Figure. Historical computer speeds. Note that the iPhone 11 Pro is a fast as the world's fastest super computer from 1996.
I have to admit I was recently impressed when I read an article in Forbes (Dec 31, 2019, p50). The impressive or more precisely stunning feature of the article was a graph that showed computer speeds. In this graph (shown above) the iPhone 11 Pro is as fast as the fastest super computer of 1996. Wow! I’m sure all of this will add to the hype surounding artificial intelligence and autonomous vehicles. To me the artificial intelligence (AI) and autonomous vehicle (AV) threat is just that: hype! I think these two technologies will add much to our lives, but will not put us all out of work.
In my opinion, what AIs and AVs lack to replace us is context. I have thought much about this in trying to combat the “irrational exhuberance” surrounding both of these technologies and have decided that lack of context is it.
What do I mean by context? Let me start by saying that AI experts such as Ray Kurzweil focus mostly on the computing power of the human brain and compare it to that of computers. He believes, to a first order that when computers have a computing power greater than a human, humans will begin to be eclipsed by AI. With this in mind, Kurzweil has predicted that by 2029 a robot will claim to be human and its claim will not be disputed. I'm not sure how Kurzweil measures the compuing power of the brain, but let's assume that he can.
I think we can agree that a computer/robot starts with no context. It can compute fast, but it knows nothing about the world. It doesn’t know that a dropped ball will fall, that Spring follows Winter, that the moon orbits the earth, or that a human needs to breath, eat, and drink. Certainly, these things can be programmed in or learned by machine learning; however, the amount of context that humans have is mind-boggling. This context helps us to solve myriads of everyday problems. As pointed out by Stephen Pinker, no AI can empty a dishwasher. An AI robot might be able to be programmed to empty a specific dishwasher, but to bring a robot AI into you house and say “empty the dishwasher,” and expect it to happen is science fiction. Not too impressive, as a well-disciplined child can easily handle this task.
This means that the hope of an AI robot that can completely clean a house is many decades away. How does this context situation affect AVs? Something like 90 to 95% of driving requires little context; for example, driving on a main road. However, the 5 to 10% of driving that requires context is broad reaching and requires much context. Consider this scenario:
You leave your garage, driving on gravel driveway that leads to a poorly marked country road on which your house is located. During the night, a moderate size branch has fallen onto the road. You look at the branch and decide that you can drive over it without incident. As you travel down the road, a UPS truck has pulled off to the side of the road, partially blocking it. You pull up behind the truck and see that it is broken down. The driver waves you on, but you have to look carefully down the road to assure no vehicle is coming in the other lane. You pass the truck and continue on your way only to be pulled over by the police. You roll down your window and are told that the bridge in town has been washed out by last night’s storm.
You follow the unmarked detour and finally make it to the main highway. After last night’s storm, the temperature dropped and the water on a bridge has frozen into a glare of ice. You slow down to drive over the ice only to find that in a valley up ahead the cold, still air has created a dense fog. Visibility is only 50 ft. You turn your fog lights on and negotiate the fog for ½ a mile. You can then travel the remaining 130 mile trip to the airport. The last 10 minutes of your trip to the airport parking lot has similar context challenges.
The scenario above has many challenges for an AV. We are decades away from solving all of them for a Level 5 completely autonomous AV.
So two things we shouldn’t worry about is one, massive job displacements by AIs; and two, truckers being out of work by AVs. These concerns are decades away, if ever. I should point out that many well-respected scholars disagree with me. Just recently. Daniel Susskind published a book, “A World without Work” in which he forecasts how the world should adjust to AIs doing everything. I would agree that there will be some job displacements, but as AIs become more common, we will be surprised at all of the things they can’t do, in addition to emptying a dishwasher.
In the meantime, all of the electronics required for AIs and AVs, as they progress, will still be assembled by us.
I predict a bright future.