What Is Artificial Intelligence?

Artificial Intelligence has been a topic of conversation in the tech industry for a while now. There’s been a lot of conversation around AI taking people’s jobs, increasing security concerns, and much more. Not to mention, it seems like there’s a TON of new technology coming to the market touting their AI capabilities.

The idea of AI can be traced back to the late 1800s when people explored the ideas around what would be needed to create an artificial human brain. In the 1800s, our understanding of the brain and the necessary technology wasn’t there yet. However, in the 1930s and 40s Alan Turing came up with the core theories around artificial neurons – which laid the foundational theories around AI. Over the years, more folks took a stab at creating AI, but the technology still wasn’t there yet.

Fast forward to the 1990s and early 2000s and finally the technology caught up. Computers could store a lot more data and they were finally able to process that data fast enough. This has allowed folks to finally start to explore the ideas pioneered by Turing and others decades prior.

But what really is AI?

Well, artificial intelligence is defined as intelligence – perceiving, synthesizing, and inferring information – demonstrated by machines. You’re likely already very familiar with AI technologies, as we’ve been using them for decades already.

Some common examples are search engine algorithms, such as Google, Bing, Yahoo. Other examples are speech processing applications such as Siri and Alexa. But the big examples that have been stealing the show are generative AI like Chat GPT, Dall-E, and more.

So how do all of these work?

At the core of AI models, we have something called ‘deep learning.’ Deep learning uses multiple layers of algorithms to extract higher level information from a huge amount of data. From here, these models can ‘learn’ based on more data and mistakes made in the ‘learning’ process. Multiple deep learning models are combined to form a neural network.

An example commonly used here is the idea of feeding a deep learning model a bunch of animal pictures. The models will be able to pull some common items (e.g. Eyes, ears, mouth, shape, color, etc.) The algorithms will identify these features and ‘learn’ that animals with fins are likely to be fish and not likely to be a cat, and animals with wings are birds, etc. However, it may make mistakes when it sees the shape of ears on say a German Shepard and mistakenly classifies it as a cat. More human oversight and feeding the platform more pictures would correct that. As these additional data sources are combined, a neural network can be created to best classify dogs vs cats vs fish vs birds.

From here, we can get into ‘machine learning.’ Machine learning uses algorithms that function on the idea that strategies, algorithms, and inferences that have previously worked will continue to work in the future. This requires a little less data but tends to need more human oversight.

Going back to our animal classification AI example – we’ve trained the algorithms to realize that animals with wings are birds. So, the AI is humming along, under the idea that anything with wings is a bird and gets classified appropriately – until we introduce a butterfly. Butterflies have wings, and the AI would assume that because it has wings that it must be a bird. But we know that isn’t the case. So, a human could come in and say ‘if an animal has wings AND the animal has feathers, then the animal is a bird.’

So, what is ChatGPT and those other platforms making the news lately?

Well, these platforms are generative AI. Their purpose is to generate things for us. The fundamental theories and science behind them is similar to what’s been described before, but the headline grabbing element here is the fact that these tools can create. They’re not answering questions like Siri or Alexa, or suggesting a search result. They’re creating content, based on the prompt from the user, with all the lovely machine learning, artificial neural networks, and deep learning models behind them. That whole concept is pretty awesome, and the fact they’re hitting the masses is a unique challenge to us all – hence the big attention-grabbing headlines.

And there you have it folks!

Deep learning models form the foundations of AI by providing a lot of data and pulling key information from the data. We’ve added multiple models together to form an artificial neural network with areas of focus. And then we’ve added machine learning to make assumptions and inferences from this data. And for the visual folks out there – here’s an image from IBM that illustrates these concepts:

Please keep in mind that this article is SUPER high level. This field of science and technology has been around for a while and it’s evolving rapidly.

Additional Reading: https://www.ibm.com/blog/ai-vs-machine-learning-vs-deep-learning-vs-neural-networks/