This article is based on a talk by Emerj.com (formerly TechEmergence) CEO Daniel Faggella at a recently held independently organized TEDxURI event. Part of the talk was on the impact of artificial intelligence on job security.
When I speak to my business readers or with businesspeople at conferences and events, the topic is always the same: The applications and implications of artificial intelligence in business
When I talk to family members or friends of friends, the topic is also remarkably predictable: The automation of jobs by artificial intelligence
This lead me to devote an entire TEDx talk to this topic and to spend three long weeks digging through hundreds of interviews to find the common threads and themes about job automation from all the experts we’ve spoken to over the last five years. I’ve decided to distill these ideas into a single article—and here it is.
Below is the full video of the TEDx presentation on which this article is based. What follows is a breakdown of what I consider to be the three most important elements of job automation:
Job Automation – What Researchers Say
We periodically carry out interviews and surveys of those that work with the hard science of AI. To keep up with current trends and issues, we talk to those who have a deep understanding of AI and its attendant technologies.
For one of our surveys, we polled over 30 researchers—most of whom have PhDs in various disciplines including artificial intelligence, computer science, robotics, and linguistics—about what they believed would be the risks of AI in 20 years and 100 years.
Various responses ranged from general mismanagement of AI to killer robots. However, of the 33 respondents, 12 believed that the biggest risks of AI would be to the economy through automation.
Below, respondents explain how they believed AI would change the job landscape:
“Drastic changes to the employment market due to increased capability of autonomous systems (it’s more of a certainty than a risk and we should already be factoring this into education policies).” — Dr. Helgi Helgason, PhD in Artificial Intelligence, Reykjavik University (Iceland), VP Operational Intelligence at Activity Stream
“Massive unemployment and resulting huge wealth disparities.” — Prof. Lyle Ungar, Professor of Computer and Information Science, Bioengineering Science, Genomics and Computational Biology, Operations and Information Management, and Psychology at the University of Pennsylvania
”Massive unemployment. Machines will be singing the song, “Anything you can do, I can do better; I can do anything better than you.” — Dr. Nils Nilsson, PhD in Electrical Engineering, Stanford Professor of Engineering (Emeritus) in the Department of Computer Science at Stanford University, Kumagai Professor of Engineering (Emeritus) in Computer Science at Stanford.
It is apparent from the survey that many AI researchers believe that the AI risk to job security is a real concern, and they believe it is already happening today. This is supported by numerous other interviews and surveys carried out by TechEmergence at various points in time.
Three Factors Ensuring Job Security – Context, Coordination and Connection
A rundown of recent news articles, as well as an MIT meta-analysis on job loss predictions, indicates that AI will certainly take over quite a few jobs in the future. However, there are many reasons why machines will not make all humans obsolete, and three factors stand out for people that have made it their business to know.
Insights on this topic have emerged from interviewing numerous executives and researchers and diving back into previous surveys on TechEmergence that discuss the impact of AI on jobs in different sectors. From our research, we’ve found three factors that could ensure job security and fend off automation in the future.
Factor 1: Context
Most people believe that machines will take over mostly blue-collar jobs, such as in factories and manufacturing. While machines can improve some efficiencies, some tasks will require more context than machinery can handle.
In the domain of welding, it is possible to automate the task in a repeatable or predictable production line. There has to be a specific kind of input that goes through a specific process to produce a specific output.
This type of rote work is particularly suited to automation. However, it is much harder to automate the repair of an automobile, as the welder has to make a judgment call based on the situation. For example, the welder would need to decide whether they would bang out a damaged portion of a car or cut through it and replace it.
Many skilled workers, in particular, are not automatable. Take a plumber, for instance. The work is not predictable at all. When a plumber shows up at a house, many factors come into play that the plumber has to consider. The plumber may need to know the age of the boiler, the type of pipes, seasonality of the problem, and so on.
Only after getting the answers to these kinds of questions can the plumber even start to figure out what to do. This type of job involves a lot of context as well as physical dexterity. Robots are not able towalk like humans yet, so changing a pipe under the sink is probably out of the question.
Another skilled type of work considered blue-collar is clothing and shoe production. Two companies claim to have developed AI-based software that can guide a robot to make shoes and sew clothes. However, these systems require humans to work alongside them and can only handle very specific and simple production lines. Much of the designing and implementing still depends largely on humans to work.
Contrary to popular belief, white-collar jobs may actually be more automatable than many skilled blue-collar jobs. Quite a lot of automation implementation is going on in the world of white-collar workers, from life insurance agents to medical personnel.
However, the most interesting thing about automation in these sectors is that the same rule applies as for blue-collar work: if it is repeatable and predictable, it is automatable. It does not matter how much a person is paid to do the work.
A good example is in finance. An auditor, for example, that mainly goes through financial reports looking for variations or anomalies in order to pass them to the next person in the workflow is doing an automatable job. It is just a matter of detecting glitches in the reports, which an AI system can do. The same holds true for many bookkeeping and accounting tasks, where the work primarily involves handling and processing data.
