Robots are entering every aspect of our professional and personal lives, be it in cleaning our rooms, delivering our shopping to fighting our wars. As the robotics sector develops, more and more autonomy is being built into these machines, where the simplest aspect of autonomy is decision making. One can define levels of autonomy by the degree of decision making- for example, the thermostats in your office, is autonomic at the lowest level- it is constrained to one specific predefined action, but can decide when to apply heating/ cooling. Here we define 5 levels of autonomy of machines or man-made systems
- Lowest – fully programmed and supervised (the Thermostat)
- Next – some decision making, fully supervised
- Higher – some decision making- no supervision
- Still higher- fully independent in decision and action
- Highest – fully independent in decision and action including group cooperation
Before going into examples, we also need to understand what components comprise an Autonomous System ( which is a generalization of both robotics, and artificial intelligence i.e. decision making robots and algorithms) These AS systems can then be broken into groups according to function and technologies.
As you will see, the basic model for an autonomous system is an animal – and to take an extreme, let’s take a cat as an example (and anyone owned by a cat will agree that they are fully independent and make autonomous decisions). In the list below, the groups are described by examples in brackets from our cat
Sensors (Eyes, Ears, Nose, Whiskers, Toes, etc.)
Assessment and Decision making tools and algorithms (Brain)
Actuators (Muscles, Limbs, Teeth, Claws etc.)
Communication tools (Mouth)
One of the main themes for discussion at present is what level of decision making to allow such AS in different areas of life. These thoughts bring up scary prognoses, starting from local results of failures in large scale systems (electrical or water grids) and crashes such as drones colliding with airliners, and up to the singularity notion and “humans producing a super-race that will exterminate Homo sapiens”. On the other hand, more and more types of activities reduce dangers to humans, in areas such as space, etc. At the present, most advanced AS are at level 3, with limited areas in which level 4 is achieved. The latter are usually limited in scope, such as Algotrading programs and other very time- sensitive systems.
How does all this relate to management and the Drucker Forum?
The answer includes several dimensions- first, as more and more functions are devolved to AS, or “Thinking Robots” the present definition of knowledge workers will have to change. Already now, big-data programming is more thorough in identifying complex patterns than most humans so that one can assume that knowledge based work will involve smaller groups , needing different kinds of management. In fact, it is not clear whether if the main role of the human workers will be more about managing and leading change, with the AS only analyzing the consequences of such changes and making the adjustments required to comply successfully with these changes.
Examples of this trend can be found in some firms, such as ride-sharing and allocating programs, which autonomously apply customer feedback to future operations such as downgrading the priority in allocating a driver who was reported late/impolite. Other types of autonomous big-data applications morphing into Big Brother appear under names such as metrics-driven status of employees.
A different management aspect will arise when AS reach level 4 and 5. How can we control such systems effectively, without losing the advantages of AS driven decisions and acting? Exact definitions of requested actions will have to be used in order to obtain the wished for results (the programmers among you will appreciate the implications of having a missing or superfluous comma in a program).
There have been fears that introduction of AS will result in massive unemployment. While this may be partially true, many new types of jobs, both at the top, and the bottom of the knowledge worker range can be expected. Predicting what these would be is almost impossible as the ramifications of the AS-dense world are still unfolding.
An important question is how to include ethics into the decision making process of AS. This raises even more basic questions – is there a universal definition of ethical behavior, and how should that be translated into the AS world? How does a dual-use (civilian and military) AS cope with the different approach to adversaries?
The time has come to look at the questions of applying AS in the business world, and of regulation and legal aspects of these developments, such as:
What are the absolute limits to AS independence (is there such a limit?)
How can human ingenuity and “strokes of genius” be merged into AS activity?
How does one deal with responsibility for actions by AS?
We truly live in challenging times.
About the author:
Daniel Weihs is Distinguished Professor, Israel Inst. Of Technology; Head, Autonomous Systems Program; Member , Israel Academy of Sciences; Foreign Member, US National Academy of Engineering; Past Chief Scientist, Ministry of Science and Provost Emeritus; Technion, Israel Inst. of Technology