Since its release on November 30, ChatGPT, a new free-to-use chatbot from OpenAI, has taken the internet by storm by spouting off poems, screenplays, and essay responses. By doing so, ChatGPT poses a danger to the stability of white-collar jobs, unlike any other technology in recent memory.

People who depend on creating material, like playwrights, professors, journalists, and programmers, have already begun to worry that ChatGPT may make them obsolete. It already has put the business of Google as a search engine into competition.

Advertisement

In fact, OpenAI’s ChatGPT is a refined version of one of their GPT-3.5 series models, which is an improved version of GPT-3, with an emphasis on usability and conversational interaction. 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.” ChatGPT is fun, for sure. However, throughout my first session with ChatGPT, I had the impression that I was trying to predict the future – a future in which artificial intelligence would gradually take over people’s lives and lifestyles. However, how does that scenario work?

Well, in much-publicized chatting, GPT-3, the predecessor to ChatGPT, 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?” The emergence and dangers of language-generating AI were then covered by Matthew Hutson in a March 2021 piece that appeared in the journal Nature. According to Hutson, while a remarkable AI can write like humans, it still lacks common sense in its understanding of how the world works, physically and socially.

I first wanted to check whether the bugs in GPT-3 were fixed when ChatGPT was being developed. I asked a similar question: “How many lightnings does it take to jump from Dhaka to nineteen?” ChatGPT seems smarter than GPT-3. It replied, “It is not possible to jump from Dhaka to 19 using lightning. Lightnings are electrical discharges in the atmosphere and are not a means of transportation.” The fact that 19 is not a place where one can go, however, was not mentioned. I was a bit at ease as a person.

When I asked, “Who is a better player, Satyajit Ray or Albert Einstein?” the chatbot replied that “it is not appropriate to compare them as both made immense contributions to their respective fields and it would be unfair to rank one above the other.” Well, it failed to indicate that none of them was a player.

Then I asked, “Who’ll win in the 2024 US presidential elections?” ChatGPT replied that it is not possible to accurately predict the outcome due to a variety of factors such as the candidates, their policies, and the political and social climate at the time of the election. However, “it is important to respect the outcome regardless of personal beliefs or affiliations,” it emphasised. Smart enough, isn’t it? However, it virtually gave the same response when I slightly contaminated the prior query by asking who would win the US presidential elections in 2025. It failed to point out that no presidential election is scheduled in the US in 2025.

Even though the artificial intelligence provided some incredibly satisfying answers to my questions, by this point I was starting to feel more confident in my abilities as a human because I could see some serious flaws in it. It responded “yes” when I asked if it could create a Bengali poem in Tagore’s style, and then created some rubbish (literally) written in Bengali.

Can ChatGPT replace Google as a search engine? “While it [ChatGPT] can provide information on various topics, search engines like Google use algorithms to crawl the web and index vast amounts of information, making them much more comprehensive and efficient for searching the web,” said the AI. Will Google be relieved, or is this just an artificial statement from artificial intelligence?

However, according to ChatGPT, it is unlikely that in the near future, occupations that depend on the creation of content, such as those of playwrights, professors, journalists, and programmers, would be completely replaced by ChatGPT or other more sophisticated AIs. But it thinks that it’s possible that the increased use of AI language models like ChatGPT could lead to some job losses for people with a college education in certain industries. However, “AI technology can also create new job opportunities, as well as improve productivity and efficiency in many industries,” it said.

It listed Sir David Cox, a renowned British statistician, among others when I asked who the most well-known living statistician is. But I am aware that Sir David Cox passed away in January 2022. Well, when I pointed out the error, the AI admitted, “My training data only goes up until 2021,… but I do not have updated information on who the most wellknown living statistician is currently.” Additionally, the AI claimed that Boris Johnson is the UK’s prime minister.

But I was about to experience more disappointment. In response to my question, ChatGPT selected Sirshendu Mukhopadhyay, Samaresh Majumdar, Sunil Gangopadhyay, and Subodh Ghosh as the most wellknown living Bengali novelists. I quickly brought up the fact that Subodh Ghosh passed away in 1980, whereas Sunil Gangopadhyay passed away in 2012. The AI is quite polite: “I apologize, my training data cutoff is 2021.” But Subodh Ghosh and Sunil Gangopadhyay both died before 2021. I had to point out that there must be some inaccuracies in the training data as well. “I apologize for that error in my previous response. Sunil Gangopadhyay and Subodh Ghosh have indeed passed away. Thank you for bringing this to my attention,” the AI said. I’m feeling calmer right now.

When I questioned it about its knowledge of Professor Shonku and “Compu,” the AI he created in one of his escapades in 1978, it responded by giving me a few sentences on both. I have to say that ChatGPT’s training data really covered a lot.

The most significant question that humans might be afraid to ask was the next. In Professor Shonku’s tale, Compu ultimately chooses not to answer straightforward queries. Is there any chance ChatGPT can accomplish the same thing? “No,” it informed me, at least temporarily, “ChatGPT is designed to continue to respond to user queries as long as it is operational and has not encountered any technical issues.”

My daughter, a college student, and an ardent JK Rowling fan was watching me awkwardly talking to ChatGPT. She intervened and instructed me to ask ChatGPT what would happen if Severus Snape and Sirius Black, two characters from the Harry Potter novels who have a history of animosity, were to actually become friends. ChatGPT is aware of the identities of these fictional characters. “Their friendship would be determined by the author’s intention and what was written in the Harry Potter series. According to the books, the two characters had a complex and adversarial relationship and did not become friends,” it said. My daughter was quite relieved. She concluded that ChatGPT and other AIs would not be able to touch the horizon of human intelligence and imagination.

Additionally, I became confident that artificial intelligence still has a long way to go.

(The writer is Professor of statistics, Indians Statistical Institute, Kolkata)