SCMP: So fast, so good: Hong Kong users of ChatGPT are bowled over, but what about accuracy, ethics, integrity? (Part1 of 2)

[Hamartia Antidote]

Hong Kong users say ChatGPT is so fast, so good, but what about accuracy, ethics?

As artificial intelligence-driven tech tool disrupts the way things work, some worry about the downside.

www.scmp.com

• As artificial intelligence-driven tech tool disrupts the way things work, some worry about the downside
• Some experts say chatbots like ChatGPT will continue to evolve and become even more like humans

Illustration: Brian Wang

In the first of a two-part series, Oscar Liu and Cannix Yau highlight concerns being raised as various sectors in Hong Kong are disrupted by ChatGPT, the artificial intelligence tool that has taken off globally.

The growing popularity of ChatGPT and other artificial intelligence (AI) tools in Hong Kong has sparked a debate over their usefulness and the challenges they pose to human endeavour.

Businessman Keith Li King-wah, 49, said he could not believe how quickly ChatGPT drafted a proposal for a set of technology-related workshops aimed at students.
He fed the chatbot some relevant details from Chief Executive John Lee Ka-chiu’s first policy address and Financial Secretary Paul Chan Mo-po’s annual budget speech, and it responded right away.

“It literally took a minute to give me a comprehensive proposal with a programme name and objective, recommended activities including Web3 workshops, technology competitions, and sections for guest speakers and field trips,” the chief executive officer of Innopage, a mobile app developer, said.
“Although it was just a framework, people could easily tweak it to make it viable.”

ChatGPT is changing the way businesses operate. Photo: Elson Li

His company has embraced ChatGPT, which attracted 100 million users worldwide within two months of being unveiled last November by Microsoft-backed OpenAI, making it the fastest-growing consumer app in history.

“We also use it in our marketing meetings where my colleagues and I simply bring our laptops and ask ChatGPT for ideas for promotion activities, social media content and advertising strategies,” said Li, who is also chairman of the Hong Kong Wireless Technology Industry Association.

ChatGPT, which is capable of providing humanlike responses, is programmed to draw mainly on information from the internet and other sources.

It is, however, at present limited to material available up to 2021 and it can also make mistakes in responses to queries.

Users marvel at its ability to perform a wide range of tasks, from answering questions to writing essays and offering solutions to problems, but the tech tool has also raised questions of integrity, ethics and cheating.

The Hong Kong government has signalled it will set up a task force to look at new tech tools and consider if they should be regulated through legislation.

The city’s technology minister, Sun Dong, has also said Hong Kong planned to develop its own AI-based chatbot service to keep up with the global innovation race and would consult the public this year on the establishment of an AI supercomputing hub to attract top talent and technology companies.

Eason Lai, a global technology strategist at Microsoft in Hong Kong, said he believed the city government wanted to understand and research more about GPT technology and come up with guides to show individuals and businesses when and how to use AI the right way.

“We cannot prevent new technological innovation from emerging,” he told the Post. “Having rules to prevent or even block it could cause Hong Kong to lose out to competitors, especially as the government is positioning it as a hi-tech, innovative city.”

‘Observe impact before introducing laws’​

Ronick Chan Chun-ying, a finance sector lawmaker, said AI-based tools played an important role in the industry’s financial technology development.

“AI-based technology has been used to provide different banking services such as replying to customer inquiries and smart services for automatic teller machines,” he said. “From the industry’s perspective, I don’t see any disadvantages of using ChatGPT or how it will be a threat”

Chan added he saw no need to regulate the use of ChatGPT in banking.

“I hope that the government can give the financial industry more room to apply ChatGPT technology so we can produce quicker and more precise responses to customers,” he said. “It could also save operating costs by relieving the strain on the banking industry’s manpower.”

The Hong Kong Institute of Bankers, a non-profit institution that offers training and certification services to the sector, said it had not adopted ChatGPT yet as it was still studying the impact of its application.

