We Must Stop Approaching Artificial Intelligence As a Technology


Artificial Intelligence is not Automation

We make a major error when we think of AI as a technology, or perhaps as just another technology. This error can turn out to be a costly mistake. Specifically, using the term “technology” to describe artificial intelligence can influence us to ignore the perils of AI, to accept it as we have accepted other technologies, and to incorrectly assume that in the past 3 centuries we have acquired enough experience to deal with technological revolutions. AI is different and hence it should be approached completely differently. Most importantly it should be approached humbly – and with an understanding that this is not something we have dealt before.

The anxiety about automation is rising again. In 2013, Daniel Akst wrote an article highlighting that humans have experienced this anxiety before and that it is happening again (Akst, 2013).

We are getting more and more concerned about the potential job loss resulting from automation i.e. the concept that as machines replace human labor, more and more people will be out of job.

Recently, Frey and Osborne estimated that just about half the jobs can be impacted by the rise of artificial intelligence related automation (Frey and Osborne, 2013).

Brynjolfsson and McAfee, two MIT professors, have studied the emerging phenomenon in great depth (Brynjolfsson and McAfee, 2015). Specifically, they observed that four measures of economic health: per capita GDP, labor productivity,the number of jobs, and median household income tend to grow in tandem but in the last 15 years GDP and Productivity have grown while Income and jobs have declined – a concept they call the Great Decoupling.(Interview HBR, 2015).

What Brynjolfsson and others, for example Autor (Autor, 2015),have observed is that while it is possible that there will be jobs, the need for skilled jobs will grow while the demand for unskilled or lower skilled jobs will diminish. Also, what was considered a “high skill” job today may be downgraded to low skill job tomorrow.

As recently as last month Jason Furman, Chairman, Council of Economic Advisors to the White House, referred to the “shift in the impact of automation on the labor market”.

While I completely agree with the above analyses – I would add one factor to the analyses that, I believe, has been ignored. The problem with the above analyses is that we are approaching the technological transformation only from the angle of “automation”.

It is natural for us to approach technology in this limited sense. Our social sciences developed under the shadows of the industrial revolution. Our understanding of technology, technological unemployment, technology economics, and impact of technology on economics, politics, and society are all shaped by a certain concept of machines replacing humans. For the most part, this concept of “automation” implies a task involved in the production or delivery of goods and services and where once it was performed by humans, it is now performed by machines.

It is important to make a distinction here between machines and humans. Machines are mechanical devices, that, in our current thinking, perform very specific and limited tasks, completely under the command of humans, do not have their own thought process, and their performance potential is completely predictable due to repeatable nature of their function.

Imagine the following scenario. A child is drinking milk and she spills it on the floor. When we think about vacuum cleaner, we don’t expect that the vacuum cleaner will notice the spill, will plug itself into a power outlet, will quickly rush to clean it; once it approaches the spill it will realize that it is not capable of sucking in the liquid and then it will figure out a way to sweep and soak in the milk.

We know that our concept of machine, in the above example a vacuum cleaner, is limited to be used for cleaning dirt and other stuff from the floor and not for cleaning liquid spills (specific and limited use), that the vacuum cleaner will only become operational under human authority and can’t plug itself in (under human control), that it will not start cleaning dishes (completely predictable function and potential, repeatable tasks), and that it doesn’t have its own thought process to create, learn, or develop. Thus when we think about automation, we tend to think in terms of physical automation and not necessarily cognitive automation. Thus the basis of our analysis always tends to be derived from the limited concept of machine and its potential.

The advent of the information revolution should have compelled us to think differently. Unfortunately, we continued to apply the concepts of the industrial automation in the rise of information technology. We paid less attention to the cognitive augmentation of the information technology and more importance to the “automation”. For example, consider the following:

  • While the Great Recession of the 2008 has been analyzed from various angles, we have yet to thoroughly analyze it from the perspective of the role played by the rise of information technology in accelerating and deepening the recession. The spread of risk in the global economy, the price transparency, the ability to create exotic and highly complex financial derivative products, the bad news spreading fast, the behavior of markets, etc. has all been made possible due to the “availability” of information.
  • Since the maturity of the information revolution, the interest rates have been low, and the economy has been run by the constant government intervention (e.g. quantitative easing programs, saving large institutions, and now promise of launching mega-infrastructure projects) to sustain the economy. Some economists call it “the new normal”. The question is: is there a connection between the information economy and the government’s inability to raise interest rates and the need for constant influx of government interventions?
  • Is there a link between the information revolution and the manufacturing job loss? After all the ability to discover, transact, and exchange goods, services, and currency were facilitated by the rise of the information technology. Hence, the loss of manufacturing jobs to China is more related to information technology, than perhaps to the cost savings. In other words, simply cheap labor and production capability in China may not have been enough to drive manufacturing there. The ability to transact, to identify suppliers, to manage the supply chain – have all lowered the economic transaction cost and hence increased the ability to transact (more along the lines of thinking of Douglas North, Robert Coase, and Oliver Williamson).

Just like the rise of information technology was a lot more than just “automation” and its economic effects on job loss were far more complex than simply the ones resulting from “automation”, the impact of artificial intelligence will be far more complex.

The biggest mistake we can make is to assume that artificial intelligence is all about automation. It is about cognition, consciousness, information, and awareness.

We don’t have any experience in dealing with those types of machines. And therefore, we must approach this humbly and with an extremely high degree of awareness.

I will write more about this in the upcoming weeks.


Akst, D. (2013) What can we learn from past anxiety over automation? by Daniel Akst — Summer 2013: Where Have All The Jobs Gone? | The Wilson Quarterly 

Autor, D. H. (2015) Why are there still so many jobs? The history and future of workplace automation. The Journal of Economic Perspectives.

Brynjolfsson, E. & McAfee, A. (2015) Will humans go the way of horses? Labor in the second machine age. Foreign Affairs. 94 (4), .

Frey C. B. and Osborne M. A. 2013 The Future of Employment: How Susceptible Are Jobs to Computerisation? Oxford University http://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf

Interview HBR (2015) The Great Decoupling. Harvard Business Review. (June), 68–74.