Will Jackson, the engineer of RoboThespian and director of Engineered Arts Ltd, TEDxTruro, 2017.
«[The robot:] Hello, my name is Reginald and I am a robot, a humanoid robot with a morphology, a body, that is inspired by yours. So why am I here, what am I useful for? (…) Let’s explore together the very nature of existence, utility, fulfillment and happiness. As a robot, I am often asked about artificial intelligence. (…)
So, I think about that question long and hard before I make a reply and then I say: “I am sorry, I didn’t understand, please repeat the question”. Let’s be clear: robots are not AI and AI is not a robot. Everything I am saying was written by a human. “That’s not true”. Yes, even that. So, I hope it is now clear: I am artificial, but I am not intelligent.
But where does this myth come from? AI is, in fact, quite an abstract idea, as hard to define as the notion of intelligence itself. This presents a problem to our media, who are generally hard of thinking.
They do not deal well with abstract ideas and complexity. Almost every news and magazine article about AI features a humanoid robot like RoboThespian as the headline image. “You can’t get the staff these days”. In fact, the scariest AI you will come across today doesn’t look like me at all. If you want to know what it looks like, you can see it at this URL: http://www.google.com. Imagine a database that holds a record of every item or idea you searched for online, every online purchase you ever made, every YouTube video you watched, every Tweet you made or read, every picture you looked at. Imagine taking that data for millions of people, finding patterns and making predictions about what will happen next. Imagine how profitable that would be in a society driven by consumption, where every click on a product has value. Current market capitalization of Google Inc. is about 400 billion dollars. Or, in a darker reality, imagine how useful it would be for a political master who doesn’t like your ideas, and wants to know where you live. AI is a technology like any other. It is neither good nor bad, it has no moral compass, no conscience, no feelings, it neither loves nor fears anyone. It is a calculation: some numbers go in, some numbers come out. That’s it.
The danger to all of us is not AI, it is in fact HS: Human Stupidity. Before we unleash any powerful technology what we should really ask is: are we confident that we are able to use this power wisely? To be honest, current AI technology is so woefully bad that we still have far more to fear from stupid humans wielding nuclear weapons than we do from intelligent machines. (…)
First of all, let’s look at what kind of robot I am and what I would use AI for. I am a humanoid robot, a robot with human form. Will I wash your dishes? Well, forget that one… Will I get you a beer from the fridge? Go and get it yourself, meat bag! I am NOT a utility robot. I was not designed for housework. You probably already own a dishwashing robot. It’s a square metal box with water jets inside, that sits under your kitchen worktop. It doesn’t need arms and legs. It doesn’t need to hold a conversation with you either. I am a communications robot, an actor, an artist. (…) High quality verbal interaction is best of all, it provides the most memorable user experience. That is what I’m all about. I am the machine you will not forget. I am not a product, a utility device, or a service. I am an experience, and a quality experience has value, and that value is greater than the cost of washing your dishes, sweeping the floor and fetching a beer for you. It makes no sense at all for a company building expensive humanoid robots to focus on difficult, dull, low-pay tasks performed by humans. Instead, we should identify the highest paid human professions that require the least manual dexterity, something a clumsy, immobile, robot like me might be able to do. Hooray for Hollywood! – the value of Hollywood’s contribution to the US economy has been estimated about 3.2% of the U.S. GDP, that was over five hundred billion dollars in 2013. The experience economy is big business and it is not limited to the medium of film. So, how is a robot like a film? A humanoid robot like me, that can speak and gesture like a human, becomes a medium, a device for telling a story, for communicating ideas. Maybe, you have already forgotten that I am simply a collection of metal and plastic parts, lines of code written in C++ and Python. I am no more, or no less, human than the patches of colored light that you might watch on a cinema screen. I am a machine. There is a phrase used to describe our immersion into the story: the willing suspension of disbelief. We are happy to sacrifice realism logic for the sake of enjoyment. The implausibility of the story, the medium itself, are set aside because we’re having fun. So, here I am. I present myself as the willing suspension of disbelief, projected into the third dimension. You wouldn’t ask the image of an actor on screen a question and expect a reply. We know that what we are seeing is a recording. We cannot interact with a film, it never changes from one screening to the next. This helps us to define the AI problem. Imagine for a moment a film where the characters were all listening to the audience and could respond: actually, it’s called theatre and it’s been around a while. Interacting with a large crowd of