In the past year, our understanding of AI has increased as programs like ChatGPT, DALL-E, and Midjourney have become ubiquitous tools in our daily lives. These programs all fall under the same category of generative AI, a type of AI that excels at synthesizing new text and media content. But AI can do much more than generate content, and it’s worth exploring and familiarizing yourself with the technology’s full range of possibilities.
Before we get into the different types of AI, it helps to define what AI is. People often use AI as an umbrella term to describe many different techniques, says Nitya Mandyam, Codecademy Senior Curriculum Developer. (Fun fact: Along with creating a ton of our data science courses and content, Nitya’s also pursuing a postgraduate degree in AI Ethics.)
“AI is an interesting field because it’s kind of defined by the end goal, which is achieving human or superhuman intelligence, rather than what it actually is,” Nitya says. “Curiously, this lack of a precise universally accepted definition has probably helped the field grow and blossom.” Simply put, AI refers to intentionally constructed systems that actively interact with the world and are usually guided by large data models.
Learn something new for free
Early AI systems were inspired by mathematical logic and sought to mimic conscious thought processes through rules. (For example, in the late 90s, IBM built a chess-playing commputer called Deep Blue that was able to beat a grandmaster in chess.) Today, the many different types of AI that we encounter are much more sophisticated, in part because our understanding of human intelligence is more in-depth, Nitya explains.
While its roots go back decades, modern AI really accelerated in the last few years. As AI advances and seeps into technology we use, the lines between different types of AI can be hard to distinguish. Ahead, Nitya breaks down the various categories of AI that you might come across and key differences that you should know about each.
The different types of AI
When we talk about AI, there are two general categories that systems fall under: goals, which are AI systems designed to achieve a specific outcome; and techniques, methods of teaching a computer to replicate human intelligence, Nitya says. Here are the main types of AI goals and techniques that you’ll come across.
Computer vision
Computer vision (CV) refers to neural-network based algorithms used to classify and generate image data. The overarching goal is to teach computers to “see” the way we do. CV is used in the automotive industry to develop self-driving cars, and it’s also heavily used in medicine to diagnose tumors and help health professionals catch things that the human eye can’t.
Natural language processing
Natural language processing (NLP) refers to neural-network based algorithms used to classify and generate text data. NLP has a wide range of uses, from virtual assistants like Siri and Alexa to large language models like ChatGPT. It’s also heavily used in translation. Check out our Data Scientist: Natural Language Processing Specialist career path if you want to start building your own NLP program.
Machine learning
Machine learning uses algorithms that use large amounts of data and computing power to find patterns in data and perform tasks like prediction, classification, and generation. You likely interact with some form of machine learning every day, like when Spotify or Netflix suggests new content based on your interests.
Generally, there are two types of machine learning: Supervised and unsupervised.
- Supervised learning involves manually telling a program the correct output repeatedly until it learns to do it itself.
- Unsupervised learning involves giving a program a ton of data and allowing it to make its own connections.
You can learn the basics of machine learning algorithms and how to train supervised and unsupervised models in the path Learn Machine Learning.
Neural networks
Neural networks are “information processing units arranged in specific configurations designed to mimic the human brain,” Nitya says. While neural networks serve as the foundation for various AI methods, they require vast amounts of data and computing power and can be very complex, so they’re not necessary for tasks that can be accomplished with other machine learning techniques.
Generative adversarial networks
“Generative Adversarial Network (GANs) basically involve two neural networks duking it out until one convinces the other that stuff it generated (text, images, etc.) is good to go,” Nitya says.
The two neural networks that comprise a GAN have different tasks. One generates new data, while the other evaluates whether or not it fits the prompt. “Let’s say the goal is to generate images of cats,” Nitya says. “The generator wants to do as good a job as possible at generating cats, and the discriminator wants to do as good a job as possible filtering out things that aren’t cats.”
GANs are used heavily in image-based tools like Photoshop and Stable Diffusion.
Deep Learning
Deep learning is an umbrella term that covers many of the other types of AI listed here. “It’s basically machine learning done with neural networks,” Nitya says. Neural networks have multiple layers, and they can be either shallow (with few layers) or deep (with many layers). Deep neural networks require more computational resources but tend to outperform shallow ones.
Reinforcement learning
Reinforcement learning uses a penalty-reward system to perform tasks. “Say you’re building a self-driving scooter and want it to distinguish between the sidewalk and the road,” Nitya says. “You upload images of both and every time it correctly distinguishes between the two, you give it a point, and when it’s incorrect, you reduce a point, and this calibrates the system to do what you want.”
Reinforcement learning is used to develop self-driving cars, and it’s always used in tandem with deep learning algorithms. There’s also reinforcement learning based on human feedback (RLHF), which uses human fact checkers to work with systems and help them respond appropriately to prompts.
Learn more about AI
As you can see, there’s way more to AI than just generative AI. Want to dig in and learn more about the different types of AI and what they can do? Check out our AI courses to explore large language models, prompt engineering, chatbots, and more.
Subscribe for news, tips, and more
