Build a system that generates something new

- The last few years have seen a big rise in interest around artificial intelligence, and it's not because your Netflix account is better at making recommendations. It's not that you noticed your package arrived a few days earlier. This type of artificial intelligence has been working for years. This increase in interest is because of something called generative AI. Up until the last few years, most of the work in artificial intelligence has been in what's called predictive AI. This is when the system looks for patterns. Then it tries to do a better job predicting the outcome. This could be a facial recognition system that predicts a person identity. It could be a retail website that predicts customer purchases. It could even be a reinforcement learning system that predicts the next move in an ancient game. These predictive systems have been the quiet engine that's built up many today's largest companies. LinkedIn got its start by predicting your business context. Google has become very good at predicting which website closely matches your search. Facebook can predict which news stories you'll find the most interesting, but newer generative AI systems are no longer just predicting your behavior. Instead, they're generating something new. These systems can produce very natural sounding text. They can generate lifelike images and videos. This puts artificial intelligence in a whole new context. It's now doing tasks which were normally reserved for humans. It's creating articles and even writing poetry. It's not just boxed into the back room predicting what you'll buy. It's suddenly a creative partner that can write, draw, chat, and research. That's why for many people, it looked like AI went from serving us to competing with us. It's important to remember that even though these new capabilities are impressive, they still use most of the same technology. These new generative AI systems still use deep learning artificial neural networks. They still use machine learning to identify key patterns. They're building on technology that's been around for decades. That's why it's important to learn about the important steps that led to generative AI. One of the most important ingredients, this new technology is access to massive amounts of data. Companies like OpenAI have looked through trillions of words and billions of images. This allowed systems like ChatGPT to have what seems like a very human conversation. Once it sees patterns, it can come up with something entirely new. But to get there, the system needed to move beyond just supervised and unsupervised learning. It can't rely on predictive data models that have been trained for just one task. Newer generative AI models will become some of the most valuable properties and technology. In the next few years, you're even going to see lawsuits and new regulations. These generative AI systems will shake up ideas about who owns the knowledge that went into these models. So even though these new systems might seem like they're coming out of nowhere, there's still an extension of technology that you've already seen. It's a significant advancement, but not entirely new. But this small step has been a huge leap in AI capability.