Why I Use the Term “Supervised Learning” When Creating an AI Strategy
Recently I’ve been telling friends a joke:
Q: What’s the #1 use case for AI?
A: The pitch deck.
All kidding aside, AI is at the hype-cycle peak. Many investors are (mistakenly, IMHO) asking founders for an “AI strategy”.
Unfortunately hype creates confusion and impedes discussions with customers.
AI means … self-driving cars … job-stealing robots … the end of humanity … Siri …?
Terms like AI can mean anything depending on context.
I’ve started using the term supervised machine learning — or just supervised learning — to overcome confusion and help clients develop an AI strategy.
Artificial Intelligence, Machine Learning, and Deep Learning
Let’s start with a few simple definitions.
Artificial Intelligence (AI) is a general term for intelligent machines and can mean just about anything. Is your calculator AI? I don’t see why not.
Machine Learning (ML) is a type of AI in which computers learn to perform tasks without being programmed to do so. ML engineers use data to build complex systems.
Deep Learning is a type ML which uses large neural networks. Deep learning achieves better results through large data volumes and fast computers.
Supervised Learning
Ok, now let’s cover supervised (machine) learning.
Supervised learning is a machine learning technique of training algorithms to complete tasks using labeled training data.
Supervised learning maps inputs to outputs. For example:
In supervised learning engineers take thousands (or millions) of input/output data pairs and use them to teach the algorithm how to do the mapping.
Everything else is research
Of course supervised learning isn’t the only type of machine learning. For instance, Google’s AlphaGo used reinforcement learning to beat the world’s best Go player.
But techniques like reinforcement learning are still in the research phase and haven’t made it to commercial adoption.
Supervised learning is the only relevant topic for turning data into $$$ TODAY.
Almost all practical uses of machine learning are in supervised learning.
Use supervised learning to create your strategy
Supervised learning leads to discussions of strategy and helps develop business cases. I work through 3 questions when helping clients build an “AI strategy”.
- Can we make money mapping inputs to outputs?
- Is the problem appropriate for supervised learning?
- Do we have the data we need?
Working iteratively though these questions gets us out of AI buzzword-bingo and create plans for turning data into $$$.
Had enough buzzword bingo? Email me at kevindewalt@kevindewalt.com and let’s talk about your company’s AI strategy.
