5 minutes 48 seconds
00:00:00 - 00:00:11
Artificial intelligence and machine learning. What's the difference? Are they the same? Well, some people kind of frame the question this way. It's AI versus ML.
00:00:11 - 00:00:27
Is that the right way to think of this? Or is it AI equals ML? Or is it AI is somehow something different than ML? So here's 3 equations. I wonder which 1 is going to be right.
00:00:28 - 00:00:54
Well, let's talk about this. First of all, when we talk about AI, I think it's important to come with definitions, because a lot of people have different ideas of what this is. So, I'm going to assert the simple definition that AI is basically exceeding or matching the capabilities of a human. So, we're trying to match the intelligence, whatever that means, and capabilities of a human subject. Now, what could that involve?
00:00:54 - 00:01:18
There's a number of different things. For instance, 1 of them is the ability to discover, to find out new information. Another is the ability to infer, to read in information from other sources that maybe has not been explicitly stated. And then also the ability to reason, the ability to figure things out. I put this and this together and I come up with something else.
00:01:18 - 00:01:42
So I'm going to suggest to you this is what AI is and that's the definition we'll use for this discussion. Now what kinds of things then would be involved if we were talking about doing machine learning? Well machine learning I'm going to put that over here, is basically a capability. We'll start with a Venn diagram. Machine learning involves predictions or decisions based on data.
00:01:43 - 00:02:11
Think about this as a very sophisticated form of statistical analysis. It's looking for predictions based upon information that we have. So the more we feed into the system, the more it's able to give us accurate predictions and decisions based upon that data. It's something that learns, that's the L part, rather than having to be programmed. When we program a system, I have to come up with all the code, and if I wanted to do something different, I have to go change the code, and then get a different outcome.
00:02:11 - 00:02:33
In the machine learning situation, what I'm doing could be adjusting some models, but it's different than programming. And mostly it's learning the more data that I give to it. So it's based on large amounts of information. And there's a couple of different fields within, a couple of different types. There is supervised machine learning, And as you might guess, there's an unsupervised machine learning.
00:02:34 - 00:02:59
And the main difference, as the name implies, is 1 has more human oversight, looking at the training of the data, using labels that are superimposed on the data. Unsupervised is kind of able to run more and find things that were not explicitly stated. Okay, so that's machine learning. It turns out that there's a subfield of machine learning that we call deep learning. And what is deep learning?
00:03:00 - 00:03:26
Well, this involves things like neural networks. Neural networks involve nodes and statistical relationships between those nodes to model the way that our minds work. And it's called deep because we're doing multiple layers of those neural networks. Now, the interesting thing about deep learning is we can end up with some very interesting insights, but we might not always be able to tell how the system came up with that. It doesn't always show its work fully.
00:03:26 - 00:03:49
So we can end up with some really interesting information, not know in some cases how reliable that is because we don't know exactly how it was derived. But it's still a very important part of all of this realm that we're dealing with. So those are 2 areas and you can see DL is a subset of ML. But What about artificial intelligence? Where does that fit in the Venn diagram?
00:03:50 - 00:04:09
And I'm going to suggest to you it is the superset of ML, DL, and a bunch of other things. What could the other things be? Well, we can involve things like natural language processing. It could be vision. So we want a system that's able to see.
00:04:09 - 00:04:37
We might even want a system that's able to hear and be able to distinguish what it's hearing and what it's seeing, because after all, humans are able to do that. That's part of what our brains do, is distinguish those kinds of things. It can involve other things like the ability to do text-to-speech. So, if we take written words, concepts, and be able to speak those out, So this first 1 involved being able to see things. This is now being able to speak those things as well.
00:04:37 - 00:05:05
And then other things that humans are able to do naturally that we often take for granted is motion. This is the field of robotics, which is a subset of AI. The ability to just do simple things, like tie our shoes, open and close the door, lift something, walk somewhere. That's all something that would be part of human capabilities and involves certain sorts of perceptions, calculations that we do in our brains that we don't even think about. So, here's what it comes down to.
00:05:05 - 00:05:29
It's a Venn diagram, and we've got machine learning, we've got deep learning, and we've got AI. So, I'm going to suggest to you the right way to think about this is not these equations. Those are not the way to look at it. In fact, what we should think about this as machine learning is a subset of AI. And that's how we need to think about this.
00:05:29 - 00:05:48
When I'm doing machine learning, in fact I am doing AI. When I'm doing these other things, I'm doing AI. But none of them are all of AI, but they're a very important part. Channel, so we can continue to bring you content that matters to you.
Omnivision Solutions Ltd