Artificial Intelligence and Machine Learning: Part 4: Complexity

In this series I am writing several blog articles about Artificial Intelligence. Today’s article concentrates on the complexity of AI/ML algorithms and approaches. Do they necessarily need to be complex to be effective? Probably not…..

My experience starting in the mid 1970’s was that AI was very complex to implement. Not only were computers of that time completely inadequate in terms of  processing power and memory capacity, there were very few programming languages that could do the type  of programming that was needed to do AI.  I remember spending hours poring over code in CLIPS and LISP to write even simple programs. The only sources of information on AI were books from the library (Remember this is pre-Internet). This meant that generally if you wanted to implement an AI algorithm you had to become both an experienced programmer and an AI expert.

Today things have changed. Computers have plenty of processing power, speed, and memory. Anyone can spend a few hours online and quickly understand the basic precepts of AI and ML In addition programs can be written to automate the knowledge acquisition task which for years required special skills.

The tools available to do Artificial Intelligence and Machine Learning programming are largely available through the Internet for free. These include programs for things like neural networks, Bayesian Networks, Adaptive Resonance Theory, Rule-based expert systems, Finite State Machines, and a host of statistical software that unfortunately (Due mostly to companies like IBM) is still considered by some to be valid for creating AI/ML applications.

The ready accessibility of these tools and information greatly simplifies the task of implementing AI/ML in DoD programs, however companies still cast an aura of mystery around AI/ML with their customers quite possibly because they are new to the technology as well. A little research on the tools and techniques available will go a long way towards demystifying AI/ML and give buyers the knowledge they need to keep from being trapped by contractors that want to make easy  money on AI/ML projects.

In the 5th and final blog article I will introduce the Cognitive Object Reasoning Engine (CORE) Machine Learning toolset that Government Program Managers can use to meet most, if not all of their AI/ML needs. Built under the DoD Small Business Innovative Research (SBIR) Program in 1995 the Government already has limited rights to the software which is extremely easy for any non-programmer, non AI-expert to use. It is also one of the few AI/ML toolsets that provides validation of the results, making it perfect for most DoD applications where AI/ML is needed.

Andy Bevilacqua, Ph.D.
Cognitive Optical Psychophysicist

Read all 5 parts of this series

Leave a Reply

Your email address will not be published. Required fields are marked *