Robots have been automating blue collar occupations such as manufacturing. But some heavily cerebral occupations have been immune to automation. But that's starting to change. Robots are going to automate some professional occupations
They list manual dexterity as one of the human skills that robots have a hard time with. But it really depends on setting and risks. A robotic fruit picker has easier safety requirements than a robotic nurse. A robotic fruit picker also has an enormously shorter list of tasks than does a robotic nurse.
Highly routine professional tasks are more prone to automation. A substantial chunk of accounting can get automated, especially as more of the data arrives in automated data flows with standard formats.
One claim in the article is that social intelligence is not going to get automated any time soon. But a lot of companies are hiring machine learning modelers to predict a lengthening list of things about what people might do. Those prediction models often can beat humans in specialized categories of predictions about what we want or what we might do. For example, models can predict the rate of criminal recidivism and even odds of becoming a criminal at date of birth. Such models will become much more accurate when DNA sequencing costs fall another order of magnitude and everyone gets their DNA sequenced.
As his work has been put into use across the country, Berk’s academic pursuits have become progressively fantastical. He’s currently working on an algorithm that he says will be able to predict at the time of someone’s birth how likely she is to commit a crime by the time she turns 18. The only limit to applications like this, in Berk’s mind, is the data he can find to feed into them.
Rather than think of a single thing called social intelligence it is better to think of a long list of tasks that involve social intelligence. Then for each task ask how susceptible it is to automation. Jobs that require workers to make predictions about human behavior (e.g. parole boards and some pieces of police detective work) are certainly susceptible to automation for the prediction part of the job. That's true for a large chunk of marketing (e.g. who to send each sort of credit card offer).
But an institution has to be willing to use and accept the output of prediction models. That's not always the case. College admissions could be done much more accurately with prediction models than with human application reviewers. Models could greatly exceed the performance of current admissions staff. But colleges do not want to use the models. A college of lower rank that wanted to raise its ranking and did not mind using models could come up with models for who to offer scholarships for and at what dollar amount to get higher ranked students. A college could rank all applicants using a prediction model of likely college performance. The opportunity exists for the college willing to automate and use a model.
|Share |||Randall Parker, 2016 July 30 04:21 PM|