In my book on Configuration Management Database, I have discussed the hierarchy of DIKW (Data, Information, Knowledge and Wisdom) and further to that in another book, I have expressed my apprehension on the capability of technology to produce wisdom. I have argued that although the technology is extracting lot of knowledge from the data and information but not achieving the wisdom. There is a difference between knowledge and wisdom. In old scriptures, knowledge and wisdom have been used interchangeably because in old times knowledgeable persons were invariably wise but in modern times it is not true anymore. A person can be knowledgeable but may not be wise. Knowledge can be learned but the wisdom ought to be earned.
With the advancement in the area of “Trusted AI” I am hoping to see the possibility of “Wise AI” in near future. Being trustworthy in my opinion is one of the key characteristics of being wise.
In order for the AI systems to be trustworthy, we must ensure that it addresses all the dimensions of trust and thus inspires confidence. Although there are multiple dimensions for the trusted AI such as explainability, security and privacy, transparency, and accountability, but I would like to discuss the dimension of fairness as I consider it as the most important and the most complex.
AI systems source the data from the world that carries biases and prejudices, and these biases and prejudices can easily enter AI systems through training data. The AI systems pick these biases and prejudices, encode them, and may even scale them up. This risk brings in the necessity of calibrating and instrumenting AI for fairness. Building fairness in AI will enable us to break the chain of human biases.
We all know that biases and prejudices can eclipse the wisdom of a person but a person with strong thinking and reasoning can overcome these weaknesses and become wise. Trusted AI need to be continuously learning with newer and newer data but has limitations of thinking natively. Will that time come when machine thinking could be matched with human capability?
Yet another characteristic of wisdom is righteousness. As such, there is a systematic decline of righteousness in the society with the rise of unreason and it is being pushed back by populism. So, will anyone be interested in teaching AI system to be righteous? The prevailing trend is to be popular rather than be righteous. At times you have opportunity to be righteous as well as popular, but these kinds of opportunities are diminishing in the business and society. So, I will leave the same question on the table- will that time come when it will be viable to build a “Wise AI”?