ADVERTISEMENT
ChatGPT AI text tool: A fad or promise?Some are using the chatbot to draft emails, blogs, contracts, and codes. Others are scrutinising its limits
Barkha Kumari
DHNS
Last Updated IST

A new Artificial Intelligence (AI) chatbot, ChatGPT has clocked over 1 million users since its launch on November 30, and Bengalureans have given it a go already.

It produces human-like text in seconds in response to prompts fed into a web browser — from poems to essays, emails to codes, film scripts to self-help tips. Reports say it is more conversational and intuitive than existing models as it can answer follow-up queries, give authoritative-sounding answers and turn down inappropriate requests. Currently, it is free to use.

ChatGPT is built by Open AI, the US-based AI research laboratory founded by Elon Musk. GPT expands to Generative Pre-trained Transformer because it has been trained on a vast amount and variety of text.
Bengalureans weigh in on the future of AI writing tools.

ADVERTISEMENT

First rush

Shivani Vijay has used ChatGPT to write “an email to my boss asking for holiday (sic)”, “an apology to a friend who is angry at me”, and a summary of her qualities for a CV. The startup professional hasn’t put the responses to use though.

“I tried it for fun but the responses are accurate and polished,” says the 25-year-old, who may use it “for assistance” going forward.
Naga Subramanya’s first prompt to the chatbot was ‘How do I get my 18-month-old son to eat food by himself?’ “Google search shows a whole lot of information. But ChatGPT provided a checklist of dos and don’ts. It covered all bases in one response,” shares the finance director of a startup. Generating a poem for his wife is on the cards. “The thought counts,” he quips.

But Avik Dey feels it is a herculean task to dislodge Google. “Google gives a plethora of options while the chatbot makes a decision for you, which can be useful or otherwise. For instance, I asked ChatGPT to formulate a mathematical puzzle and answer. The answer was wrong!” says the lead software engineer with a product firm. Plus, he reminds its database is smaller than Google’s, which limits the responses.

Use case

Mahesh N’s travel company has generated 10 blogs using ChatGPT — first versions took 30 minutes. These will go up online in a week after the team manually adds locally-sourced information. “Freshers take a month to produce 10 blogs,” he comments. The chatbot is also yielding “impressive” taglines, and meta descriptions, which required hiring experts earlier. This doesn’t mean job cuts. “The same people have more time to do better work now,” he says.

Narasimha Kaushik’s company runs AI-powered fact-checking in real time. “I have been experimenting with ChatGPT to generate software programmes that do image processing and AI tasks of basic to medium complexity,” he says.

He is excited about generative AI tools which have birthed a new job profile called “prompt engineer — someone who harnesses such tech to get desired results”. “An AI art won a competition recently. A Booker prize for AI-written novels doesn’t sound far-fetched,” he adds.

Rohit Regonayak, who has an ebook distribution company, says, “Publishers are surely concerned about how to tell an AI-generated book from a genuine one. The problem is bigger in the self-publishing segment.” He feels plagiarism may not be an issue because “ChatGPT has been trained to create original content”.

For a project for a US-based information company, he tried to generate data on US legislators in specific tabular formats. The turnaround was immediate — he didn’t have to scour multiple sources, assemble names of the individuals, their party, their tenure, and create a table. “But the results weren’t always up-to-date (as it is trained on pre-recorded data) and tracing the data source is difficult,” he says about the limitations.
Concerns about the ease of generating hate content, sophisticated spams and competent-looking CVs were voiced by all as also the loss of low-level jobs. However, these Bengalureans foresee its adoption in content generation for training manuals, blogs, contracts, press releases, marketing punchlines and software source codes.

ADVERTISEMENT
(Published 09 December 2022, 00:27 IST)