The Simple Economics of Machine Intelligence

Quotes:

> All human activities can be described by five high-level components: data, prediction, judgment, action, and outcomes.

> When the cost of a foundational input plummets, it often affects the value of other inputs. The value goes up for complements and down for substitutes.

-------------------------

The Cost-Per-Observation [1] has been on a race toward zero as I and others have written about. That covers the data portion of the author's five level components of human activity. What they are asserting in the article is that machine learning is going through a similar cycle and will become cheap and ubiquitous. The author's reassure us that although it will decrease the value of prediction (for example diagnosis) when visiting the doctor's office. It will increase the value of the other components, that is judgement and action.

I am not so sure that judgement and action are not on the table as well for commodification. In the case of self-driving vehicles we clearly need the car itself to take action (drive) for this to be real so doesn't that imply that the intervening step of judgement has also been covered? And, if we are encompassing and automating the full chain from data to prediction to judgement and action for the task of driving a vehicle why should we trust reassurances that judgement will increase in economic value as the price of prediction falls in other industries and activities?

Interested to hear the reflections of fellow HN readers.

[1] http://computationalimagination.com/article_cpo_decreasing.p...