There’s no bigger hype topic than artificial intelligence. If you have been following the news on one single day during the last four years, you will have seen at least one article on AI, Machine Learning or Deep Learning. These topics are often contextualized with the question: “Will you lose your job soon? And the general answer, especially for simpler professions, is often “yes”.
But a research result from the 1980s shows that it’ s not that simple. And in the end, exactly one group could have a guaranteed job – people with “simple” jobs.
68 years for a hairdresser appointment
Intelligent are the Facebook algorithms, our cars and the NSA. Microsoft is also intelligent, just like Amazon’s purchase suggestions and this computer that beat someone in “Go”. And what is intelligent if not Google! Their language assistant can now even arrange appointments at the hairdresser’s by phone.
What AI can and cannot do
But artificial intelligence is by no means a hip trend topic of recent years. Depending on what you read, the beginning of the research is either attributed to a six-week workshop of the Rockefeller Foundation in 1954 (ultra-mysterious) or to Alan Turing and his world-famous Turing test around 1950. More than 60 years of research until a computer can call the hairdresser on its own – Wow!
If you let the irony aside, you will actually find a multitude of achievements of artificial intelligence, which are then somewhat higher than the appointment by telephone: in 2017, for example, an algorithm discovered a planet in a solar system 2500 light years away. In the same year, Google developed an artificial intelligence that can program artificial intelligences. And of course there’s Bob and Alice, Facebook chatbots who first lied to each other and then invented their own language.
If one had to summarize the totality of all problems solved by an artificial intelligence, it could look something like this: ” Well, as a human being I would have needed a really long time for that.” And the reason for this is Moravec’s paradox.
Machine does what human cannot do
Moravec’s paradox was defined in the eighties by Hans P. Moravec (surprising, isn’ t it). A man from Austria who constructed his first robot at the age of 10 and submitted a kind of externally controlled cat robot with whiskers for his master’s thesis.
Essentially, the paradox can be described as such: Artificial intelligence can quickly learn and take over things in particular that humans find very difficult, such as recognizing abstract patterns or performing mathematical calculations.
At the same time, machines would find it very difficult to do things that a toddler would take for granted, such as recognizing another person and their intentions, moving freely in the room, or concentrating on interesting activities in the immediate vicinity.
As proof you can watch this video showing the running robot “Atlas” by Boston Dynamic: https://www.youtube.com/watch?v=vjSohj-Iclc
This is the non-plus ultra of current research on motoric skills. And whooosh: The hairdresser appointment looks impressive again.
I see something you don’t
As further evidence to the thesis that AI cannot learn things that are easy for humans, there is a nice categorization of the current progress of artificial intelligence on Wikipedia. While the categories “Optimal” and “Super-Human” are filled with games that require abstract thinking (Chess, Rubik’s Cube and obviously Go. Seriously, do you know anyone who plays Go?) the entries in the category “Sub-Human” read like the to-do list of a newborn: Object recognition, face recognition, voice recognition, running.
The explanation for this paradox is stunning and plausible: we massively underestimate our innate abilities. In fact, the skills mentioned above are not an easy task, but the fact that we have mastered them almost by birth is the result of thousands of years of evolution and optimization. Or as Moravec himself describes it:
We are all prodigious olympians in perceptual and motor areas, so good that we make the difficult look easy. Abstract thought, though, is a new trick, perhaps less than 100 thousand years old. We have not yet mastered it. It is not all that intrinsically difficult; it just seems so when we do it.Hans P. Moravec
“No, YOU will lose your job”
Knowing that it is especially the processes that seem simple and self-evident to humans that present great challenges to machines and their developers, it is all the more astonishing that the general consensus on the future of work is that particularly simple tasks could be replaced early on.
Thomas Erwin of KPMG for example, says in the in-house talk show klardenker live: “If I can describe exactly what I do to earn money, I have a good chance of losing my job within the next five to ten years. (from 33:00: https://videostream.kpmg.de/kpmg-klardenker-live-zukunft-der-arbeit) On the website willrobotstakemyjob.com (a grandiose piece of Internet), the profession of “cook” is threatened with 96% probability of extinction.
I think it’s much more likely that software will analyze all company and industry data so accurately that it can provide me with a better basis for making decisions than a management consultant, than a machine making Amatriciana sauce at my favorite Italian restaurant. Or that the robot from the Boston Dynamics video delivers the letters to us (although that would really look funny).
An unemployed world
Ultimately, I believe that most professions could sooner or later be carried out by machines. The only question is: who does it happen to first? And with Moravec’s paradox in mind, the answer is probably not “cook” or “gardener”, but “data analyst”, “marketing manager” or “management consultant”.
And then, of course, there is a final question: How bad is that actually – a world without work?