According to Stephen Hawking, AI (artificial intelligence) could lead to the end of the human race. As AI systems have been outperforming human-beings in various domains, should we start to worry whether humanity is on the way out?
When in 2011, IBM’s Watson achieved a victory in the competition ‘Jeopardy’, it might have been not a big surprise as a computer like Watson could easily access an encyclopedia. Yet, for those not involved in the field of AI, what may be difficult to interpret with regard to such a big achievement is that to understand the clue well enough to make a correct guess requires more than an encyclopedic amount of knowledge. Although many AI researchers may have assumed that making such a guess would extend the capabilities of AI systems, they were wrong.
On the other hand, this should not be interpreted as Watson or any other computer being capable of achieving a human-level understanding of the world. It should also be taken into account that Watson assumed that Toronto was a city in US rather than in Canada. Such a simple mistake may leave AI researchers scratching our heads when it comes to AI systems with superhuman levels of performance.
In a similar vein, researchers from Carnegie Mellon were able to fool a facial recognition system into thinking that a man who was wearing eyeglasses was actress Milla Jovovich.
Given such cases, it is obvious that AI systems are involved in a different problem-solving process in comparison to human-beings. Although AI systems can acquire some statistical patterns which can be used to a great extent, they are lacking a common sense which cannot be conveyed to them in any technical way. In addition to this, in the case of a change in which the way data is produced, misleading results may occur as well, such as the case of the man wearing the eyeglass frames and being mistaken for an actress.
Within the light of this information, we can gain insight into what type of jobs may be more appropriate for AI systems to perform. While tasks that require responding to the same kind of standardized input over and over, such as the diagnosis of medical images, may be more relevant to AI systems, unpredictable tasks for which an understanding of individuals and the broader world is required given the messy nature of the task, AI systems may not be the right choice.
Given the progress of AI in a short period of time, future breakthroughs are easily thinkable. So far, human-beings could not be outperformed in terms of developing a robust and flexible understanding of the world. If AI systems could acquire such an understanding, one of the most complicated issues in the field of philosophy, namely the nature of consciousness, will become very pertinent. Yet, in the near future, there seems to be no clear path toward developing such systems. Despite the fact that today’s AI systems have been proven to cause many disruptions in various domains, they still do not make humanity obsolete.