A newly released artificial intelligence chatbot has gained traction this week, with its ability to answer back in a conversational style.

ChatGPT by OpenAI is a new language model trained on vast amounts of data from the internet written by humans, including conversations, so the responses it provides sound human-like.

Users on Twitter and TikTok have been sharing their interactions with the algorithm, as they get it to generate movie scripts, wedding vows and recipes.

However, there's concern from New Zealand's teachers that the software could be used for cheating.

Philippa Wintle from the New Zealand Association for the Teaching of English told 1News she and her colleagues had been having discussions around the future of this technology.

"I'd be one of many teachers who has come up against this technology recently or introduced to it recently, and you think 'this could be amazing in my teaching practice, but holy heck this could be terrifying also'."

1News asked the software to write a high-school level essay on the novel To Kill A Mockingbird, which is a standard in NCEA Level 1.

ChatGPT provided an essay, with quotes from the text, within a few seconds.

After grading the paper, Wintle said "it's pretty good, which is scary, but really exciting when you think about how we can use this".

She said at NCEA Level 1 it would be given a merit grade "because of the integration of examples, the moderate unpacking of examples, because of the convincing vocabulary".

Wintle said it's inevitable the technology will be misused by students.

"It's not even plagiarising, because it's original ideas, but we can't ignore the fact this is happening.

"We have to understand it, and we have to integrate it into our programmes to make our programmes more exciting and relevant, and to enhance the good work we're already doing."

Andrew Chen is an expert in AI and a research fellow at the University of Auckland.

He said this is an example of a new piece of AI that's easily accessible for the public.

"With this new system, it's really giving these new text responses that have probably never been seen before," he said.

"So, it's writing new sentences that have never been written by a human and doing so in a very convincing way."

He said it's possible for programmers to create another piece of artificial intelligence, that can be used to recognise text written by a computer, but warned it was an "arms race".

"At some point, we will have AI systems that produce text that are so convincing, produce images that are so convincing, that no human or no other AI will be able to tell that it's not real."

Chen said while it's easy to look at this technology as a negative advancement, he said there are some useful applications.

"One that I'm quite interested in is that, for example, people who have dyslexia who might otherwise struggle to communicate in written language, they might be able to use an AI like this, to give it an input prompt in a way that makes sense to them, and then have something that's produced for them that might meet the standards of other people."

However, Chen warns that algorithms can still be untrustworthy, especially if they've been trained on data that is biased.

"The current AI systems that we have, they produce very convincing text, but they're still not 100% correct, and there are definitely gaps and how it applies its knowledge."

OpenAI said ChatGPT is not connected to the internet and admits "it can occasionally produce incorrect answers".

"It has limited knowledge of world events after 2021 and may also occasionally produce harmful instructions or biased content."

1News has approached OpenAI for this story.