Rise of the Machines
Most people who watch sitcoms would inevitably have come across The Simpsons, the quintessential story of a working class American man and society. Homer Simpson, the father of The Simpsons, drives to work in a nearby power plant, where he is a nuclear safety inspector. Most of his job involves checking key operational parameters for anomalies. What work would a man like Homer do if he was born 150 years ago? And what if he was born 50 years later? Let’s make some guesses –
1850 — Homer, the average man, would probably work in a coal mine. He would walk to work, dig coal (or if he was lucky, supervise the men who were), get paid in cash, and presumably went outdoors or played board games for fun. He eats food he brings from home, or the food served there by mess workers. The broom at home helps clean the dust.
1925 — The standard American job early in this century was factory work. By now, he may have had some form of transport to get there. Leisure evolve to introduce modern games like basketball and baseball. Books and amusement parks were on the rise too. He still usually eats food he brings from home, or the food served there by mess workers, but takes packaged food sometimes as well. The mighty broom continues to rule.
2000 — Today, Homer goes to work in a car. He sits in his office, observing numbers that the machines output. He gets paid electronically to his bank account, and has his TV to accompany him at home. He also has options like video games and social media. When no one has cooked food at home, he just puts a ready packet into the microwave and eats that. Now that his house has gotten bigger, he uses a vacuum cleaner since it’s cheaper than getting someone to do it for him.
2050 — Homer takes a self-driving car to his workplace, where all the machines look after themselves internally. For entertainment, there are a myriad of devices. There is a machine that prepares any food one wants. The house is self-cleaning. Wait, why is he going to work again?
Although purely hypothetical, this does give us an idea of what way our jobs are progressing. He goes from more labour-intensive work to less labour-intensive work. From mining coal to using machines in factories, they are permeating more and more into our lives. Where we used to be dependent on other human beings, we’re not depending on technology. Be it social media replacing the mailman, or Google Maps replacing the person we ask on the way, technology seems to be doing a significant amount of work for us. But can they do all our work, leaving us with nothing to do?
Several economists say no — humans will always have work to do. They believe that the theory of comparative advantage, advocated by Ricardo in the 19th century, holds true in all cases. Even if machines can do better at everything we do, there is still a case for us to do what we are best at. This is essentially what comparative advantage is — if country A produces more wheat and more steel than country B, it still makes sense for them to trade. Though it seems counter-intuitive, it is widely accepted and is the entire basis for international trade. David H Autor, an economist says that while machines are getting better at what they do, the value of the work we do complements it. Quoting him, ‘Typically, these inputs each play essential roles; that is, improvements in one do not obviate the need for the other. If so, productivity improvements in one set of tasks almost necessarily increase the economic value of the remaining tasks’. Since we’re sticking on to only the theoretical question of whether all human work can be replaced, we can ignore social considerations for now. But I still think Autor’s assumption of comparative advantage fails when we bring machines and technology into the picture.
Let’s assume a two member society — a human and a robot. Suppose the human can lay 1000 bricks a day or cook 20 kg of pasta a day (for the sake of the argument). A robot can lay 10000 bricks or cook 100 kg of pasta. Clearly, the robot has an absolute advantage, whichever task it does. The theory of comparative advantage tells us that it still makes sense for each to specialise and then trade. (If you’re not able to wrap your head around this, here’s a great link). But what if another robot could be produced at a moment’s notice? Then, the human’s output would be much less than both, and it makes no sense for the human to do either task anymore! Clearly, the ability of robots (and technology in general) to be replicated at a short notice and low cost is the key to my belief as to why comparative advantage may not hold.
With the advent of the motor vehicle, there came about jobs at petrol bunks, mechanics, windshield cleaners and so on. These are the unexpected jobs that came up with the development of the motor vehicle. Whether it recreated all the jobs that it took away is a side issue; the fact is that machines couldn’t do all the work. Autor argues along these lines — there will be jobs created that we cannot envisage with the advancement of technology. These are the complementarities that progress offers. He invokes the idea that as one activity becomes more productive, all the other activities in the process bear more value. Thus, human beings will always have a role to play by filling these gaps. But with the pace at which technology is moving, the complementarities they require are also becoming more technical. Machines now require more machines to work better. However, humans are still required to build those ‘other’ machines. This point leaves me deeply conflicted on whether robots can ever fully replace us. For every 1000 Homers who have nothing to do, there is 1 person building that self-driving car. However, artificial intelligence might change this.
The Advent of Artificial Intelligence
Before that, let’s step back and consider what intelligence is. How do you know what a notebook is? Some are big, others small. Some spiral-bound, some hardbound. But when we see a notebook, we just know it’s a notebook. But imagine describing a notebook to someone who has no intelligence. Notebooks come in varieties of shape, size, cover and colour. It takes intelligence to know what a notebook is. This is Polanyi’s Paradox — We know more than we can tell. This is traditionally seen as the reason why computers can’t replace human beings. However, this is changing quickly. Some evidence of this –
· AlphaGo, a program developed by Google DeepMind, beat Lee Sedol, the best player of Go in the world in March 2016. Go is an abstract strategy board game considered more difficult than chess, wand requires strong intuition and intelligence.
· Every time you use Image Search on Google, you’re actually using technology from Google DeepMind, and is today at a 5.5% error rate. If that sounds like a lot, consider this — human error rate is roughly the same.
· One standard for artificial intelligence has been the Turing Test. Alan Turing, a famous British mathematician (more famously, The Imitation Game was based on him) proposed that if a machine could successfully fool a human counterpart during a conversation that it was human, the machine would have shown true artificial intelligence. This was over 75 years ago. There are several systems that have come close and even beaten it (though this is disputed). Either way, we’re coming very close.
Why is this becoming a bigger talking point by the day? The amount of data that we’re generating every day is massive; the world’s data is doubling every two years. It’s predicted that by 2025, human brains will have the same power as the human brain. With the advancement in computing ability , the large amount of data there is at the disposal for machine learning algorithms is massive, which can be used to build better AI.
In lay man terms, with the amount of data and computing power that we have, we can use brute force — show a computer a million pictures of a rock, and it’ll be able to identify the next thousand correctly — to develop intelligence in computers. This means that Polanyi’s paradox will be circumvented. Computers will now know just as much as us. And without any of the human drawbacks, like the need for rest, food or familiarity — it isn’t hard to envisage these robots taking away all the work we presently do. And when Elon Musk says we should be worried about AI, we probably should sit up and take notice.
Bottom line, it seems likely that all the work we currently do, can be taken over by robots in the future. Whether that necessarily means doomsday is near is more of a social question — it depends on how quickly human society adapts to the changes, while taking everyone on board. Because one thing is clear — the transition will be difficult. Jobs created in the 70’s and 80’s are already being taken away by artificial intelligence. Human civilization has never been in such a rapid state of flux. In Saturn’s Children by Charles Stross, once human society has robots doing all the work, we retreat into ‘internal pleasures’ and lose interest in the outside world, eventually leading to our extinction. Let’s hope we can prove him wrong.
If you want to explore this topic more, here are some great links -
PDF: We made a fancy PDF of this post for printing and offline viewing. Buy it here. (Or see a preview.) Note: The…waitbutwhy.com