Very few jobs are entirely automatable. It makes sense, therefore, to ask which tasks will be automated – Kai-Fu Lee (2018) has a very good answer. Lee thinks that “AI creates a mixed bag of winners and losers depending on the exact content of job tasks performed”. This means that in one job, more tasks will be computerised, whereas, in another, very few tasks are automatable.
Currently, AI can do amazing things. For example, think about a self-driving car. However, the current version of AI is narrow AI. It is better than humans in specific tasks in a single area, for example, image recognition. However, narrow AI lacks the capacity to perform in many domains as humans do. Humans can read, understand language, speak, solve math problems and compose music. This is known as cross-domain thinking. AI lacks cross-domain thinking and struggles with tasks that require creativity and complex planning.
What does this mean for the risk of job replacement? Kai-Fu Lee (2018) says that the risk of job replacement is different for physical labour, that is, the repetitive tangible and physical execution of tasks and for cognitive labour, which requires mental processes solve problems that do not necessarily require physical intervention. Cognitive, non-physical jobs that comprise routine tasks such as accounting and financial advisory have already been automated. More jobs in this category will be automated. Creative or strategy-based jobs that require mental processes will remain relatively safe in the near future, even if they do not involve interaction with people. Examples are a medical researcher, qualitative researcher and artist. Jobs that require a lot of human interaction, such as social worker or wedding planner, have a low risk of replacement because they require a lot of social interaction even though they require low levels of creativity. Cognitive tasks that can be optimised with algorithms and data and requires little human interaction are in great danger of replacement. A radiologist and financial planner are examples.