How I Explain ML Algorithms to Business Managers....

(in less than 7 minutes)

I think of algorithms as the core part of machine learning and AI, with these algorithms divided into traditional AI methods and GenAI methods. Traditional AI methods include supervised learning techniques, unsupervised learning techniques, and reinforcement learning, but more recently, self-supervised techniques powering apps like ChatGPT have gained prominence.

When it comes to supervised learning, it's all about the availability of labels. If labels are available, great; we use supervised learning techniques like tree-based models and regression models, etc.

If not, then I turn to unsupervised learning techniques, such as K-means clustering and naive bayes, for example.

If there's a complex environment that requires action and there's a reward loop involved, then that's likely a good candidate for reinforcement learning techniques. Good examples of this include Tesla's Autopilot and DeepMind's AlphaGo.

More recently, we've seen a lot of attention on self-supervised learning techniques in the GenAI space, like the base large language model that powers ChatGPT.

I've tried to pack as much information as possible into a short video. Check it out below.

If you have any questions, feel free to drop them. If you're a practitioner, please add your insights as well. Thank you.