What is artificial intelligence or AI exactly? If that’s a question you’ve been asking yourself, then you’re in the right place because that’s the topic of this article. AI is becoming more important than ever, especially understanding what it is and what it can do and how it can benefit both people and businesses alike. That being said, there’s so many different definitions of what AI is out there. And so today I wanted to spend some time really walking you through sort of the way that I think of it, and also explain it to other people.
We’re going to be coming out with a lot more articles talking about AI, machine learning and other related topics. All right. So let’s dive in.
So what is AI?
Well, before we dive into what artificial intelligence is, I like to start with the definition of intelligence itself. You know intelligence like how we think of for humans or animals, that sort of thing. So intelligence can be roughly defined as learning, understanding, and using the knowledge learned to achieve one or more goals. So think about that.
As humans, when we’re first born, as babies, we don’t know very much, right? So we have to start learning things. And we learn from our parents and we learn from school, we learn from our friends. And we also learn from trial and error from experimenting with different things to see what’s going to happen.
So as we’re learning, we start to understand the world around us and the environment more and more, and we make sense of it. And then we use that knowledge learned to do things, whether those things are have conversations with our friends or figure out how to get to work every day, or do the tasks that we need to do while we’re working, right, we use the knowledge learned all the time to do different things.
So as a natural extension to that, artificial intelligence is simply intelligence exhibited by machines. So if we can get a machine to sort of learn somehow then understand what it learned in order to do things, like, let’s say, make a prediction, determine if an email is spam, determine whether or not there’s a cat or hot dog in a picture, then one would say that’s intelligence exhibited by machines.
And you might hear the term artificial intelligence also called cognitive computing, machine intelligence, and so on. Those are all very valid terms as well, and sort of synonymous with artificial intelligence. So again, AI is intelligence exhibited by machines, which means learning, understanding, and then using that knowledge learned to do something.
Now, AI has a history, actually came from the 1950s as a concept, but it was largely modeled originally around theories of how the brain works, and so the human brain. And so there’s this history in neuroscience, psychology and so on and so forth. But it’s also highly related to fields like computer science, mathematics, and statistics due to the fact that people have tried to sort of replicate the ideas of how the human brain works using computer algorithms and code and mathematics. So it’s very highly technical as well.
AI can be thought of in terms of three broad categories, the first is artificial narrow intelligence or ANI. The second is artificial general intelligence or AGI. And the third is artificial super intelligence or ASI. In the first case, artificial narrow intelligence, also called weak AI, that’s most of what we see today. So most AI today is sort of like a one trick pony. You might train some algorithms, some AI models, to do one thing, like make a prediction or make a recommendation, something like that, but it can’t do other things as well necessarily. Nowadays, there’s this thing called multitask learning.
So people are working towards having the ability of AI to do multiple things at once. But for the most part, AI today is actually very narrow and very so-called weak. AGI, on the other hand, is this idea of AI that has the same intelligence as humans. So is able to answer questions, understand things, comprehend things, and so on, like humans do. Today, we’re currently nowhere near that. And then finally, ASI, or artificial and super intelligence, is this idea of artificial intelligence that exceeds human intelligence.
And that’s where you get into these concepts of things like the technological singularity or, again, AI that just becomes way smarter than humans to the point that it could get out of control and completely start self-improving itself and turning. That’s where a lot of the killer robots and Terminator scenarios come into play. But again, we’re nowhere near even artificial general intelligence, which is, again, this idea of AI that’s at the intelligence level of humans.
And in fact, most AI isn’t even what a lot of people think it is. So what I mean by that is a lot of times AI has this perception of being sort of self-improving, self-regulating, self-guiding, self-learning, and then just keeps getting smarter and smarter on its own, increasing its intelligence until you get sort of like the C-3POs of the world or Ex Machina or Westworld for those of you that are fans of that show. In reality, most types of artificial intelligence, and we’ll talk about that more in future videos, exactly what those types are and the different categories, but most of them aren’t in any way sort of self-learning, self-guiding, self-improving.
There is one area of AI called reinforcement learning that would be the closest to that concept where you have a certain kind of AI, there’s this idea of an agent that’s existing in an environment. Think of video games, like a Pac-Man game. You have Pac-Man and it’s in an environment which is the board, the level that you’re on at the time. And Pac-Man is kind of going, or Ms. Pac-Man, is going around and trying to eat all the dots. So it’s trying to take different actions and it gets points as a reward as if eating the dots. And also, whenever it eats the fruit and becomes invincible, that’s sort of the state of the game changes because now the Ms. Pac-Man is invincible.
