The promise and peril of ChatGPT, a remarkably powerful AI chatbot

Analysts and experts predict that Open AI’s new ChatGPT will bring everything from the “death of the school essay” to the dawn of a new era of communication. But what is ChatGPT, and how can it change our lives?

Ask ChatGPT a question and it quickly summarizes an answer in grammatically correct and punctuated paragraph. Within two weeks of its launch on November 30, millions of users tried the large language model artificial intelligence application. In fact, it received so much attention that the system exceeded its user capacity from time to time.

If you ask ChatGPT what it is and how it works, it will tell you: “As a large language model trained by OpenAI, I generate answers to text-based queries based on the large amount of text data I’m trained on. I don’t have the ability to access external information sources or communicate with the Internet, so all the information I provide is derived from the text data I was trained on.”

As remarkable as ChatGPT appears to be, the system does not “think” and is incapable of coming up with original ideas. It works by closely mimicking human language and packing the potential to make writing tasks faster and easier in a way never seen before.

“This particular tool is remarkably great at figuring out what a user wants and putting relevant things in a really logical and clear way, to the point that some might be fooled into thinking it’s sensible,” says Tinglong Dai, a professor of business management at Johns Hopkins Carey Business School and an AI expert. “No other AI has been as impactful as ChatGPT. It really opened the window to the latest developments in large language models with something most people didn’t realize was possible before.”

We have already become familiar with some forms of AI. Roomba can map your home and vacuum your floors. DALL-E will create images from descriptive text. Siri or Alexa can complete a multitude of tasks with a simple voice command. What makes ChatGPT so engaging is the seamless use of human language. While his essays are impressive, Dai said the system is not foolproof.

“It is also very deceptive in that it is unable to tell if what it is writing is accurate,” he explained. “In fact, just based on my own extensive testing, I’ve found that it makes tons of factual errors, but it does so on some sort of confident, authoritative, people.”

Dai also noted that ChatGPT’s responses seem to become increasingly repetitive and even “defensive” when the same question is asked over and over again. ChatGPT’s answers also vary depending on the language used to ask the question, because its answers will reflect the language of the source material from which ChatGPT draws to formulate its answer.

Risks or rewards

Like many tools at our disposal, ChatGPT holds great promise and terrifying potential. Will it replace work or make it even harder for consumers to distinguish fact from fiction?

“This tool could pose a serious challenge to democracy because it means that the cost of creating misinformation will become incredibly low, so that it will be almost impossible for humans to detect content created by AI,” Dai said. “Say that you can even make AI more authentic by inserting typos and other errors and biases that make it seem even more authentic and personal. I think that’s the scariest part.”

At the same time, Dai said, there is likely to be a premium on authentic writing and real thinking, which only humans can provide now.

“Writing is not just writing; good writing reflects good thinking,” Dai said. “By taking a shortcut, I worry that people could lose that really valuable thinking skill.”

Bring AI to the classroom

The Carey Business School is the only major US business school that requires full-time MBA students to study AI.

“We want them to be intimately familiar with AI theory and what underlies all the AI ​​tools on the market,” says Dai, who designed the course. “We want our graduates to know they will not be replaced or fall victim to false promises of AI. Instead, we want students to think critically, and learn to use AI to improve their competitive advantage, and what distinguish them from others—humans and AI agents alike.”

So how can we be sure we know the source of what we read? Dai says disclosure is important, and developers should provide tools for tracking, so they know if they’re reading something a human wrote or not.

“If you can combine a product like ChatGPT with real-time information, or even with data that is not immediately available on the Internet, you can imagine billions of people wanting to use it,” Dai said. “I really think it could be something at least as big as social media, or as big as the smartphone market. Industries need to figure out ways to work with AI in the future. They need to look at being augmented by AI, and not not be replaced. by AI.”