TechnoPath News
Industrial Robot Arm

Machine Learning: The Need for Diverse Data

This area of our technical evolution requires much out of you and I.

“Amidst the politics, environmental concerns and social conflicts, we must see that our most important task currently bestowed upon us will be helping our artificial intelligent machines “overstand” the definition and value of humanity.”

At the core of the principles of machine learning, is a variable of diversity.  From a technical perspective, the word “diversity” can be void of human’s bottle-neck construct of race and introduced as a quantification involved in calculating the intelligence quotient of an AI entity. The more diverse … the more intelligent.  Simple math.

Dr. Subbarao Kambhampati, Computer Science Professor at Arizona State University, (Dr. Rao) drops a gem about the biggest flaw detriment being introduced to artificial intelligence.

Dr. Rao provides a very clear example of the need for diverse data in machine learning.  What should make mankind optimistic is the current opportunity to understand this need now and set the proper protocol.

AI is being applied to your life NOW!

Self-driving cars, predictive legal/crime assistants, and banking algorithms are currently applying artificial intelligence to our lives.  All areas will benefit from humans taking the time to introduce diverse data sets.  Deep machine learning of “what it means to be human” has already begun.

The Wells Fargo Example

A recent banking misfortune at a Wells Fargo ATM and bank branch caused me to place more importance on my roll in educating our current and future business leaders about the perils of biased data.  My interactions with Wells Fargo staff surrounding the erroneous generation of ATM code #SignatureIrregularity forced me to look at how artificial intelligence could help an unethical goliath company do a lot of damage to the American small business community while making a tremendous profit.

Wells Fargo may have applied artificial intelligence in a predictive manner to select check deposits from clients with the following data criteria:

  • least troublesome non-payroll, personally endorsed incoming check deposits
  • coming to “low-risk of damage to WF” clients who just have to accept detrimental action with no discourse
  • set a period long enough to execute alternate transactions and short enough to contain negative action from the WF client

In a situation like this, if no one speaks up for small businesses with very modest financial flow, a future of seeing American small businesses dropping like flies could be right around the corner.

Diverse data is the quintessential key to raising the overall intelligence of all business solutions.