Bill Gates, Elon Musk, Steve Wozniak and many other high profile people have expressed concerns over the potential risks AI might pose. Why so? Well, it seems we are exceeding predictions over the speed at which AI is developing and possibly sleepwalking in to some bad outcomes.
Current AI can be described as weak or narrow AI mainly because of the limited use some AI algorithms are put to, such as facial recognition. Could a facial recognition algo fly a plane for instance? Obviously not, so the next milestone will be artificial general intelligence, AGI for short.
It is considered that AGI would outperform humans in most respects, even giving rise to recursive self-improvement. This could lead to a blossoming of intelligence leaving human intellect in the dark ages. However, would super intelligent AI exhibit human emotions? Why should it? Possibly to gain understanding of human emotions and maybe mastery over them?
Let us consider autonomous weapons. AI systems programmed to kill. I would not be surprised if they are available off the shelf now. It would not take much development to make drones autonomous except for refuelling and servicing. With GPS navigation and recognition engines, already known technologies, perhaps they are flying around as I am writing this!
There’s an obvious downside, what if they fell in to the wrong hands? Would there come a time when there was a balance of power between opposing countries in terms of autonomous weapons? Would autonomous weapons target the enemy’s arsenal of these things or the very humans that created them?
Possibly the first thing we humans will notice from the rise of the machines are our jobs going. People who do repetitive jobs may find themselves being replaced by a machine. This would be great news for a manufacturer wanting to become more competitive and productive but for how long? Well, as long as people have the money to buy the product. If half the labour force engaged in low skilled repetitive jobs become unemployed, there won’t be the money around.
The politicians will all say the labour force can be retrained, but to do what? Computer science is a high level discipline where proponents number the tens of thousands and not the millions employed in low skilled jobs. I love and appreciate good music but all the training in the world would not turn me in to a musician.
Tony Blair, a controversial British politician and a Marmite figure, maintains Brexit pales in to insignificance compared with the future impact of AI on our lives. He may be right on this one.