Machine Learning AI blows my mind.

So I have a formal education in classical engineering. What that means is I solve problems in heuristic, formula-based ways derived from foundational principles. Let’s say someone is throwing a ball and you want to know how far it is going to go. The way I would solve that problem is look at the velocity the ball is thrown at, the angle it was thrown, the mass of the ball, and the gravitational pull of the Earth, and use a physics formula to calculate the distance.

But…

There are other variables at play here. The wind, the air resistance of the ball, even the relative humidity. So there’s a different way to solve this problem. What if instead we had an observational experiment of someone throwing a ball, and we gathered ALL the data. Weather, time of day, date, the type of ball, etc. And we observed how far the throw went. Now let’s say we have tons and tons of these events to train an AI model, hence machine learning.

Now, instead of running a physics calculation, I take all the variables from the current event of throwing the ball, compare that against a trained model, and the model will give us an inference of “the ball will travel X meters to a degree of confidence of Y.”

Once you recognize this algorithmic difference of using inference instead of heuristics and formulas, you realize that computer science problems can suddenly be solved in completely innovative ways. For example, in classical computer science, there’s a problem called the “Traveling Salesman”. Basically, when trying to calculate the shortest distance between two paths including specific nodes (such as GPS navigation), it has been mathematically proven that there’s no easy way to solve this problem other than brute force of “try every possible path”. But with Artificial Intelligence and Machine Learning, you can solve the problem in a totally different way that is much faster. You can look at the historical data of millions of times people have driven that route, including things like weather and time of day and local events (such as sports or concerts), and you can use inference to find a least-time path.

Like I said, AI and ML blows my mind.