If you ever have a chance to tour a modern car factory in person, or if you watch a video on YouTube that takes you inside one, you will notice an interesting bifurcation. Some parts of the factory are completely robotic.
For example, the welding line that puts the body of a car together can be devoid of humans, except for the few humans who are there to monitor and repair the welding robots. The paint shop can also be completely robotic, as can something like engine manufacturing.
However, if you walk over to the general assembly area of a car factory, where the interior of the car gets installed, there are still lots of human beings. There are some robots yes, but there is nothing like the total robotic domination seen when welding, painting or assembling engines.
If we step back and look at the “job market” from 30,000 feet, we see this kind of bifurcation in many different ways. Agriculture is now highly automated, but restaurants are not.
Or think about Blockbuster’s business model, which at the peak employed 84,000 people. It was destroyed and replaced by a combination of Netflix (which mailed DVDs to people) and Redbox (which uses very compact DVD robots). And then streaming services have destroyed DVDs.
But if Blockbuster employees all disappeared, why not grocery employees? Modern grocery supermarkets are largely unchanged in the number of employees they have when compared to 100 years ago. In fact, the whole “service sector” of the economy has been highly resistant to a robot takeover.
Or think about home construction. Houses 100 years ago were “stick built” by tradespeople like carpenters, roofers, painters and electricians. In 2021 the process is very nearly the same.
People used to vacuum their own floors too.
Why are some jobs lost to robots but not others?
So what gives? Why do some jobs get completely automated out of existence in today’s economy, while other jobs do not?
One big reason is the flexibility of human vision systems and human hands.
Think about something as simple as restocking a shelf in a grocery store. It is trivially easy for a human being to look at the shelf and move new products from their boxes to the correct shelf position. But up to this point robots have not been able to come anywhere close to human-level vision, so no robots are restocking shelves right now . . .
To put it another way: An industrial robot can weld two body panels together while being blind. The panels are precisely positioned by jigs and the welding path is always the same.
But it is a completely different story when it comes time to restock groceries – the task impossible for a blind robot.
These same advances are also crucial in the world of
A tsunami of job loss
The problem for the job market is that we stand on the brink of a revolution in robot vision. It has taken humanity a long time to create inexpensive computing power at a level required for robot vision, but we are finally arriving.
For example, Google’s new gen4 TensorFlow chip can now perform 200 teraflops. GPU and CPU chips are advancing at a similar pace. With this kind of computational horsepower and the right algorithms, high quality computer vision systems will soon be here.
Take Walmart as an example of what is likely to happen. Walmart employs 1.6 million people in the United States. Quite a few of these people – the shelf stockers, the floor cleaners, the remaining cashiers, etc. – are likely to be replaced when robots can see and manipulate things like humans do.
And as soon as Walmart makes a move, so will Target, Kroger, Publix, etc. Amazon warehouses are currently a mix of robots and people. The humans will be gone once robots can see and manipulate things.
Much of the construction industry is vulnerable, as are myriad factory jobs. Even the cooks in restaurants and many of the waiters/waitresses will be replaced, as well as parts of the healthcare industry.
It will not all happen overnight, but good computer vision systems will soon become a real gamechanger in the employment marketplace.
What does it mean?
The problem is that quite a few jobs in the US economy have been protected by the lack of vision systems. A look at Bureau of Labor Statistics data can give us a sense of how many jobs are in play once effective computer vision systems arrive.
Here are several examples:
- There are 7 million construction jobs available in the economy. A million of them or more are repetitive tasks that could be handled by robots.
- 12 million manufacturing jobs are available in the economy. Several million of them will easily fall to vision-enabled robots.
- There are 15 million jobs in retail. Perhaps half of them are easy for vision-enabled robots to handle.
- Many of the 4 million transportation and warehousing jobs will be replaced by robots.
- There are 14 million jobs in Leisure and Hospitality, many of which robots can handle once they have vision.
Looking across the board, it is easy to imagine that 10 million jobs could be lost to vision-enabled robots in the first wave.
Which industry will be first? This is an interesting game to play. Think about all the jobs where vision-enabled robots could make an impact, and then try to predict the first job category that will see significant job displacements.
Will it be grocery stores? Restaurants? Construction sites? Hotels?
And when will it happen? Three years? Five years? Ten?
At an intellectual level it will be fascinating to watch the transition to a robotic workforce unfold. But if the transition is not handled gracefully at the societal level, it could be tragic for those who are caught in the unemployment crossfire.