It can eat the ghosts, and so on and so forth. And so even though humans normally are the ones playing Ms. Pac-Man and controlling the joystick and moving around this sort of environment, think of if AI can do something similar, but on its own. So basically like trying over and over to figure out the best way around to eat the dots, when to eat the fruit, when to go after the ghosts, how to maximize the points, if that’s the ultimate goal, because more points you get more lives, right? So that would be sort of like reinforcement learning and this idea of the self-improving, self-guiding type of AI.
But again, the vast, vast majority of AI is really much simpler than that. It’s not self-improving. And we’ll talk more about that in future articles. Another concept that’s related to artificial intelligence is this idea of cognition and cognitive functions. So what is cognition? Well, the Oxford Dictionary defines cognition as, “The mental action or process of acquiring knowledge and understanding through thought, experiences and the senses.” So it’s kind of similar to what we said about intelligence in general, this idea of learning, understanding, and using the knowledge learned to achieve one or more goals.
And the part about the senses is really interesting here because if you think about it with humans, we have all of these different senses. We have, we can see things, we can taste things, smell things, touch things, and so on. And when we do that, we get all this data coming in through our nervous system. So as we’re sensing the world around us, that data comes in through our nervous system and it’s passed along through these what they call neural networks. And our brain, which actually is in a cavity that’s completely silent and completely dark, receives these patterns of signals, basically, coming through our nervous system that result from this incoming data through our senses.
Well, artificial intelligence works in a similar way, when you’re training certain models. You have input data it flows through, and especially in the case of deep learning and neural networks. And these networks learn things from these patterns and that learning allows a model, let’s say, to make a recommendation or predict something. Now, again, going back to cognition though, cognition is also associated with a lot of other topics that we as humans are very familiar with, things like remembering, memory, thought, thinking, let’s see, awareness, comprehension, intuition, understanding, apprehension, attention, and so on.
These are all sort of commonplace. We often don’t even think of them as humans. But to get machines to actually mimic these kinds of human cognitive functions and cognitive behaviors or functions of cognition is very, very difficult. And that’s a big part of the reason that we’re so far away from artificial general intelligence. Even this idea of understanding, that’s a very deep idea and it takes a lot more to understand things, truly comprehend and understand things. Often when you don’t have all the context or all of the sort of information that you need, humans have an ability to sort of fill in those gaps and kind of derive an understanding of a situation or something that someone’s saying to you.
Even if you don’t have all the details, you have enough memory and information stored in your brain that you can kind of piece things together to have a pretty good sense of what’s going on in a way that machines are nowhere near able to do. Another thing that’s really interesting, going back to that definition of cognition, so again, “The mental action or process of acquiring knowledge and understanding through thought, experience and the senses,” this idea of understanding through thought is pretty interesting.
Think about that. We as humans, sometimes we learn certain things or we have certain experiences in our lives and we’re trying to make sense of it. And sometimes we just sit there and we think through something for a while, we’re trying to put pieces together, two and two together in our mind. And by thinking of it, sometimes we have this aha moment and we go, “Oh yeah, I get it now.” Right. It’s almost like we’ve learned something new and understood something new completely on our own just by thinking, by recalling different bits and pieces of information that we have stored in our brain, piecing it together and creating new intelligence, new learning, new understanding, and the ability to use that new understanding to do something, right.
So in our case, again, we use our intelligence every day, whether it’s carrying out the tasks at our job, or dealing with issues with our families and friends, or dealing with our own health issues, or whatever it is, but that’s super interesting. But you know today, again, going back to this idea of sort of narrow artificial intelligence or weak AI, these models, you want to make a recommendation, you want to predict something, you want to cluster things together and create groups so that you can better market or target people through personalization or whatever the case may be, but you very much are trying to accomplish one specific goal with a certain AI sort of task.
But the AI doesn’t have the ability to just kind of go outside the bounds of that and put things together based on other domains or areas of knowledge or data or something like that. It’s very much dependent on the domain and the data that it’s using related to that domain at any given time for any given task.
And so that’s another one of the areas where there is a big gap between sometimes what people think of as AI and the reality of AI. Now, one of the questions that comes up time and time again, is what is the difference between AI machine learning, data science, neural networks and deep learning? And those are great questions. And you’re in luck because upcoming videos are going to be exactly about that. Hopefully this article has given you a much better understanding of what artificial intelligence is and isn’t.
reference – What is AI? AI Explanation and AI Types Overview
Web enthusiast. Thinker. Evil coffeeaholic. Food specialist. Reader. Twitter fanatic. Music maven. AI and Machine Learning!