So we do know precisely what constitutes intelligence?
SHAPIRO: No. The assumption that you can reproduce human intelligence in artificial intelligence is totally wrong. The idea of transferring biological algorithms to mathematical algorithms shows just how simplistic we humans are.
Are you serious?
SHAPIRO: Of course. We have huge shortcomings. People are dying of starvation around the world, there are wars and dictatorships. Another example is climate change, which we are unable to get under control. We obviously can’t come up with solutions to serious man-made problems. On the contrary, we’re doing everything we can to wipe out our world and ultimately ourselves.
So we’re simply not viable. And that’s why we need AI?
STADLER: That alone is surely not the sole motivation, but AI will enable us to use our resources much more efficiently. But self-learning machines can undoubtedly also make our everyday lives much easier.
HANSON: There are things that we humans simply cannot do. Machines and robots can, for instance, lift much heavier loads. They’re more precise and much faster than we are. They don’t get tired, get sick or need a vacation.
Are humans becoming superfluous?
HANSON: No, because we ultimately also have some outstanding strengths. Not everything is based on logic. Humans tend to make personal decisions in particular based on gut instinct. This intuitive did-everything-right feeling is alien to a robot. It’s not about replacing us humans, but about extending our human potential. Assisted by robots, we can achieve more, become better and surpass ourselves.
STADLER: That applies equally to using AI in the working world. Robots are already working hand in hand with humans in our factories and are taking over strenuous tasks.
SHAPIRO: All of the technology that humans have developed thus far has been aimed at extending our possibilities. For many years, humans have had the benefit, for example, of medical implants that prolong life. In the future, though, this will go much further still. We will extend our cognitive skills, our intellectual capacity. We will, for instance, be able to assimilate things and knowledge, as and when required, with a pill. Such as a foreign language or a particular skill.
Does that mean in the future I’ll board a plane to Beijing, pop in a pill and be speaking fluent Mandarin when I land?
SHAPIRO: Believe it or not, you won’t need a dictionary anymore. You will become the dictionary yourself.
Sounds tempting. And when will all this be possible?
SHAPIRO: It’ll be some time yet. There are many obstacles to overcome before we can actually turn all science fiction technologies into reality. We still don’t have the right IT systems and what we’re still missing in AI research is an interdisciplinary approach. This is why I don’t believe that machines will be equal to us or even exceed our abilities in the foreseeable future.
So at the end of the day it’s all just fiction?
HANSON: It is in fact still very early days. AI is already much more reliable than we humans when it comes to analyzing medical images, or rapid stock exchange trading. But the breakthrough will come. And it isn’t an issue for our children’s children, it’s something we ourselves will experience.
Mr. Stadler, can you imagine, given these enormous prospects, soon working with an android as a colleague on the Board of Management?
STADLER: As a next step, I would prefer to have a woman on the Board of Management …
… but hypothetically speaking, an android certainly acts with less emotion than we humans, right?
STADLER: Absolutely. A touch of rationality wouldn’t hurt in certain decisions made by humans. It’s the interplay of the two that makes all the difference.
Mr. Shapiro is skeptical about crediting humans with intelligence, while Mr. Hanson is working on the robot of the future. What is Audi doing, Mr. Stadler?
STADLER: We’re looking at machine learning, among other things. This means a computer, in our case the car’s operating system, learns to act without being preprogrammed explicitly for a certain situation. Machine learning is essential for piloted driving and enables the car to act autonomously even in unforeseen situations. The car initially learns from specific situations, but can later generalize what it has learned. The more miles it clocks, the better it becomes. We are working hard on this issue and that is also why we went to one of the world’s most important conferences for experts in AI in 2016. We presented a model car that uses machine learning to develop intelligent parking strategies. In the next step we will transfer that to a real car. The goal is the intelligent car that can make independent decisions even in complex situations.
Isn’t the expertise for programming artificial intelligence actually in other sectors?
STADLER: You might not necessarily associate AI with a car manufacturer. But to push piloted driving forward, we need to assimilate AI as a core competence. Meantime, consortia such as the one between Audi, Mercedes and BMW with the HERE map service are becoming increasingly important as a way of pooling expertise.
Is that why Intel suddenly decided to join this alliance?
STADLER: Intel brings enormous expertise in developing and optimizing hardware and will support us decisively in our future projects. Together we want to develop a digital platform so that we can update high-resolution maps in real time.
Automobile literally means self-driving. Why has it taken the car industry over 130 years to discover automated driving?
STADLER: The dream of the self-driving car is as old as the dream of perpetual motion. Only so far, we haven’t had the technology to fulfill this dream. Now solutions are emerging to some of the problems that have for a long time been regarded as insurmountable. And so, enhanced computing power is finally allowing us to utilize the huge amounts of information and take the next step toward piloted driving.