Here’s everything you need to know about ChatGPT
Lesson summary
Hi there everyone, I’m Jeff and this is Plain English, where we help you upgrade your English with current events and trending topics. JR is the producer. He has uploaded the full lesson, including the free transcript, to PlainEnglish.com/574.
Coming up today: ChatGPT is an artificial intelligence app that was released to the world in November 2022. By February 2023, it had one hundred million users, making it the most quickly-adopted new technology in history, by some measures. But if a hundred million people are using it, that means 7.9 billion people are not—so if you’re not yet using ChatGPT, that’s okay, you’re in good company. On today’s lesson, we’ll talk all about this new AI sensation.
In the second half of the lesson, I’ll show you how to use the English expression hit home. And we have a quote of the week.
Now, at the end of the lesson, I’m going to tell you about a really fun new activity that you can do together with me and other Plain English listeners—and it’s totally free. But before I tell you what it is…it’s time for you to learn about ChatGPT.
ChatGPT and the next step in artificial intelligence
How do you make a delicious carrot cake? Is the tone of this e-mail too formal? What caused the fall of the Roman Empire? Can you please read me a bedtime story? I generated this computer code, but it’s not working. What could be wrong?
You can ask any of these questions—really any question you want—and a new tool called ChatGPT will answer you, right in your web browser. The answers are cogent: they sound natural, as if they were written by a knowledgeable person, just for you. It’s like having a research assistant, a conversation partner, or even a therapist, all for free.
Welcome to the future. We’ve been hearing about artificial intelligence for a long time. But now, for over a hundred million users, AI has hit home . ChatGPT came out on November 30, 2022. Within two months, it reached one hundred million active monthly users. UBS, a bank, said in a research note that this was the “fastest-growing consumer application in history.” Surely, many, many more have joined since then.
ChatGPT may be new, but the technology that drives it has been in development for a long time. So let’s take a step back and explore how a tool like ChatGPT came into being.
ChatGPT is made by OpenAI, a research lab based in California. OpenAI has two arms: a non-profit research laboratory, and a for-profit arm that seeks to commercialize its innovations.
GPT stands for Generative Pre-trained Transformer. Don’t worry about the name. It’s a type of large language model, or LLM. An LLM uses deep learning techniques to understand and generate language. The models are trained on billions of inputs: GPT used books, articles, and all the content available on the internet as training.
The model then predicts what should come next in a sentence or a conversation. It analyzes the prompt and, based on its learning from the massive amount of input it has received, it predicts what the answer should be.
Because its input is so vast, GPT understands language. “Understands” might not be the right word; let’s say, it can process grammar. If a user asks, for example, “What is the capital of Canada?” GPT would recognize “What is” and it would recognize that it had been asked a question.
GPT looks into the language and determines what knowledge you really want. So for example, you can ask, “What is the capital of Canada?” or you can ask “What is Canada’s capital?” or you can say, “Tell me the capital of Canada.” An LLM analyzes language, so it knows the information you’re looking for, no matter how you phrase it.
Once it recognizes that it has been asked a question, it knows to search its massive training set for information that will satisfy the question. In all of the billions of books and web pages, it knows to look for text that sounds like this: “[Something] is the capital of Canada.” This is an easy one: any LLM should produce Ottawa in response to that prompt.
Another key feature of LLMs is that they remember the context of what came before. So if you ask “What is the capital of Canada?” it answers “Ottawa.” If you then ask, “What about Quebec?” it remembers what you asked before. It knows you asked for the capital of Canada, so if you say “What about Quebec?” it knows you’re asking for the capital of Quebec. And it answers correctly, “Quebec City.”
Large Language Models have been around for a while. But there are two innovations that are making them popular right now. The first is simply computing power. For an LLM to be useful, it needs a massive amount of input, training data. An LLM based on a small data set is not very effective. And analyzing that much data is not cheap. A new generation of computer chips has finally allowed large models to process the massive inputs needed to make them effective.
Second, OpenAI, Google, and others are finally building sleek interfaces that allow consumers and companies to use their models. Before, these models existed in research centers and in companies with a big budget for advanced technology. But now, consumers can use these models directly. And even small companies are finding ways to let their customers access the models in unique ways.
