Artificial Intelligence (AI) got a huge public relations boost in November 2022 when a Silicon Valley start-up OpenAI released its breakthrough text and essay generating system, called ChatGPT, to much fanfare. Type in a question and the chatbot will reply in seconds with a few paragraphs of facts and commentary. The AI chatbot can also write essays, compose speeches in the style of a famous person, draft in different languages, summarise documents and write code. It is able to process natural-language prompts and offers conversational responses. The bot remembers the thread of your dialogue, using previous questions and answers to inform and improve on its next responses. It derives its answers from a staggering 175 billion parameters and 570 gigabytes of text.

In just five days after the bot’s unveiling, ChatGPT had garnered 1 million users, making it one of the fastest consumer-product launches in history. OpenAI has received a US$10 billion investment from Microsoft, with the tech titan announcing a new version of its search engine Bing incorporating the same technology that underpins ChatGPT. Bing has only 9 per cent of the world’s search market, while Google dominates with 84 per cent, but people are already asking if ChatGPT poses a deadly threat to the world’s default search engine. Earlier this month, Google’s share price lost US$100 billion in market value after its own new chatbot shared inaccurate information. Now, ChatGPT has at least 100 million users.

Let’s talk about the hype

Is ChatGPT really a ground breaker in AI?

We took it for a spin and asked the bot about the Keppel Offshore and Marine (KOM) corruption case, a news story which was very much developing at the time. ChatGPT was able to summarise the case, detail the legal settlements and provide commentary that the “government has been criticised for its perceived failure to prosecute individuals” in connection with the case. It went on to explain that “although the company reached a settlement with the authorities in the United States, Brazil, and Singapore in connection with the investigation into allegations of corruption and bribery in Brazil, some have argued that the Singapore government did not do enough to hold individuals accountable for their role in the alleged misconduct.” In fairness, the bot also pointed out that the government has defended its actions.

ChatGPT’s response was written with sufficient detail and correct emphasis on salient points to be useful to anyone wanting a quick precis on the issue. And it accomplished this in seconds.

Chest beating comments and commentary abound on the implications of this new development, including whether AI will now be able to replace humans in their jobs. It’s a legitimate concern.

Which jobs will become employment history?

A few roles can be easily supplanted by ChatGPT, once they have worked out the kinks in the bot.

Clearly, as the above KOM example demonstrates, the chatbot is able to do a large chunk of the work currently assigned to junior reporters preparing news reports. For the same reasons, the more rudimentary part of speechwriters’ and researchers’ work can also be handled by AI. The Singapore Government has announced that it is rolling out ChatGPT for civil servants to conduct research and to draft reports and speeches. Marketing personnel needing to compose sales or promotional content and flyers to potential customers, including in languages they do not speak, will also be affected. Clearly, simple customer service queries can be easily handled.

Lawyers are not exempt. The AI that powers ChatGPT could potentially be applied to the processing and summarising of voluminous documents, putting junior associates and paralegals charged with due diligence reviews out of a job. The bot can draft first cuts of a statement of claim and generate generic advice based on simple factual matrices.

ChatGPT agrees.

When we asked it what jobs it is likely to replace, it informed us, with no hint of rancour or glee, that data entry and transcription tasks which involve entering large amounts of text-based information into a computer, simple customer service enquiries, translation services and content creation such as articles and blog posts would be well within its capability.

How should we respond to ChatGPT?

Singapore has, for years, been trying to upskill ourselves and climb the economic value chain. It is already the case that our more basic jobs are being taken over – currently by lower paid foreign workers, whether working here or as offshore outsourced service providers. In future, some of these roles will be played by AI.

Our strategy to combat this latest development need not change.

AI, as it currently stands, needs to deliver output based on historical data as amassed on the internet. It is largely still backward looking and as such, does best when addressing a situation or problem for which its pattern recognition powers are honed. It also looks at hard data points – albeit trillions of them – and cannot respond to fast changing situations, read human facial emotions or body language or apply critical thinking.

For example, ChatGPT may be able to spit out legal research on a point of law in milliseconds but it would not be able to unpack a complicated dispute and advise clients on the best commercial option taking into account the (often unspoken) fears and desires of the individuals involved, feelings which would not be based purely on logic or precedent. Solutions which are highly customised, fixing complex one-off problems and dealing with the messy world of irrational human reactions are not its core strength. They need to be ours.

Singapore must respond to this challenge by continuing to invest in innovation, training and upskilling. We should recognise that any task which is repetitive is likely to be taken over by a robot. Hence our working population should be at the forefront of cutting edge technology and service offerings. And with such phenomenal technology like AI on offer, we would be foolish not to use it to simplify our work process. ChatGPT and the like can become a tool for us to be more effective. To start, we should consider which tasks it can perform effectively and accurately and revamp our businesses to incorporate those cost savings. Given sky-high manpower costs from tight labour supply, having technology absorb some of the work that is time consuming and tedious is an advantage.

But how we integrate AI into our work systems is also important. Often it is doing the rudimentary tasks that makes us better at our jobs when the assignments get more complex. For example, a lawyer learns what can go wrong on a transaction from doing legal due diligence. If AI is going to accomplish back-breaking reviews in minutes, our professional training needs to find ways to replicate the same learning process in some other form.

The advent of AI will also improve human output. Those fearful that ChatGPT will replace them at work can take comfort from the fact that the novelty of purely bot-generated content is already palling. As Steven Pinker, the Johnstone Family Professor of Psychology at Harvard University and noted author pointed out, there is already skepticism over auto-generated content: “journalists have already stopped using the gimmick of having GPT write their columns about GPT because readers are onto it”.

So while bot-created essays or commentary will be a good starting base, saving individuals hours of grunt work, the real value comes when human thought and creativity are added to the product. 

This analysis is based on the capability of ChatGPT in its current infancy, but bots will not just limit themselves to cold hard fact generation; Bing has reportedly been spewing emotional messages, hinting at a future where humans can discuss their thoughts and feelings with software. As bots become more complex, so will the legal, regulatory and safety issues surrounding them. Jobs will be created to deal with these problems, and to innovate products and services for a world where bot output is a part of the landscape.

ChatGPT is a phenomenal breakthrough and should be embraced by businesses, just like we did the mobile phone, computers and email. And until it becomes sentient and self-aware, humans – so long as we keep thinking about how to stay relevant and continue adding value – are safe.