ChatGPT: A chatty, amazing writer

ChatGPT

Shock and awe were my reactions after my first encounter with ChatGPT.

Words sprouted on the web page as soon as I finished typing my request for a parody of Shakespeare. In a flash, ChatGPT churned out a poem addressing the Bard that went:

“Let’s see, how could we twist and bend your work,

And make the old seem new, the sombre more absurd?

“Perhaps, dear Shakespeare, you’d be a rapper on the scene,

Spitting rhymes and beats, with iambic pentameter lean.

You’d bust a flow about young Hamlet’s troubled mind,

While Ophelia’s love for him you’d eloquently rewind.”

The amazing wordsmith was lightning-fast, writing, at my request, in the style of Shakespeare, Jane Austen, Hemingway and PG Wodehouse.

Imitating Shakespeare, it wrote about fame:

“Fame, thou art a fickle mistress,
A fair-faced lie that doth deceive.
Thou art a fleeting breeze that doth come and go,
A sound that doth echo and then leave.”

The chatbot shared jokes, too, albeit blandly:

“Why was the math book sad?

“Because it had too many problems.”

ChatGPT is not perfect. It can slip up.

Asked to write a sonnet, it wrote a poem, not a sonnet, which is a poem of 14 lines with a specific rhyming scheme.

Also, asked about differences between British and American English, it wrote:


“There are also some differences in the way grammar is used in American English and British English. For example, in American English, the past tense of the verb ‘to get’ is ‘got’, while in British English, it is ‘got’ or ‘gotten’.”

On the contrary, “got” is British and both “got” and “gotten” American.

Launched by the artificial intelligence research laboratory OpenAI, headquartered in San Francisco, ChatGPT seems to have an American orientation.

Asked which sentence is correct — “England are playing France” or “England is playing France” — the chatbot claimed the first is incorrect, the second correct.

But “England is playing France” is American English while “England are playing France” is British English.

Corrected, the chatbot thanked me for pointing out its mistake.

To its credit, when asked, ChatGPT makes a fair assessment of its strengths and weaknesses. It says: “ChatGPT is a very powerful language model that is capable of generating human-like text in real-time. “

It agrees it is somewhat similar to the predictive text we can use to write on mobile phones. But predictive text suggests words and phrases based on what we normally use. It is limited by the user’s vocabulary.

ChatGPT, on the contrary, to quote it, “has been trained on a large dataset of human text” “with billions of parameters”.

“My training dataset includes a wide range of text from a variety of sources, including books, articles, and websites, from many different countries where English is spoken,” it says. “This includes text from the United States, the United Kingdom, Canada, Australia, and New Zealand, as well as text from other countries where English is widely spoken, such as India, Singapore, and Malaysia.”

Indeed, it wrote a bit of Singlish, too, at my request: “Yo, wat’s gucci man? How’s life been treating you lately? I just finished my shift at the hawker center and man, I am starving lah. Thinking of heading to the food court to get some char kway teow. You want to join me or what?

“However, it is important to note that ChatGPT is not a perfect writer,” the chatbot itself says, “and it may make mistakes or produce text that is not entirely coherent or appropriate. It may not always be able to generate text that is as sophisticated or nuanced as that produced by a human.”

Nevertheless, its ability to write essays and code has prompted concern that it could be used for cheating.

Jobs may be under threat, too.

ChatGPT says it could be used in chatbot applications for customer service, e-commerce and entertainment because it could carry on a “virtual conversation”, that is, chat with people.

The economist Paul Krugman in the New York Times wrote that artificial intelligence and automation could reduce the need for some knowledge workers, such as those involved in data analysis, research and report writing.

When I put the question to ChatGPT, it said: “It is possible that language models like me may become more advanced and able to generate text more quickly and accurately than human writers in certain contexts. However, it is also likely that human writers will continue to have a role in creating original content and expressing personal perspectives and experiences.”

Indeed, while I was impressed by the speed, flow, clarity and smoothness of ChatGPT’s prose, I was also struck by the absence of vivid expressions and imagery found in good writers.

It did use memorable words and images when asked to imitate great writers like Shakespeare, when it was quoting them.

But when it was chatting or writing something else at my request, it used conversational language and what may be called good business English – clear and smooth but bland.

That may be because most of the content it has been trained on is written in plain language.

ChatGPT has its limitations. It hasn’t accumulated knowledge beyond 2021. It can’t be used for searching like Google, it says, but answers your questions with instant write-ups that may be like entries in Wikipedia except that, unlike Wikipedia and Google, it offers no links.

Yes, there is concern over the software. But when was an innovation accepted without any hand-wringing?

Even the invention of the alphabet was mourned by the Greek philosopher Plato. He was worried that the use of text would threaten traditional memory-based arts of rhetoric.

This was pointed out in a recent article by the New York Times columnist Zeynep Tufekci.

Just as humanity has evolved from hunter-gatherers to knowledge workers, technology is evolving too, assuming roles played by humans so far.

If man can write, so can software, with a little help from human inventors, of course.

Man and machine have become partners and could get closer still in a future that resembles the Nobel Prize winner Kazuo Ishiguro’s novel, Klara and the Sun, where Klara, a robot, an Artificial Friend, is close to Josie, a young girl, speaking the same language.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

You may also like

Blog at WordPress.com.

%d bloggers like this: