The Future is now: Intro to AI for data analytics

- [Instructor] Before we dive into specific AI tools like ChatGPT and Google Bard, it's worth taking a step back and understanding the broader AI landscape. Artificial intelligence is a very broad term that relates to machines and computer systems that mimic functions associated with human intelligence like decision-making, image recognition, self-driving cars, and more. Within the field of artificial intelligence lies machine learning. These are going to be the models that act as the brains of artificial intelligence systems. They help computers learn with minimal human instruction and get more accurate when exposed to more data. And that more data piece is critical in explaining why we've seen an explosion in AI and machine learning in the past few decades. With the proliferation of the internet, mobile devices, internet of things, we're now generating more data than we ever have and that's enabled machine learning to become more accurate than it ever could have been prior to some of these inventions. And within machine learning lies deep learning. This is a very complex family of algorithms that's designed to mimic the human brain and learn almost exclusively without human intervention. Large language models like ChatGPT and Google Bard fall within this category. And speaking of the human brain, there's still one area where these models are not able to approach what humans are able to do and that brings up the conversation between weak versus strong AI. Currently, we're in this state of weak AI, which is defined as AI that can only perform specific tasks. So for example, ChatGPT could tell you how to drive a car if you asked it, but if you asked it to drive your car for you, it couldn't do that even though it surfaced all of this knowledge that you'd need to drive a car. Strong AI, or artificial general intelligence, would be able to do that at some point in the future. This is often what's depicted in science fiction and often what is very scary to many folks about this field. These systems would be able to learn and perform any task that a human could do. So if you could ask ChatGPT, "Can you go buy me groceries today?" And it was able to solve all of the problems that it needed to without your intervention, that would be much closer to strong AI than we are today. As of now, we haven't seen any strong AI systems implemented, but this is likely on the horizon and possibly within our lifetimes. And just in case you were wondering, both machine learning and deep learning have their roots in statistics. A lot of the algorithms rely on statistical concepts, but there's one major philosophical difference between these fields. Statistics is very much concerned with being able to confidently say, yes, there is a relationship between variable A and B and that relationship is X. Machine learning and deep learning don't care about those relationships, they just care about the accuracy of output. And because they don't care about these relationships and they certainly don't care about humans understanding these relationships, they're able to be much more complex than the same statistical algorithms that they might be based off of. This has led them to be called black box models because we know what goes in and we know what comes out, but we have no idea what goes on inside these algorithms. They're often picking up millions or billions of nuances and data that we really can't understand or detect. But now let's take a look at generative AI specifically.