Imagine how conflicting feelings can be experienced by a machine learning expert creating an artificial intelligence (AI) system that one day, and perhaps even very soon, will be able to independently create new AIs. And at the same time, these AIs will be more effective than those created initially by man himself. An era is coming in which machines will create their own replacement.
At the moment, machine learning specialists are highly valued in developed labor markets, however, when the world begins to create software that can "learn to teach" their own kind, the day is not far off when such specialists will become completely unclaimed.
In a similar situation, groups such as Google Brain, OpenAI, DeepMind, as well as departments of the most prestigious technology schools and institutes that develop machine learning systems may soon find themselves in order for these systems to create machine learning systems themselves in the future. And what is worse, the first signs of this can be noted already now. For example, researchers at Google Brain have developed a program capable of creating AI systems, whose task is to measure the level of performance of language processing programs. The test showed that a program written by a machine copes with this task better than software written by humans.
According to MIT Technology Review, the head of the Google Brain development team, Jeff Dean, sees “automated machine learning” as the most promising research project for his team.
“At the moment, when solving a problem, you rely on your own experience, the available data and the actual calculations themselves. Can we exclude 'experience' from this order when it comes to machine learning?”Dean asks.
If AI turns out to be able to consistently cope with tasks at levels comparable to those demonstrated in the Google Brain experiment, then one day, self-created and self-learning AIs could lead to faster creation and adaptation of new technologies.
And yet, while the field is still more of an interest only to enthusiasts, there is already a growing number of people around the world worried that the growth and development of AI systems could ultimately deprive many of their livelihoods.
Automation is designed to change not only the economy, but the very principle of capitalism as a whole, a principle that has not changed over the centuries. In the long term, machines will indeed become cheaper than hired workers. After all, the bosses no longer have to worry about when to give and pay their subordinates vacations, insurance, pay salaries and give many other things that are required and expected by employees from their employers. Yet this cheaper and more efficient workforce will require huge sacrifices from us.
Promotional video:
The largest sector of the economy that will be the first to experience the effects of automation will be manufacturing. Especially in developing countries. The importance of this problem was raised even in his farewell speech by former US President Barack Obama:
“The next wave of disorganization in our economy will not come from abroad. It will come from an unrelenting pace of automation that will make a lot of middle-class jobs simply irrelevant,”Obama said.
And many industry experts agree with these words. Moreover, it is not only low-skill jobs that will suffer. Systems are already being developed to replace, for example, film directors, songwriters, journalists and many others in the future. And now, when there are already AI systems that can create some programs that function much more efficiently than programs written by humans, we need to become more attentive to this problem and, finally, come to an understanding of what may lie ahead of the horizon.
NIKOLAY KHIZHNYAK
