Watching the recent advancements in large learning models like GPT-4 unfold is exhilarating, inspiring, and frankly, a little intimidating. As a developer or code enthusiast, you probably have lots of questions — both practical ones about how to build these large language models, and more existential ones, like what the code-writing chatbots mean for the future of programming careers.
Our brand new free course Intro to ChatGPT will get you closer to some answers. We’ll walk you through what generative artificial intelligence (AI) is, how it can be used (and when it shouldn’t), and ultimately help you understand why this technology is worth learning.
The reality is that the “age of AI” is poised to transform the way we live and work, and aspiring technologists like you need to be prepared to meet the moment. That means learning new AI skills, thinking critically about AI’s ethical implications, and getting hands-on experience interacting with AI tools. Intro to ChatGPT is an excellent way to get started, no matter how much coding experience you have.
Still skeptical about all the AI hype? Here are a few reasons why aspiring developers should know about ChatGPT.
Learn ChatGPT for free
The technology driving the trend
ChatGPT is a consumer-friendly AI chatbot created by the AI research company OpenAI. It’s a type of generative AI that uses algorithms to create new text-based content. You can give ChatGPT a prompt in plain English (like, “What are the most popular coding courses to take on Codecademy?”), and it’ll try to provide you with a fitting, original response. ChatGPT is wildly popular, and hundreds of millions of people worldwide are finding ways to put it to good (and not-so-good) use.
The GPT in ChatGPT stands for generative pre-trained transformer. “Pre-trained” refers to how generative AI models are built with training data, which can consist of various datasets of different websites, images, text, videos, and more, explains Sarai Fernandez, Codecademy Computer Science Domain Manager.
ChatGPT’s ability to accurately interpret our prompts and respond conversationally stems from GPT4 — one of the best large language models (LLMs) to date. LLMs are trained with huge datasets; as you feed a model with training data, it learns to recognize patterns and associations and uses probability to make predictions through unsupervised learning. “When we use ChatGPT, we’re not really getting new data — it’s already there,” Sarai says. “You type something in, and then it goes back to all the stuff it already learned and tries to give you a good response.”
Understanding AI’s limits
More often than not, the answers you get sound correct, relevant, and polished. But ChatGPT is far from perfect, and is prone to errors or nonsensical answers.
One major consideration is that ChatGPT can’t differentiate accuracy from truth, Sarai says. Broadly speaking, ChatGPT is making an educated guess about what you want to know based on its training, without providing context like a human might. “It can tell when things are likely related; but it’s not a person that can say something like, ‘These things are often correlated, but that doesn’t mean that it’s true.’” Put another way, ChatGPT is very good at generating language, but that’s not the same thing as having original, intelligent, factual thoughts.
In our course Intro to ChatGPT, you’ll learn about the risks and limitations of AI, including the real-life consequences of using biased or outdated training data.
Why programmers should learn about AI
Many programmers are using ChatGPT and other code-writing AI tools as part of their programming workflow so they can get more done. In fact, when GitHub surveyed developers who use its AI tool Copilot, they found that devs were more productive, completed repetitive tasks faster, and were able to focus on more satisfying work.
ChatGPT can be really helpful with simple or tedious tasks, Sarai says, and it can also come in handy when you’re stuck on a problem or forget something important. For instance, say you needed to know how to create a random number generator in Python. Instead of digging through Google and Stack Overflow, you could ask ChatGPT for a code example and have reference material in seconds.
But to make the most out of ChatGPT, you’ll need to know how to communicate with it properly. AI isn’t perfect, and there’s a logic to writing prompts (which you’ll learn in our Intro to ChatGPT) that’ll help ChatGPT provide you with the best response. In fact, there’s actually a career — prompt engineering — that’s centered around writing specific prompts to help test and develop an AI’s capabilities.
The code ChatGPT writes might require some editing, and the chatbot is far from coding entire apps on its own. “There would definitely be places where things would go wrong,” Sarai says. “Is its code secure? Does it know that this library is deprecated? How does it handle user authentication and authorization?” It’s up to human developers like you to know what code is supposed to look like, how a program is supposed to work, where bugs came from, and what it takes to fix them, she adds. Doing the dirty work and figuring things out yourself is all part of the learning process.
“My grandpa would say this: You need to be smarter than the tools you work with,” Sarai says. “So you need to be a better programmer than ChatGPT is.”
You need a strong understanding of the tools you work with to be a good programmer, so check out our free course Intro to ChatGPT to learn more about what it can do, then continue sharpening your AI skills with the courses below. (And sign up quick — they’re only free for a limited time!)
- Intro to Machine Learning
- Intro to Cloud Computing
- Learn Linear Regression with R
- Learn Text Generation