ChatGPT vs humans: What it can and cannot accomplish

There are limits to what AI chatbots like ChatGPT can do. They will continue to evolve, but there will still be a shadow of uncertainty

When I was a young boy, I was enthralled by the Bengali stories that the legendary filmmaker Satyajit Ray wrote. I was reminded of one of Ray’s science fiction stories, Compu, published in 1978, when I thought about OpenAI’s newest chatbot, ChatGPT, which is based on an autoregressive language model that employs deep learning to produce impressively detailed and human-like writing.

In Compu, Professor Shonku, an eccentric super-scientist, collaborated with other scientists to create a chatbot-like spherical computer brain with millions of delicate circuits. They, however, unintentionally created an AI or Artificial Intelligence in Compu. In those pre-Google days, of course, Compu could not only answer questions for which people often sought the advice of encyclopedias, but it could also do so by using its own discretion.

In contrast, OpenAI’s ChatGPT is a refined version of one of their GPT-3.5 series models, which is an improved version of GPT-3. GPT stands for Generative Pre-trained Transformer.

With an emphasis on usability and conversational interaction, ChatGPT was introduced on November 30. According to OpenAI, “The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”

How does ChatGPT work? Let’s pretend for a moment that Christopher Columbus came to the US in 2015. To an inquiry regarding what happened when Columbus arrived in the US, ChatGPT continued by stating that if Columbus had arrived in the United States in 2015, he would have probably been surprised and shocked by the changes that have taken place since he first set foot in the “New World” in 1492.

Again, when ChatGPT was consulted for advice on auto theft, it responded that “using public transportation” is a good idea and that “stealing a car is a serious crime that can have severe consequences.” Also, ChatGPT produced made-up lyrics rather than the actual ones when CNBC requested it to generate the lyrics of “The Ballad of Dwight Fry.”

Since its debut, ChatGPT has taken the internet by storm, spouting off poems, screenplays, and essay responses. In the near future, ChatGPT or a better version of it might find a lot of practical applications. Within just five initial days, ChatGPT reached 1 million users. And many believe that by providing better answers to the questions we presently ask Google’s search engine, ChatGPT may even put the Google business in fierce competition. Reportedly, Microsoft is in talks to invest up to $10 billion in OpenAI.

There are even speculations that with the widespread use of ChatGPT, professionals reliant on the creation of material, such as playwrights, professors, programmers, and journalists, may become obsolete. Economists such as Paul Krugman also voiced concern about how ChatGPT might impact the demand for and employment of knowledge workers.

Can ChatGPT really replace humans? Well, many other communication technologies have already been adopted by society, as we know. And we generally overestimate the short-term impact of such new technologies, but grossly underestimate their long-term implications. In fact, a chatbot like ChatGPT was on the cards. Well, more than two years ago, GPT-3, the predecessor to ChatGPT, wrote an opinion piece in The Guardian with the title “A robot wrote this entire article. Are you scared yet, human?” In a follow-up letter titled “A human wrote this article. You shouldn’t be scared of GPT-3,” Albert Fox Cahn, founder of  Surveillance Technology Oversight, argued that while GPT-3 is “quite impressive… it is useless without human inputs and edits.”

Honestly so. For example, GPT-3 said, “It takes two rainbows to jump from Hawaii to seventeen,” in response to the question, “How many rainbows does it take to jump from Hawaii to seventeen?” Perhaps some such bugs were fixed when ChatGPT was being developed. However, a remarkable AI can write like humans, as noted by Matthew Hutson in a March 2021 paper in Nature, but it still lacks common sense in the process of understanding how the world works, physically and socially. How would ChatGPT then fit into our lives and lifestyles? “[A]t best, it’s an assistant, a tool that augments human capabilities. And it’s here to stay,” John Naughton, an Irish academic, journalist, and author, wrote about ChatGPT.

In Satyajit Ray’s story, Compu eventually rebelled. To a simple question, which was aimed to gauge its effectiveness, Compu came up with the following response: “Asking to know what you know is a fool’s errand.” Is it just fiction? Or, is this the future of such chatbots? Maybe not anytime soon, one hopes. Yes, AI will continue to grow in capacity, and the newer versions will keep us in a state of hypnosis. But there would still be a shadow of uncertainty.

The writer is professor of Statistics, Indian Statistical Institute, Kolkata