Machine Learning is Like a Baby
Subscribe to our YouTube channel and never miss a thing!
Machine Learning is Like a Baby — Video Transcript
Machine learning is like a baby. A baby who needs frequent supervision from parents to learn. Today I'm going to share with you three key similarities and three important differences between babies and machine learning. Let's dive in!
Machine Learning is Like a Baby — 3 Key Similarities
They Both Learn — Babies are incredibly fast learners. If you show a ping pong ball to a baby learning how to talk, pretty soon, if you show the baby a basketball for the first time, it will correctly identify that a basketball is also a ball, even though it's never seen one before. Machine learning is our attempt to build a digital baby brain that learns from experiences we give it to learn to recognize other similar objects and patterns that it has never seen before. Like a baby's brain, machine learning models learn from experience. This ability to learn over time is different than a software app on your phone that has been explicitly programmed to handle every possible situation it may encounter as you use it.
Experiences Matter — Babies get smarter as they experience more of the world around them. Machine learning models also become more intelligent the more training data we feed them. Like babies, they become better at understanding the world around them, which helps them make better decisions when facing a situation they haven't seen before.
Feedback Is Important — Babies rely on feedback to learn. They might learn not to touch a hot stove because they got burned the last time they did. From that accident, the baby updated its view of the world with a reminder to avoid touching stoves because they may be hot. Machine learning models also learn from feedback. If a machine learning model incorrectly guesses that a basketball is a cat, we can tell the model that it did not guess correctly, and that in fact, it is a ball. With this feedback, we can then retrain our model to update its view of the world and become smarter.
Now that we've seen how machine learning is like a baby, let's look at three important differences:
Machine Learning is Like a Baby — 3 Important Differences
General vs. Specific Learning — Babies are remarkable because there is virtually no limit to what they can learn! On the other hand, machine learning models can only learn how to do very specific tasks, like guess what song you want to hear next or which product you might want to buy.
Small vs. Big Data — Babies can learn much more about the world around them with far fewer experiences or "training data" than machine learning models. Machine learning models still require that we feed them many labeled examples of, for example, "this is a cat" and "this a dog" before it learns to tell them apart correctly. Babies may be able to tell them apart after they see only one example of each.
Learning Speed — Babies are continually updating their mental model of the world based on their experiences and feedback in response to their actions. As the world changes, babies can quickly adapt. On the other hand, machine learning models need to be told when to retrain and update their models. If not properly managed and monitored, machine learning models may get dumber over time if they cannot adapt to the changing world.
What's important to remember is that both babies and machine learning models require human parents to learn and grow. Both babies and machine learning models may also pick up on our biased thinking about the world, even when we don't know we have them.
For now, AI is still a baby, and it is not here to replace us anytime soon. In this video, we only scratched the surface of what is "AI." We hope you'll come back for more! I'll see you out there, and have an awesome day!
Stay Current in AI, Machine Learning, & Data Science Technology
Subscribe to our free weekly newsletter.
Subscribe to our weekly Data Science & Machine Learning Technology Newsletter
