AI will take over many jobs but there is room for humans

Tomohiro Inoue, an associate professor of economics at Komazawa University, graduated from the Faculty of Environment and Information Studies at Keio University before studying at the Waseda University Graduate School of Economics. He specialises in macroeconomics. His written works include “Jinko Chino to Keizai no Mirai” (Artificial intelligence and the future of the economy).
Inoue speaks about the future of jobs in the age of artificial intelligence during an interview with The Yomiuri Shimbun. 

To what extent will AI be able to do human jobs?
Groundbreaking new technologies are coming out, and artificial intelligence is becoming quite practical. AI image recognition technology, which determines what is shown in an image, is becoming especially widespread.
Take this example of how such technology is used in London. AI constantly analyses security camera footage. When it determines that someone has left a bag on the street and walked away, it notifies police, who rush to the scene within minutes to dispose of the bag. This is because the bag could have a bomb in it.
A human being will immediately notice if someone has walked away from a bag. Now AI can also recognise actions that have been taken, just like a person.
Image recognition technology has enabled digital cameras to automatically focus on people’s faces, and is already part of our daily lives. But I do not think we have realised even one-thousandth of its underlying potential.
Leading AI expert and University of Tokyo Prof Yutaka Matsuo often says that the implementation of image recognition technology means that machines now have eyes. This includes the ability to discern what they see with their “eyeballs”.
Video cameras and sensors have long existed, but they couldn’t discern exactly what they were seeing. Now that machines have eyes, they will likely evolve tremendously just like Cambrian Period creatures. Even now we cannot imagine our lives without washing machines and vacuum cleaners, but this is something different. Machines are going to play active roles in every aspect of our surroundings.
Robots are already being used in factories, but they only do regular jobs in fixed places. When machines have eyes, we will be able to use robots even in agriculture and the service industry, in situations requiring case-by-case judgement.
General purpose AI now exists, but it is still not as effective as human beings. A human can play Go or shogi, and can also do office work and converse with people.
There are quite a few groups in the world like DeepMind developing “general-purpose AI” with broad intelligence similar to that of humans. The Czech company GoodAI and the Japanese nonprofit Whole Brain Architecture Initiative aim to complete general-purpose AI by around 2030. We cannot know the future, but my gut tells me that by then we will have general-purpose AI that can clumsily imitate various human behaviours. I imagine it would be difficult for it to surpass human functionality in all areas.

What issues will AI face in the future?
Beyond image recognition, linguistic comprehension is a major obstacle. We now know that AIs must connect their understanding of language to perceptions and insights.
For example, take a cat. As a narrower concept of a mammal, a cat has eyes and ears on its face, whiskers and a tail protruding behind. Thinking about how to codify and describe this is 20th century AI research.
But a human understands the meaning of a cat through all five senses, including visual information in the form of its appearance, auditory information in the form of its meows, and sensory information in the form of how it feels when touched. Twenty-first century AI will surely learn to understand language in this way.
If image recognition and other technologies improve rapidly through deep learning, computers may be able to understand the true meanings of words. Even so, sayings and expressions can be difficult to understand. For example, in Japanese we have the expression “want help even from a cat”, which means to be so busy you are desperate for help. It includes the implication that cats are not useful creatures. Even the seemingly simple concept of a cat has a rather deep meaning. This is why linguistic comprehension is a big hurdle. Yet if we can overcome that obstacle, we may be able to create general-purpose AI.

Will AI take away jobs?
Americans are already losing their jobs due to the spread of IT.
Mid-income office jobs such as accounting, travel agency and call-centre staff positions are decreasing.
If general-purpose AI becomes a reality, these jobs may be completely wiped out. General-purpose AI can perform at a level very close to humans, and can work in any occupation. There is no need to make specialised AIs for accounting or securities trading. A general-purpose AI can do everything. On the other hand, jobs involving 1) creativity, 2) business management and 3) hospitality are unlikely to disappear.
If you come up with a unique new product, you’ll have to judge whether the product will sell. If a similar product has done well in the past, AI can make this judgement, but an unprecedented new product has no past data to work with. You’ll be able to make such a judgement. Your own sense and intuition are key.
AI will be able to take over many management jobs, but someone will still have to use human intuition and thinking to make final judgements and respond to unexpected situations. In the field of hospitality as well, humans are better suited to calming others and having a lively conversation because of their capacity for empathy.
This means it will become increasingly necessary to improve our thinking ability and sensitivity. I don’t think cramming-oriented education is completely useless, but we Japanese have not really refined our critical thinking abilities.
Mathematics is one way to polish one’s thinking ability, but philosophy is also the product of the thoughts of various people since ancient times. Philosophy is looked down on in Japan. We need to steadily adopt thought patterns or something like that, and we need to think properly on our own. In karate, for example, first you learn form. When you’re actually fighting, you make your own judgements in the heat of the moment.
We absolutely should not say that sensitivity and culture do not matter because the AI age is coming. Rather, these concepts will be the source of added value for future businesses and workers.