The Rise of AI: Understanding Machine Learning

Pete Weishaupt
3 min readDec 1, 2022

The rise of Artificial Intelligence (AI) is a trend that will have significant implications for your business. Machine Learning (ML) is ground zero for unleashing the potential of AI for businesses, governments, and consumers.

Industries and Institutions vary, but the fundamentals stay the same. We’ll look to the Joint Artificial Intelligence Center (JAIC) publication, “Understanding AI Technology” by Greg Allen to get a better understanding of Machine Learning.

“AI is not an elixir. It is an enabler.” Lt Gen John N.T. “Jack” Shanahan.


Investors need to understand: What is AI? How does it work? What are the types of Machine Learning, and how do they differ? What are the risks and limitations of AI?


Artificial Intelligence is an umbrella term covering a broad swath of technology. When people say “they’re using AI” at work, they usually mean systems that use Machine Learning to automate processes. ML is a few years away from being an autonomous Intelligent machine able to make decisions on its own.

The important point is, while machine learning systems program themselves, human intervention is critical to the learning process. This human intervention includes choosing data and algorithms, setting the learning parameters, and troubleshooting problems.


The big reason? ML lends itself to automation tasks too complex for human programmers. The number of rules required is impractical, if not impossible, to accomplish with a human coder. ML is well suited for content generation, language translations, pattern recognition, and speech transcription.


There are multiple reasons the pause in the Fourth Industrial Revolution is now accelerating. I’ll highlight two of the big ones:

Big data

When machine learning was being developed decades ago, few applications had enough training data to be useful. Fast forward to today, and we’ve got more data than we can shake a stick at. We’re drowning in data.

Computational Power

Computer hardware is finally cheap and powerful enough to process Big Data. The switch from CPUs to GPUs, originally meant for graphics and video games, means we now have the ability to crunch calculations at blazing speeds. GPUs can speed up the algorithm training process 10 to 20x.



Supervised learning requires a human accurately label input data. Training data has images compared with the correct classification labels. Companies can purchase pre-labeled data to run an generate labels on their own data.


These algorithms extract features and group them based on their similarity. Because there’s more unlabeled data than labeled data floating around, unsupervised learning is useful for companies who want to explore their own data for insights.


Like the name implies, you take the best (or worst) of Supervised and Unsupervised learning. Often a small set of labeled data is combined with a larger set of unlabeled data.

Reinforcement Learning

Data is autonomously collected by AI agents from within the perceiving environment. It works well with games because it generates its own training data. AlphaGO is a Reinforcement Learning systems that iterated 4.9 million games in 3 days. (Consider the average game is about an hour for humans.) These algorithms explore all possible actions and learn to determine the most optimal actions what will maximize rewards.


As Greg Allen states in the Guide, organizations should not pursue AI and ML for its own sake. Allen recommends identifying specific metrics for performance and productivity. Adopting AI will require some changes to existing business processes. If companies don’t make those changes, most AI projects will deliver a fraction of the value sought.