“AI won’t end the world, but it might take your job” – WIRED
Artificial Intelligence (AI) is everywhere. At least that is the impression spread by media. A petition by leading AI scientists for banning autonomous weapons, discussions about self-driving cars, studies indicating that up to 40% of our jobs are at risk being replaced by intelligent systems, or prominent people like Elon Musk, Bill Gates, and Stephen Hawking warning about the risks for humanity, in 2015 AI appears to be everywhere. This does not only create a rather inflated image of the capabilities of intelligent systems, it also creates a very emotional, thus unreasonable debate. For someone outside of the AI or machine learning community, it is practically impossible to separate reasonable and realistic aspects from fear-mongering and Hollywood-like scenarios – frankly, even within the community, it is not always possible to do this.
Considering the impact intelligent systems will have on our economy and our society, it is time for more reason and less emotions in the discussion about how to deal with the emerging challenges. As debates about AI are often dominated by the fear that mankind will, in one way or the other, be enslaved or extinguished by intelligent machines, I want to debunk this, before describing the more pressing risks.
The unlikely Matrix-Scenario: why robots will not turn against humanity
An important insight first: robots will not become truly independent nor will they turn against us anywhere in the foreseeable future. While this is not an issue in serious debates, it is an aspect that is ubiquitous in the portrayal of AI. In movies, a super-genius like Tony Stark brings up the brilliant idea of creating a strong AI (human-like AI) and voilà there it is and turns against its creator. Modern science is too complex to allow such incredible technology leaps, so we can safely dismiss this scenario. Even the idea of a machine gaining a conscious mind is more science fiction than real science. There are no theoretical concepts for truly self-conscious machines, simply because we do not even understand how our own consciousness exactly works, or what it even is. The risk of machines turning against us, is also not an immediate problem: unless a machine is programmed to commit violence (or is not programmed to not to commit violence), it has no motivation to turn against anyone. The only risk, far-fetched as it is, is the Odyssey in Space – Scenario, in which an AI turns against humans, because they stand in the way of its objective. As unlikely as this may be, it is a scenario that at least has to be considered, when machines become more and more autonomous and their skills more versatile. Finally, remaining is the question, whether or not we want robot soldiers, drones, or similar, deciding about death or live of human beings. A question that has to be discussed by politicians, philosophers, and by a broader public. Considering the one-sidedness of this debate, I want to add one thought: while it is irrelevant for a person killed in combat, if he was killed by a human-controlled machine or an autonomous machine, it has to be taken into account that robots neither rape, nor loot, nor kill in rage or for pleasure. Unfortunately, this cannot be said about humans.
The likely scenario: how AI will change societies at their core
If AI will not kill us, why should we worry at all? Because intelligent systems will cause changes in our societies, unseen since the industrialization in the 19th century. Before the industrialization, roughly 90% of the population worked in agriculture. Two or three generations later this number was diminished to less than 10%.
Recent studies indicate that almost half of our current jobs will have been replaced by intelligent software or robots by 2050. Even though new types of jobs will be created and certain types – especially creative ones – will increase in importance, it is almost inevitable that the absolute number of jobs will decrease. Industrialization and later mass-production as well as automation destroyed innumerable blue-collar jobs, but also created many higher education jobs. This will not necessarily continue with the emergence of intelligent machines. Of course, so far untouched blue-collar jobs will become obsolete as well, for example self-driving cars will replace taxi or truck drivers. But, and here is the major difference to former industrial revolutions, AI will replace white-collar jobs. Stock brokers, engineers, and even programmers themselves, will at least partially be replaced by machines. This does not mean that their jobs become irrelevant or that these jobs will not exist anymore, but that many of their tasks will be automated, thus reducing the absolute numbers. All estimates indicate that intelligent systems will be the first technology that will destroy more jobs than it can create in long term.
What are the expectable consequences and how can we deal with them?
When, how, and to what extent these changes will hit our job markets is hard to predict, but it is all but certain that they will take place. This leaves us with two major questions: what are the implications and how do we deal with them?
On the positive side, the loss of jobs can help to mitigate the impact of the demographic development, many developed countries like Germany are experiencing. In times of a shrinking and aging work force, AI can help to maintain our economical level. Also, it should be noted that in the foreseeable future, AI will mostly supplant mechanistic, repetitive tasks, usually unloved by workers. Another positive effect could be a reduced workload and thus, reduced working hours.
Among the expectable negative effects, the foremost is reduced need of workers in industry and thus, the drastic shift in power between employers and employees. Today, work still supports the majority of the population. When intelligent programs and robots replace workers and create revenues without contributing to social welfare, social inequality will increase significantly. Currently, unemployment can usually be avoided by obtaining a higher level of education. In 2050, a university degree will probably be not enough anymore.
Unlike our ancestors in the 19th century, who were more or less overrun by the changes industrialization brought on them, we have at least a rough idea about what is going to happen. Furthermore, it is safe to assume that largescale automation will not happen within the next 10 years. This gives us the opportunity to further analyze the risks mentioned above and to find solutions that avoid massive social conflicts. But this has to happen soon, as necessary changes, especially in our education system and in social welfare, will neither be easy nor quick.
To achieve this, we need an unemotional debate, more cold-analysis and less fear-mongering.
Original: Der Text ist im FRANKly, dem Magazin des Fulbright Alumni e.V. , erschienen, der ihn uns freundlicherweise zur Verfügung gestellt hat.
Autor: Florian Grigoleit is a Ph.D. student in artificial intelligence at the Technische Universität München.