A banking insider who asked not to be named, appealed for caution in the use of ChatGPT at present as it might not always produce credible and correct answers to customers’ questions.

“But I believe that similar technology will be adopted increasingly in the industry to provide fast, sharp and accurate responses,” he added.

Information technology sector lawmaker Duncan Chiu said the authorities should watch the development of ChatGPT before it considered legislation.

“We need to maintain the most cutting-edge understanding of technology and cultivate top researchers in Hong Kong to keep up with technological development. We should not do something prematurely,” he said.

It is still unclear how ChatGPT will affect jobs and careers. Photo: Jonathan Wong

Concerns about intellectual property​

Chan King-cheung, a professor of practice in media ethics at Baptist University’s department of journalism, said he did not think ChatGPT was a big threat to journalists.

“Reporters still need to be present on site to cover events in person before writing a news story based on their observations and interviews and machines are unable to know what happened at the scene,” he said.

But he warned AI raised concerns about plagiarism and said that businesses that published intellectual property needed time to grasp its impact.

“When it comes to publishing content, transparency is most crucial. Authors can indicate AI-generated content with appropriate attribution and not treat it as their own work. That’s the basic integrity and moral conduct of authors,” he said.

Some experts said AI tools such as ChatGPT would continue to evolve and become even more like humans.

Cao Jiannong, the dean of Polytechnic University’s graduate school and professor of its department of computing, said it was inevitable that workflows would change with greater integration of humans and AI.

“ChatGPT cannot do reasoning so far and may not have the necessary emotions to handle people’s problems,” he said.

But he added that he expected advances in “humanising” the AI tools.
“By incorporating human values into something like ChatGPT, it can have empathy to figure out what is wrong and what is right.”

Professor Martin Wong Ding-fat, dean of the faculty of engineering at Chinese University (CUHK), said there were still many controversies about ChatGPT, the most important being who owned the copyright of content it generated.

“There are also issues of privacy, power balance and monopolies,” he said.
Wong said ChatGPT was one of the largest, most powerful language processing AI models created at a high computing cost and only a few companies and governments could afford to support the system.

Early adopters of ChatGPT in Hong Kong said they were surprised and pleased that the AI tool worked for them and could see its potential in their sectors.

Trainee lawyer James Lee* was won over the day he fed some details into ChatGPT and asked for a 500-word legal document.

“I was amazed because it was a well-written letter in just a few seconds. It would have taken me 30 minutes to write from scratch,” the 24-year-old graduate of the University of Hong Kong said.

Entrepreneur Simon Wong Kit-lung, the chairman of catering company LH Group, started using ChatGPT to reply to questions from customers on Facebook.

“At first, when I asked ChatGPT who Simon Wong was, it gave some wrong answers. Now I give it more precise requests and use it as my social media manager,” he said

If someone asked how to get a job with his group, he said, he would ask the chatbot to craft a suitable reply.

“But I still need to be the gatekeeper, because I need to tweak what ChatGPT generates and I am the one who presses the ‘send’ button,” he added.

Eric Hui, a governing committee member of the Hong Kong Federation of Insurers, said ChatGPT was not yet used by the city industry, but companies overseas were using it for correspondence, marketing materials, business analyses and issuing product development advice.

“Many overseas multinational firms have been trying out ChatGPT as it can improve efficiency,” he said. “But users should bear in mind that its answers are not 100 per cent accurate.”

*Name changed at interviewee’s request.

 

[Hamartia Antidote]

Part2

Who needs a teacher? I’ve got ChatGPT! Hong Kong educationists worry about future

Experts are divided on merits of tech tool in education, with concerns about cheating and plagiarism.

www.scmp.com

Who needs a teacher? As ChatGPT takes off in Hong Kong, educationists worry about impact on teaching, learning​

• Experts are divided on merits of tech tool in education, with concerns about cheating and plagiarism
• ChatGPT is shaking up the education scene and experts say everyone must just come to terms with it

Illustration: Henry Wong

In the final of a two-part series, Oscar Liu and Cannix Yau track the impact of artificial intelligence tool ChatGPT on education and the questions being raised
Hong Kong teenager Timothy Lee Chi-chung went online to get some help with his schoolwork.

