Monday, November 27, 2017

AI for policy

Humans, inconsistent, unstable and biased, have been ill-equipped to make optimal policy choices. In the modern era, rich with dynamic data, this problem has been magnified many times. Career bureaucrats and politicians with countless conflicts of interests have been running amok. In the process, they are degrading the advantages built up by generations in small steps. Just as a hedge fund manager, proud of her small victories over a long time, end up losing the entire pot in a few seconds in a massive and unanticipated discontinuity, politicians are playing with fire that could have disastrous consequences.
It is about time we delegated policy-making to machines. They have been impressive in the presence of large amounts of dynamic data and they are silent and efficient learners. Lack of ego gives them a distinct advantage over humans for their objective functions are programmed to have an unambiguous positive slope in knowledge. Failure does not seem to bother them and they always learn from it. And, they take empirical validation to be the truth and not opinions and fake news. In spite of their lack of education, they are quickly moving to a position of superiority and for the first time in history, we can look forward to a regime of rationality, driven by machines.
In this context, policy-making is an important domain for applications of AI. It is ironic that a few thousand people, making irrational decisions without knowledge or data, has become the gold standard in governance. The fact that most of them do not have a technical training but they assume to be "good enough," to make policy choices in a regime of technology, is puzzling. They appear to be disconnected from the millennials, who will be most affected by the choices they make and it is disturbing. Perhaps, they read a book or two during their summer break and understood that the internet is not a series of tubes. But, has anybody told them that it will not be enough? Can they pass a basic competency test? If not, it is time to move on.
As Sci-Fi enthusiasts hold their breath for AI to take over the world, a more common-place solution will be to replace policy-makers with it. Machines are great optimizers of policies without consideration of color, wealth, education, age, gender or orientation. Humans will have great difficulty competing with this superior knowledge.