ChatGPT, for example, is just OpenAI’s window into the powerful engine called GPT. It’s a way for ordinary consumers like me and you to use the power of the language model called GPT. And OpenAI is licensing its model to other companies. And those companies are adapting GPT to their own uses.
LinkedIn, for example, uses GPT to suggest improvements to members’ profiles. Got a long conversation in Slack, the workplace messaging platform? Use Slack’s built-in GPT function to summarize the conversation for you. Grammarly, Snapchat, Shopify, Duolingo, Khan Academy, and others are all using GPT to power their own language-based artificial intelligence tools.
LLMs like ChatGPT are not perfect. They don’t “know” anything the way humans know things. And they don’t research answers for you. They can only predict text. And sometimes they get it wrong.
In February, Google released its ChatGPT competitor, called Bard. In its very first public demonstration, Bard answers the question, “What new discoveries from the James Webb Space Telescope can I tell my 9 year old about?” Bard helpfully provides three bullet point answers. One of the answers was this: The Webb telescope “took the very first pictures of a planet outside of our own solar system.” That’s fascinating.
But that was not correct. The first pictures of planets outside our solar system were taken in 2004, long before the Webb telescope was launched. This is not an isolated incident , either. ChatGPT frequently makes up facts, gets facts wrong, and even says inappropriate things. It told one reporter he should divorce his wife. Search Google for “ChatGPT mistakes” and you’ll get a sense of how big the problem really is.
So I would not recommend you use ChatGPT to write a research paper for you or to give you actionable relationship advice. But just because it sometimes makes factual mistakes doesn’t mean it’s useless. In fact, there are a lot of great—safe—ways to use ChatGPT in your life. On Thursday’s lesson, we’ll talk about how you can use ChatGPT safely in your daily life.
Sign up for the Plain English ChatGPT Challenge
I was going to wait until the very end of the lesson, until after the expression, but I just can’t hold it in any longer. So here’s the announcement: You are invited to take a five-day Plain English ChatGPT challenge , hosted by JR and me.
The challenge goes like this. It’s five days of activities and the activities have been created by me just for you as an English learner. You will get a video and some activities to do every day of the challenge; it’s done by e-mail and on our website, PlainEnglish.com. And by the end of the challenge, you’re going to be comfortable using ChatGPT. Because every day of the challenge, you will be in there using it doing something new and different. And you’ll know exactly what to do, and how to do it, because we will tell you.
It doesn’t matter if you’ve never opened ChatGPT before: we’ll show you how to sign up, we’ll show you what to do, we’ll show you how to use it. And every day of the five day challenge, you’ll get some new activities that will help you learn about ChatGPT and help you use it in your daily life.
And the best part is, you’ll be doing this alongside and with other Plain English listeners. And you’ll be invited to a private WhatsApp group, hosted by me, where you’ll be able to ask questions, chat, and share your experiences with me and with other listeners.
We are going to do the whole thing in English, but—just like this podcast—it’s going to be about real life. We’re not going to sit there asking ChatGPT about boring grammar. We’re going to be using it in real-life situations and examples.
But it’s me—and you know me, and you know that we’ll make the videos and the prompts easy to understand for you, as an English learner. The beauty of this is, you are going to improve your English just by taking this challenge and exploring this tool in your second language.
So here are a few details. The challenge is going to start on June 5. That’s a Monday; it’s two weeks from today. And it’s going to go every day of the week, Monday, Tuesday, Wednesday, Thursday, Friday. And the activities will be about 25 minutes a day—some of you will do it faster, but I think 25 minutes is a good amount of time to dedicate to this.
So. You’re dying to know. How do you sign up for the Plain English ChatGPT Challenge? Just go to PlainEnglish.com/GPT. How easy is that? PlainEnglish.com/GPT . If you’re listening on a podcast player or Spotify, just look in the episode notes for this episode: you’ll see it there. There’s a sign up link right in the transcript, too. So there’s no excuse for not doing it—PlainEnglish.com/GPT.
Learn English the way it’s really spoken
