The breakthrough that lead to the current surge in Artificial Intelligence.
In: 0
Honestly? It’s not one single breakthrough, it’s a lot of factors that contributed to it, but I can think of three main ones; hardware becoming better, big open-source or curated datasets, and the rise of more advanced “learning” algorithms like neural networks.
The first one is self-explanatory; as hardware becomes better, it can run more complicated code. This makes it way easier to put complex concepts that take a lot of processing power into use, like we’re able to process big data sets and do more “analysis” I guess on them to get better outputs.
The rising availability of large data sets, like millions of labelled pictures or poems or whatever, allows more people and companies to have more material to teach their “AI.” Essentially how these “AI” models work is taking data, and then based on that data saying “okay, so when I see [thing1], there’s an x percent change [thing2] is involved,” and associating things with each other. I hesitate to call what we all have begun to collectively refer to as artificial intelligence… well, intelligent because it’s essentially just a pattern finder. When it finds enough patterns, it can almost convincingly imitate a human, but can’t really invent any novel ideas or anything. It gets more and more realistic with this big amount of data though.
A “neural network” is probably something you’ve heard recently, but essentially a neural network is an attempt to approach machine learning by comparing it to the human brain. While the first neural network was made, like, seventy years ago, it couldn’t do the same things that we do today with neural networks because of a lack of technology and data available for it to use— see, we’re back to the previous points. Recently obtaining that has allowed the approach to become more mainstream. Instead of using things like tree classifiers or random forests or whatever else (not super important to just explain the “breakthrough,” but super interesting [I am biased]), neural networks basically loop back on themselves to get better at guessing things, being kind of self-teaching, I guess? Not sure exactly how to explain it, but kind of like neurons in the human brain. This makes pattern recognition better, but for some things, you still want to use other methods. The big things recently have been neural nets!
That one got a little long, hope it still makes sense 🙂
Latest Answers