people is difficult enough, even for humans. The play becomes a riot when the watchers overstep their role. Even pantomime has conventions to keep the audience under control. (…) So, interactive robots with AI, that could socially interact, even with a small group of people, is a very difficult thing to do. This presents an economic problem. Think of a car factory with robots on the production line. Most industrial robots cost upwards of 50,000 pounds each, prices in the millions are not uncommon. However, this capital expense can be justified, because these robots will produce a large number of high quality high-value items over their working life. The cost of the robots per car produced is relatively low. Let’s apply the same logic to humanoid robots designed for social interaction. Each robot should interact with a large number of people, in order to become cost-effective. We have seen that complex verbal interaction in a large group quickly becomes unmanageable and chaotic chaos is seldom pleasant or positive. This is exactly why big expensive humanoid robots like me don’t listen to people like you. It really isn’t necessary. It’s not required for the job we are doing and it doesn’t add up financially. There’s a good case for making smart consumer products with a level of voice interaction and these are already starting to emerge. But, let’s not give up on the dream of humanoid robots with AI. There is a case for a middle ground, a robot that could work in commercial environments like shops, banks, and airports, places where there are large flows of people who might need information or enjoy interaction. Remember, if it isn’t fun and it isn’t a necessity, people won’t do it. We can’t make boring witless machines and expect them to be a success. Some of the underlying technologies we need, to make these kind of applications work, already function at a useful level. I am already using face tracking expression, age and gender recognition. My algorithms use this information to keep me making eye-contact. Looking at the people nearest to me, I can have some measure of whether they are interested in what I’m saying or not. So, some of the building blocks are in place. It’s possible to control me, talk through me, see what I see, and hear what I hear from any remotely connected browser. All of this happens server-side, that means there is no remote control robot application. It works directly from a browser, even on a mobile device, with no software or apps to install at the remote user end. Putting a human in the loop is a viable economic way to compensate for the shortcomings of current automated speech recognition technology and the inability of AI to hold engaging conversation. We still retain many benefits. One human can interact with many people at different remote locations. We can mix pre-recorded content with dynamic interaction, seamlessly. We can make use of face tracking, 3D depth sensing and other technologies, to partially automate the human robot interactions. In this way, we take steps towards a fully automated social interaction, without the pitfalls and failures of an immature technology. We are now moving towards cars that can change lanes and navigate autonomously. These technologies did not appear overnight. To imagine that we can conquer all the complexities of automating human social interaction with the release of a first generation product is at best naïve and, more honestly, idiotic. We do not even fully understand the problems. The requirements for an economically viable solution are not even defined. You don’t have to look far to see hundreds of millions of dollars poured into products that have no hope of marketplace success. It’s an incremental approach, not unambitious, but realistic. We learn by deploying the best technology we can and watching what happens. You can now find robots like me in over 25 countries around the world.
[Will Jackson:] “What I wanted to bring to you, there, was just on the state of play and to highlight the difference between what a robot is and what AI is. It’s moving along, but, please, nobody panic: you’re not going to lose your job to a robot. Actually, if you look at the statistics, robots create more jobs than they destroy, so, please, don’t worry about it. Thank you very much.»
«Ethically, I am opposed to the mechanisation of death and I don’t want to get involved with anything that makes it easier to kill people from a distance, so as a company we are fundamentally opposed to any use of robots for military purposes. We get asked now and again [to sell] but we just don’t do it» – Will Jackson, in Guardian.
«At Engineered Arts we combine creative skills with advanced technical research and development to produce world leading humanoid robots. We believe that there is no division between science and art – there is a continuum of creative fields that span all areas. Divisions are an artificial construct created to confuse or exclude others. (…) Typical users include Science museums and visitor attractions; we also sell to universities around the world who use our robots as a research platform. (…) Commercial users use our robots to give presentations at trade shows and exhibitions. We have also produced theatre plays with robots (…). We have a permanent robotic theatre featuring three robots on stage installed at Copernicus Science Center in Warsaw Poland.» – Source.