Recently I attended an “Artificial Intelligence (AI) and Machine Learning” event hosted by the Capital Area IT Council (CAITC) at the East Lansing Technology Innovation Center. There was a big crowd and I was immediately greeted by Jenny McCullough, the Director of CAITC, who suggested I head towards a generous variety of complimentary snacks and beverages.
Joe Ellsworth, the Chief Technology Officer (CIO) of Delta Dental was the sole Presenter. Joe lives in Seattle and has a wealth of experience, primarily as a Principal Architect. Joe has worked with Amazon, Hewlett Packard Enterprise (HPE), and Angie’s List, to name a few. He was very friendly and had genuine interest in all attendees’ interest in the topic.
The first point Joe stressed was the difference between General AI and Narrow AI. General AI refers to a machine that incorporates all human characteristics. So far, General AI has been a very challenging arena, particularly due to the complexity of the human brain. Narrow AI includes only partial features of human intelligence and is widely researched and utilized currently around the world.
Machine Learning is a sub-class of AI. Machine Learning seeks to train the machine to learn from collected data by forcing the human to not explain every rule. This is an automated process that can adjust to changing patterns and essentially fills the “Statistician” void. Actually, Joe described Machine Learning as “college statistics on steroids.” Some real-world examples of Machine Learning in play include:
- Word Suggested Auto-Type
- Stock price predication (Joe dedicated four (4) years of his life to the algorithm)
- People Match
- Advertisement Match
Expert Systems describe a machine that mimics the human decision-making ability. Machine learning is starting to show up in Expert Systems. The issue is that data needs real time to be collected. Generally, Expert Systems are divided into two (2) categories; Knowledge Base and Inference Engine. Knowledge Base represents hard facts and rules. Inference Engine applies rules to the known data/facts to form a logical conclusion.
One of the biggest issues with implementation is that people who do not understand the systems happen to be the ones feeding data into an engine in a naive fashion. In general, Joe referred to these folks as your typical lazy humans.
In conclusion, future innovation will be made via massive amounts of data. We must compile it before proper analysis is complete, and to date this is not the reality. With certain specifics, the hope is to one day predict Alzheimer’s, depression, and other widely diagnosed disorders. The evolution of AI and Machine Learning will continue to move forward with time.