How does AI "think"?
But after all, how does AI manage to “think,” differently from traditional computers?
The key to understanding AI “thinking” lies in an analogy with how the human brain works.
Our brain works with neurons talking to other neurons.
What we did was create very tiny computer devices that also talk to other devices.
It’s basically a conversation of on and off, or 0 and 1.
This way of building is called a neural network, a network of neurons.
When we replicated these devices in the same format as our neurons, we started to generate a type of analysis very similar to our thinking.
And until today, we don’t really understand what’s happening, but the result is clear.
In fact, even scientists who build complex AIs don’t fully understand how they work, just as neuroscientists don’t fully understand the human brain.
Both are at the frontier of science.
And you, as a manager, won’t understand either, because nobody does.
What you need to know is that what we do know works and generates results.
But the results are clear: we can solve problems using this technique, problems we couldn’t solve any other way.
For example, we can’t make a machine that finds all possible chess moves.
But if we try to make that machine through a neural network, that is, an AI, it will be able to think better about how to play chess.
Without still finding the best possible way.
In fact, no human today can beat AIs specialized in playing chess.
Although AI can exceed humans in specific tasks like chess, it still thinks fundamentally differently from us, with its own limitations and strengths.
When we talk about managing AI agents, you’ll understand that knowing these limitations and strengths is what will give you an advantage in this new management.