He asked ChatGPT, the artificial intelligence (AI) chatbot that has taken the world by storm, for an English grammar quiz set at the Diploma of Secondary Education (DSE) level.

In just two seconds, he was looking at a list of 19 multiple-choice questions.

“I wrote down my answers and asked ChatGPT to tell me how many I got right,” said the 16-year-old, who will sit the DSE examinations next year.

How ChatGPT will affect education has some people concerned and others excited. Photo: Dickson Lee

“It’s good enough that I have these free exercises in my pocket, it saves my parents a lot of money buying me supplementary exercise books.”

Unveiled last November by Microsoft-backed OpenAI, ChatGPT attracted 100 million users worldwide within two months, making it the fastest-growing consumer application in history.

Its most ardent fans insist it can do everything – answer questions on practically any subject, write essays, compose poetry, tackle mathematical problems, resolve computer coding issues, and even provide speedy advice for personal problems.

Tech billionaire and Twitter owner Elon Musk greeted the news of ChatGPT’s arrival by tweeting: “It’s a new world. Goodbye homework!”

That highlighted the AI tool’s potential impact on education at every level, providing students an electronic helper to do their assignments, write school essays and churn out university research papers and theses in a flash.

Academics who tested the tool said its responses to exam questions would get full marks if submitted by undergraduates.

But others have pointed out its limitations too.
The GPT in its name stands for “Generative Pre-Trained Transformer”, which means that despite its humanlike responses, the tech tool is programmed to draw mainly on information from the internet and other sources, currently limited to material available up to 2021.

Students could use ChatGPT to prepare for the Diploma of Secondary Education English exam. Photo: HKEAA

ChatGPT can get things wrong, and owns up when errors are pointed out.

In Hong Kong, students, teachers and university professors have responded to the phenomenon with a mix of enthusiasm, admiration, scepticism and concern. Experts are divided, weighing its benefits as a new educational tool against its drawbacks, especially the threat of cheating and plagiarism.

Secondary school English language teacher Michelle Chan Yuk-fan said her biggest fear was that students would find ChatGPT so easy to use that they would switch off in the classroom.

“Reading and writing are essential in learning English. I’m worried that students will use AI tools to submit their homework and neglect their own growth and improvement academically,” said Chan, who teaches students in forms four to six.

She said that while her school came to grips with the rising popularity of ChatGPT, she would try as much as possible to give her students writing assignments that have to be completed in class.

“This way, I can ensure their writing is based on their knowledge and creativity and can step in if they encounter difficulties,” she said.

‘If you can’t beat them, join them’​

Among Hong Kong’s institutions of higher learning, the University of Hong Kong (HKU) and Baptist University (HKBU) said they would consider it plagiarism if students used AI tools for their coursework and presented it as original work.

Chinese University (CUHK) said it would allow students to use AI tools “under authorisation” for their work, but they faced expulsion if caught using them improperly.

However, Hong Kong University of Science and Technology (HKUST) has gone all out to embrace ChatGPT, giving staff free rein to create their own guidelines for its use.

Professor Wang Yang, vice-president for institutional advancement, said banning ChatGPT in classrooms would be futile.

“We should explore ways to incorporate these technologies into the teaching and learning process by providing personalised and interactive learning content,” he said.

“If you can’t beat them, join them. These tools have the potential to enhance the educational experience significantly.”

He said people should find ways to adapt to the change by harnessing the potential of new AI tools while addressing their limitations.

HKUST provost Professor Guo Yike said that in the context of education, the essence of ChatGPT was whether students knew how to ask questions or give instructions to generate the best answers, rather than how AI could help them do their assignments.

