A copy of the full analysis can be downloaded by clicking on the link at the bottom of this blog entry.
Clearly, automation can perform many tasks more efficiently than people. What’s more, technologists continue to improve upon existing technologies that employ artificial intelligence, robots, and other forms of automation. As such, many jobs that are currently performed by humans will eventually be automated. Yet, there are still many tasks that automation will be unable to perform any time soon. This section attempts to clearly define tasks and jobs that automation can and cannot perform.
What Automation Can Do
Let’s start with what automation can do. Essentially, computers can perform any cognitive or manual tasks that can be clearly defined and specified by a set of rules. According to Frank Levy and Richard J. Murnane in “Dancing with Robots: Human Skills for Computerized Work”
… [A] computer can substitute for a human in performing a particular task when two conditions are satisfied:
• An Information Condition: All information necessary to carry out the task can be identified and acquired in a form that computers can process.
• A Processing Condition: The information processing itself can be expressed in rules.
Levy and Murnane then apply these conditions to come up with two broad workplace tasks that computers can perform:
• Routine Cognitive Tasks: Performing mental tasks that are well described by deductive or inductive rules…
• Routine Manual Tasks: Carrying out physical tasks that can be described using deductive or inductive rules…
After reading through the information on this subject, what struck me as perhaps the most significant limiting feature of computers is that their universes (inputs/data and outputs/actions) must be well-defined; that is, all possible options must be pre-specified.
Pre-specifying all inputs and outputs is easy to do when the tasks at hand are relatively limited in nature and/or scope. And This is essentially what characterizes most routine manual and cognitive tasks.
With the advent of globalization, the scale of business operations has increased over the past century from local/regional to regional/national, then international. Correspondingly, as markets have grown in size, successful companies have been forced by competition to become more efficient in their operations. One of the most fundamental means of becoming more efficient in business operations is by increasing the use of division of labor. Division of labor involves separating tasks into relatively narrow, well-defined sections. And when the relatively narrow, well-defined sections being undertaken at large scale are relatively homogeneous in nature — as is the case with mass production — the use of automation is often more cost-effective than human labor.
This exactly describes what had happened with the manufacturing and farming/resources sectors of the US economy.
What Automation Cannot Do
Now let’s consider what computers or automation cannot do. From my previous statement on the limiting features of automation, it follows that computers cannot perform tasks for which data inputs and/or outputs are not well-defined or cannot be fully pre-specified. Correspondingly, Levy and Murnane list three broad workplace tasks that computers cannot perform:
• Solving Unstructured Problems: Tackling problems that lack rules-based solutions…
• Working with New Information: Acquiring and making sense of new information for use in problem-solving or to influence the decisions of others…
• Non-Routine Manual Tasks: Carrying out physical tasks that cannot be well described in rules because they require optical recognition and fine muscle control that have proven difficult to program…
Complementing Levy and Murnane’s list of tasks that computers cannot perform, Erik Brynjolfsson and Andrew McAfee in The Second Machine Age list several types of broad skills that computers do not possess
• Ideation: tasks involving any form of new ideas, creativity, or entrepreneurship
• Large-Frame Pattern Recognition: tasks requiring the use of input from multiple senses to make decisions
• Complex Communications: tasks requiring the communication ofunstructured or ambiguous information
• Sensorimotor Skills: tasks requiring complex interaction with the physical world
Finally, in “10 Hardest Things to Teach a Robot,” William Harris provides more specific examples of Levy/Murnane and Brynjolfsson/McAfee’s broadly defined tasks and skills:
• Blaze a Trail: Collect data about the environment, process that data, and use it to make decisions. [Ideation]
• Exhibit Dexterity: Robots almost always work with consistently shaped objects in clutter-free environments. it will need an advanced sense of touch capable of detecting nearby people and cherry-picking one item from an untidy collection of stuff. [Sensorimotor Skills]
• Hold a Conversation: speech recognition is much different than natural language processing — what our brains do to extract meaning from words and sentences during a conversation. [Complex Communications, Large-Frame Pattern Recognition]
• Acquire New Skills [Ideation, Sensorimotor Skills]
• Practice Deception: Deception requires imagination — the ability to form ideas or images of external objects not present to the senses — which is something machines typically lack (see the next item on our list). They're great at processing direct input from sensors, cameras and scanners, but not so great at forming concepts that exist beyond all of that sensory data. [Complex Communications, Large-Frame Pattern Recognition]
• Anticipate Human Actions: Like deception, anticipating human action requires a robot to imagine a future state. [Large-Frame Pattern Recognition]
• Coordinate Activities With Another Robot: A robot working within a team must be able to position itself accurately in relation to teammates and must be able to communicate effectively — with other machines and with human operators. [Complex Communications, Sensorimotor Skills]
Starting and building a business requires all the cognitive skills that automation can’t perform: coming up with new ideas, communicating and interacting with other people to develop ideas and turn them into reality, testing the ideas in the marketplace, selling the ideas (products and services) to customers, carrying out and following up on the sales process.
Furthermore, there are many tasks that require the most basic of human skills that automation simply cannot perform (or won’t be able to perform anytime soon). In particular, the are many jobs that involve combinations of large-frame pattern recognition and sensorimotor skills (dexterity) that automation cannot perform, including the cleaning, maintenance, installation, and repair of grounds, buildings, and objects. Also, many tasks performed by personal care workers, attendants, and construction personnel involve both large-frame pattern recognition and sensorimotor skills, together with the addition of complex communications. In other words, unskilled jobs are not going to disappear anytime soon.
And of course, most jobs in the arts, entertainment, and recreation categories are, by their very nature, precluded from being performed by automation.