The is Social Intelligence, which A.I is generally quite
The rise of A.I and computerisation has many people worried that their jobs might be replaced by automation and computerisation. In “THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION?” by Carl Frey and Michael A. Osborne, Frey and Osborne look at the susceptibility of jobs to computerisation. A study was conducted, where 702 detailed occupations were examined and the probability of computerisation was calculated, with the use of a Gaussian process classifier. They calculate the amount of jobs which are at risk, but ignore the fact that new technology could create more jobs. (Frey & Osborne, 2013)
“According to our estimates, about 47 percent of total US employment is at risk.” (Frey & Osborne, 2013)
As seen in the figure 1, there are 3 main bottleneck variables, which limit computerisation.
FIGURE 1. A sketch of how the probability of computerisation varies over bottleneck variables (Frey & Osborne, 2013)
The first factor is Social Intelligence, which A.I is generally quite bad at, because of human’s unpredictable nature. Every person is different and while a regular person with an average SQ can adapt to each person, A.I is still struggling with recognizing different emotional states or body language and how to respond. Today’s A.I might not have good social intelligence, but in the near future A.I with good social intelligence is almost a certainty. A Stanford University study already created A.I, which can accurately predict human trajectory in crowded spaces, just like how people can read body language and not bump into each other. (Alahi et all., 2016)
The second bottleneck is creativity. Human creativity is one of the basic qualities a human can possess and it is very hard to replicate with A.I, but not impossible. A study from the University of Sussex from 1998 stated that we are a very long time away from achieving creativity in A.I. (Boden, 1998), but in 2016 A.I Watson from IBM analysed the movie “Morgan” and created a movie trailer. This was achieved through deep learning and analysing other works and was not truly creative. Jason Troy, CEO of Somatic said that true creativity trough A.I is not going to be achieved for a while. (IBM, 2016)
The last bottleneck is perception and manipulation. The human body is filled with millions of sensors, which can detect touch, pressure and heat. We know how to interpret these signs, but teaching an A.I to do this is much harder, since there is no big data on how people react to smelling certain things, hearing something, or picking up an object.
These bottlenecks will disappear over time as A.I. improves, but even with them present, 47% of all jobs in the U.S.A. risk automation or computerisation within the next two decades (Frey & Osborne, 2013). The ten jobs with the largest employment (FIGURE 2.), (Except registered nurses, who have a 0.9% chance of having their job computerised) have on average a risk of having their jobs computerised of 88%. All of these jobs (except for registered nurses) have a lower than average mean wage than the rest of the U.S.A. (FIGURE 3)
FIGURE 2. Employment for the largest occupations in the U.S.A
FIGURE 3. Annual wages for the largest occupations in the U.S.A
(The Bureau of Labor Statistics, 2013)
A further look into the data reveals that on average the jobs with a high chance of computerisation also tend to be jobs with a lower than average median wage. Jobs which do not require a higher level of education are also prone to be computerised. This extends the problems of rapid computerisations. Mostly the unskilled lower-class workers will lose their job. This will worsen economic inequality in a country where economic inequality is drastically bigger than in other developed countries (FIGURE 5). The inequality has been rising and with the loss of low-wage jobs, this gap will only widen further. (Frey & Osborne, 2013)