stable environment with clear rules but our society is the opposite of stable with clear rules when ai systems are unleashed outside of their small training environment into the messy world into the messy real world they often completely mess up because they cannot adapt to context unless we change our society into a predictable board game and we become robots ourselves then ai systems can flourish everywhere but is that the kind of society that we want do we want to make life as predictable as possible do we want to become optimized for the data market do we want to quantify monitor and measure all aspects of our lives do we want to become these clicking swiping data producing machines that train or feed ai systems 24 7 i would say no we don't want that okay but if we don't want that then what should we do instead of expecting more from a i and less from each other we should expect more from each other and less from a.i because technological progress is not the same as societal progress and we've seen this especially during the covit pandemic so of course we can develop all kinds of ai systems to help solving problems ai systems to diagnose covit ai systems to detect social anxiety but this is all symptom control we need to do something about the causes of these problems we need to do something about the flaws in our social economic systems that trigger these problems in the first place and keeps triggering these problems again and again and again artificial intelligence is in fact very conservative it is based it is trained with data from the past and it just reproduces pattern patterns from the past so that's not real innovation if you want real innovation you don't need better computers you need better social economic systems and unlike ai we have the potential to reshape those social economic systems because we can act on the basis of our own moral intentions environmental awareness and long-term perspectives so let's not waste that potential by becoming robots ourselves but let's address that potential and develop it further together thank you [Applause]