However, a purchasing officer may also be looking for variations in the same set of financial reports, but will then use that information to make purchasing decisions. That is contextual and makes the work much harder to automate. The auditor and purchasing officer may get about the same pay, but one is automatable, and the other is not.
Another good example is medical diagnostic AI, of which about a third of all SaaS AI companies focus. Current applications indicate that AI is not about to replace doctors for the simple matter of context.
Tests, machine vision, and medical records go a long way in reducing the huge amount of available data into manageable chunks to arrive at potential diagnoses. However, crunching the numbers and detecting patterns can only go so far. It takes that and a doctor to see, smell, hear and touch a patient to gather subtle clues that can lead to discovering the actual health problem.
Context is probably the most salient point that emerged from all of the interviews and surveys performed. In a nutshell, if it is possible for someone to do a job in isolation without having to take into account any outside factors, influences, or situation, then job security in the face of AI is not assured.
Context leads to complexity, and the real world is complex. Because of this, the best way to ensure job security is to embrace the complexities of a particular task and get involved in it in order to stay relevant.
Factor 2: Coordination
Context refers to situations, but people often create these situations, directly or indirectly. The second factor that can ensure job security in the face of AI is coordination.
Coordination is primarily about managing people. It could be a small team of scientists working to find a cure for cancer. It could be a group of 50 sales representatives that are deployed throughout a state. In either case, the human element is the most important and confounding in effective coordination.
While an AI system may match Scientist A to Scientist B as the best people to work on a certain portion of a clinical study because of their academic backgrounds, working together if they dislike each other may not provide optimal results.
In the case of a sales team, historical data may show that Sales Rep A has sold the highest number of products in a particular area. However, a new employee, Sales Rep B, lives in that particular area and has excellent relationships with retailers.
A human coordinator can understand how those dynamics can affect an outcome. An AI system will crunch the numbers and provide the logical choices, which will not always work.
That said, AI does have a place in coordination, and that is in providing venues for effective team communication. This is particularly important when coordinating people in the field or in another country.
Factor 3: Connection
In a similar way, connection is important in ensuring job security in many sectors. It does not necessarily involve connecting with people in order to manage them, but jobs that require some type of human connection are probably going to be safe from human obsolescence.
Call center agents to some degree are at risk from AI systems. Natural language processing can handle some initial interactions in the customer service tiers, such as self-service tasks (billing inquiries, identity verification), before referring to a live agent.
AI systems are also useful in customer service for sentiment analysis and customer data compilation. However, when the caller requires anything more complex than a scripted set of instructions or suggestions, a human connection needs to be established.
However, student-teacher relationships have a profound effect on student engagement, and this has a direct impact on learning. It is not likely that students, especially young ones, will engage with a robotic teacher in quite the same way as a human teacher.
Concluding Thoughts – Keep a Firm Grasp on Context
According to asurvey by Bloomberg, some of the highest paid jobs are also the most at-risk in the face of AI. These include compensation and benefits managers, accountants and credit analysts. Among the least vulnerable are doctors, teachers, and dentists.
It is immediately apparent that the least vulnerable occupations are those that require human interaction and context at some level. The most vulnerable positions deal primarily with numbers and data. It is also interesting to note that blue-collar workers (except for truck drivers) are absent from the list of survey participants in our own AI risk analysis (mentioned with a graphic earlier in this article).
As artificial intelligence progresses, programs will preempt human needs and take automatic action. The evolution of technology will be more than just upgrades in the tools used for work. It will involve different ways to relate and indeed live as human beings in an increasingly virtual world. The rolling process of creative destruction is moving faster and faster, and staying ahead of that is a major reason for keeping a grasp on context.
If you want to see your own interests represented in the future of your industry, you must have your hands around the messy context of the industry itself.
What skills are becoming more, or less, important?
How is the nature of the customer changing? How is the nature of the product changing?
Given the speed and pace of change, these aren’t questions to be handled just by founders and CEOs, but by anyone who wants to stay relevant and employable in the future we’re creating together.
The factors discussed in this article are useful for anyone seeking job security in the face of AI as a litmus test to determine whether it will affect a certain career trajectory. It would be safe to say that for most people to whom these three concepts have no relevance, where interactions are so limited that it is not a critical part of their work, or whose activities are repeatable and scriptable, are the ones most likely to be at risk of losing their jobs to automation.
There are a handful of our podcast interviews that involved specific and potent ideas about job automation that made their way into the article you see above (and the TEDx talk upon which it was based). Below are three podcasts that I found to be particularly useful to frame my research:
- Surviving the Machine Age – Technological Job Loss with Dr. Kevin LaGrandeur
- Technological Unemployment and the Role of Man and Machine with Marshall Brain
- Will Doctors Fear Being Replaced by AI in the Hospital Setting? with VC Steve Gullans