Watson, Einstein, Leonardo. These names were chosen as product names for Artificial Intelligence (AI)-based technology. These names are not random choices – they are names associated with brilliant and creative minds, reflecting the aspirations these products have. These exceptional figures are extremely individual and have had such a profound impact on humanity that we can count ourselves lucky to experience life improved by their existence.
Will we see this promise of general AI, mimicking incredible minds of human history, fulfilled over the next years or decades? Probably not.
However, recent advances in applying machine-learning to real-world problems has demonstrated that super-human performance is possible and available today in the form of dedicated and specialized AI systems. Programs designed to play chess and Go have beaten the greatest players of the games. And grocery replenishment or setting online prices by machine produces higher margins and less waste (full disclaimer – I work for Blue Yonder and we offer AI-based pricing and supply chain solutions for retailers).
Chess grandmaster Garry Kasparov recently published his book “Deep Thinking” in which he writes about his match with IBM’s Deep Blue, the super-computer that beat him at his own game in 1997. In the early days, the creators of such programs set out to mimic human intelligence. They ended up producing super-human performance by using the computers for what they can do best: computing and crunching numbers. This book review gives a great overview of this aspect. I suggest reading the full book if you are eager to learn from 20th century history.
My prediction for at least the next decades is that we do not need to worry about the “AI singularity” where machines become sentient, start thinking for us and decide humanity’s fate. Instead, we will see more of the recent successes where specialized AI systems exceed human performance in in a specific but narrowly defined case. In a recent post, experts predicted when AI-based systems will exceed human performance in various fields. For example, AI-based systems are likely to be better than the average retail salesperson in the next five to 30 years.
Machines with general intelligence won’t show up on the short- to mid-term. Even if the term AI will get into discredit as it does not deliver on expectations, the idea of what AI is may well change causing a shift in what the term means. In any case, solutions will be built and deployed to take over specific tasks from people because they demonstrate super-human performance.
During the industrial revolution, automated looms were created to take over manual labor from people to boost productivity. The looms and other machines were able to weave more, lift more, bend more than any individual or group of humans. Today, using computers and applying machine-learning to practical problems, we build machines that boost productivity. These solutions will help us bypass activities where we are limited (computing, statistical reasoning) and put our priorities on tasks where we excel (creativity, self-reflection). The impact on us as individuals and society will be profound – even without general AI.