Artificial Intelligence

Human vs Machine Intelligence

By Thomas Bolander
Associate Professor, DTU Compute, Technical University of Denmark

Even though current level Artificial Intelligence (AI) still has many challenges, there is no doubt that we will see more and more tasks being successfully automated by computers and robots in the future. Hence, it makes sense to consider what kind of human competences might still be needed in the future. We will need the following competences:

  1. Competences in seeing the potential and selecting the tasks to be automatised by AI, and, equally important, deselecting the tasks that cannot reasonably be automatised.
  2. Competences in implementing AI techniques for the tasks selected under 1.
  3. Competences to operate and collaborate with AI systems.
  4. Competences in areas that cannot be automatised.

No doubt, most of us will mainly be affected by 3 and 4. To operate and collaborate with AI systems, one does not necessarily need a deep understanding of how the systems work, but given the challenges of AI stated above, it is probably quite important that users understand the scope and limitations of such systems. If a general practitioner uses a medical diagnosis system, she has to be aware that the system cannot be expected to be flawless, and can therefore not be blindly trusted. Concerning 4, we already noted that there are certain aspects of human cognition that have so far proven exceptionally hard to simulate on a computer, most notably linguistic and social intelligence. Since almost all humans have jobs that require both linguistic and social intelligence (for communication and collaboration), not many jobs can be expected to be replace one-to-one by AI in the foreseeable future. This doesn’t imply that AI cannot lead to unemployment in certain sectors, it just means that many of the tasks most of us carry out today still have to be carried out by humans in the future.

The most important human competences of the future
As stated above, the tasks that are most easy to automatise are the most well-defined and clearly delimited ones. Those also tend to be the most routine and repetitive among our tasks. So, when trying to predict what human competences are needed in the future, we need to think about which of our tasks are least well-defined, least clearly delimited and least repetitive. Since linguistic and social intelligence are very hard problems for AI, these might become the most important human competences of the future, even for employees in technical areas like engineering. Indeed, an Australian study of the impact of automatisation on the required competences of skilled and technical workers concluded that the highest rated competences were communication, social empathy and the ability to critically evaluate digital data sources. In a case study on automation by the Danish SIRI Commission, a main conclusion was that to utilise the full potential of automation, the most important thing is to make the employees feel safe, not fearing the technology and not fearing their jobs. This is not about any specific skill set that the employees should have, but rather about their attitudes towards the AI systems. It also proves to illustrate the importance of making AI systems explainable, human-aware and trustworthy, since otherwise there is bound to be significant resistance against the use of such systems.

In addition to linguistic and social intelligence, currently humans are much better at adopting to changing norms and principles. And when it comes to creatively suggesting changes to norms and principles, we are even better. Most algorithms will at best learn and retain existing norms and principles. So, when it comes to developing our culture and decide how we want our future society, this is something that should still be designed and decided by humans.

Please find and read the complete article "Human vs machine intelligence: How they differ and what this implies for our future society"
9 AUGUST 2020