“We’re fully aware of problems like plagiarism and cheating,” he said. “The real challenge is not to find a way to forbid students from using AI or ban them from technology, but to change education and stimulate their critical thinking.”

He said tools such as ChatGPT could even transform the way examinations were conducted in future, with students tested on the questions they would ask rather than the answers they provided.

“I’ll ask you to give me the best questions because answers come from the machine,” he said. “It’s not about how to make machines dumb, but how to make humans smarter.”

The real challenge is not to find a way to forbid students from using AI or ban them from technology, but to change education and stimulate their critical thinking, says a professor. Photo: K. Y. Cheng

‘Need to recognise threats, ethical issues’​

HKU faculty of law associate professor Ryan Whalen, whose research interests include innovation policy and legal technology, said it was important for educators to adopt an open mind about the applications of future technologies, as an AI tool like ChatGPT could be useful in content production.

“It can help people do research to craft, particularly compelling paragraphs, but we need to make sure we’re still training students to do the critical thinking needed to use the tool critically, and not just rely on it,” he said.

“We need to train students to create compelling arguments, to be creative scientists, lawyers or humanists. But in future, that creativity is going to engage with AI, so we need to grapple with how to do that. It’s going to require training at the university level, and also at the secondary and primary school levels.”

Professor Eric Friginal, head of the department of English and communication at Polytechnic University, said safeguarding accuracy would be the biggest issue as new AI tools like ChatGPT were used.

“There are mistakes and the immediate and uncontrollable possibility of plagiarising information from tools like ChatGPT. You need to recognise the threats and ethical issues,” he said.

As an educator, he said, he wanted to understand the tool fully and the ways it could be used.
“Students submitting something that they didn’t write has never been good,” he said. “But with ChatGPT, it’s a little different because it’s technically writing something that the users input to the chatbot and creating something based on the user’s prompt. Human intervention is still needed.”

He said his publisher, Elsevier, a Dutch company specialising in scientific, technical and medical content, had already issued guidelines for authors about the role of ChatGPT or AI-based tools.

A company spokesman said such tools could be used to improve the readability and language of a research article, but not to replace key tasks that should be done by authors, such as interpreting data or drawing scientific conclusions.

“These tasks can only be attributed to and performed by humans,” the spokesman said.

“We ask authors to declare if and how they have used generative AI tools like ChatGPT. These technologies cannot be listed or cited as authors – who are ultimately responsible and accountable for the contents of their work.”

Hong Kong authorities have not yet weighed in on the impact of ChatGPT in education, but the government is planning to set up a task force to study if the new tech tools should be regulated through legislation.

“The special task force is a preliminary idea to deal with the opportunities and challenges brought about by AI-based revolutionary technologies such as ChatGPT,” a spokesman for the Innovation, Technology and Industry Bureau said. “We will continue to monitor the future development of relevant technologies and make appropriate responses.”

HKU law professor Whalen said it might be too soon to implement broad regulatory regimes on generative AI unless specific issues had to be addressed immediately.

“As an example, there’s apparently a subculture of producing pornographic materials but swapping the faces with someone else’s. That’s an emerging problem of AI use, and governments might want to have specific, narrowly tailored regulatory responses,” he said.

“It’s too early to try to craft a whole fulsome regulation that stipulates when and where it can and cannot be used, because we just don’t yet know the full universe of possible applications.”

Quora, a California-based social media platform with a question and answer format, has developed Poe, a chatbot partly powered by OpenAI, the technology behind ChatGPT.

Co-founder and CEO Adam D’Angelo told the Post that despite the controversy over aspects of the new AI tools, they were likely to have a transformative effect on knowledge-sharing and society.

“I can understand the challenges this creates in the short term for education, and for certain in-person tests I would expect that it should be disallowed similarly to how calculators might not be allowed,” he said.

“But I believe the right long-term solution for education, and for society more broadly, will be to allow it and make sure institutions adapt appropriately.”