Generative AI will boost local legal sector: Forum

Source: Straits Times
Article Date: 19 May 2023
Author: Osmond Chia

The rise in the use of generative AI will revolutionise the domain of knowledge management, particularly in the legal context, says the author.

I read with interest journalist Osmond Chia’s article, “More demand for natural language skills amid AI boom” (May 15).

Mr Chia said that the advent of chatbots and other artificial intelligence (AI) tools has increased the demand for people with natural language skills, and not just skills in mathematics and coding.

We are indeed at the cusp of an inflection point – certainly one with ripples into many facets of society.

Just as how the industrial revolution led to a need for new skills relating to the operation of the steam engine, the rise in the use of generative AI will revolutionise the domain of knowledge management, particularly in the legal context.

Beyond Singapore, many international law firms and legal tech start-ups are starting to take keen interest in how such skills might apply in the practice of law.

Knowledge management is a critical part of the day-to-day life of an attorney, and a mastery of prompt engineering could yield amazing results.

One example might include what dispute lawyers often engage in – distilling the convoluted facts of a dispute from legal database searches into key legal points. This thinking translates well into natural language programming, and it is because of this that I believe lawyers, and our local legal industry, are well poised to use generative AI as an aid.

This aligns well with the Government’s national AI strategy, particularly as Singapore gains traction as a top arbitration hub.

I believe the benefits of AI and learning new skills have the potential to make Singapore one of the world’s preferred legal hubs.

Marcus Ho Kan Jie

More demand for natural language skills amid AI boom

The rise of chatbots and other artificial intelligence (AI) tools has increased the demand for people with natural language skills, and not just mathematics and coding.

These skills fall under a nascent field of software development called prompt engineering – the ability to direct generative AI apps like ChatGPT or image generator Midjourney to perform specific tasks using English and other human languages.

Unlike for traditional software engineers, many prompt engineering roles advertised on job portals do not require technical knowledge.

Employers interviewed by The Straits Times say they have posted job openings for prompt engineering roles, while other firms have expanded the job scope of their employees to include such skills.

While demand is high, analysts and recruitment firms say some firms are currently expanding the roles their existing engineers play.

Reports out of the United States suggest prompt engineers can be paid six-figure salaries.

For example, US AI firm Anthropic had advertised on its site for a “prompt engineer and librarian” role that will pay up to US$375,000 (S$502,000).

AI Singapore (AISG) was among a handful of local organisations which posted hiring calls for prompt engineers.

AISG, a national AI programme launched by the National Research Foundation, also expanded the job scope of existing engineers to focus on developing language skills.

According to the job listing, the duties of a prompt engineer include designing prompts to make large language models like ChatGPT complete tasks reliably.

AISG senior director Leslie Teo said the job suits those who understand large language models and can use them to solve problems creatively.

Those hired will train AI programs to learn the cultural differences in the region as part of AISG’s goal to create more inclusive AI.

However, he said that to do this job well, some level of coding is still necessary.

“The skills are not constrained by one’s background. You need creativity and good communication, but also good understanding of how machines – and humans – work,” added Dr Teo.

When AISG product engineer Clarence Lam, 36, started a new project in March to develop a story-writing AI bot to teach children Malay, much of his work shifted to linguistics.

His team of engineers at AISG guide the AI to understand terms the children might use and suitable themes for the AI-written stories, and simplify the language it uses.

He also had to set guardrails to prevent the app from writing foul content or being misused.

Mr Lam said the role has become more conversational as engineers test different phrases to see how a bot may respond.

As part of testing, he needs to program the bot and analyse the quality of its answers before they are translated into Malay.

“Does the story have a climax? Does it make sense, or is it saying something stupid or morbid?

“The things we need to figure out in our role today involve the more qualitative aspects of what makes a good story,” said Mr Lam, who graduated with an arts major in photography in 2014, before switching to the tech industry.

He learnt to code in his school co-curricular activity, building educational websites for school projects. Since then, he has kept up with new coding languages and tools.

“This shift to prompt engineering has allowed me to use some of my creative background. AI work has evolved from people using zeros and ones to using English to talk to a computer,” he added.

AI Seer, an AI start-up developing a chatbot and other AI tools to facilitate mock interviews, is also looking to recruit prompt engineers.

The firm’s engineers analyse the bots’ transcripts and create goals for the AI, such as asking better questions to test interviewees, said head of product Shahruj Rashid.

Mr Maxim Tint, chief executive of Gtriip, a digital identity software provider for hotels, said he has hired more AI engineers in the past year. Existing engineers had to learn prompting skills to utilise generative AI.

Mr Tint, who is also the vice-president of IT association Tech Talent Assembly, added: “Creating prompts is one of the tasks.

“Usually you can’t just do prompts as a full-time job, so engineers still need to understand architecture, know how to write good code, and review the code that generative AI produces.”

Recruitment firm Morgan McKinley has received more queries for employees with prompt engineering and AI skills, said Mr Shrijesh Patel, associate director of technology and transformation.

“The interest in machine learning candidates has been high in the past few years and there has not been a sudden spike, probably because the tech market has not been in the best shape,” said Mr Patel.

“But clients have asked us for candidates with these skills.”

He added that companies are not increasing their headcount, but are likely expanding the job scope of existing AI employees to handle prompt engineering tasks.

Mr Leslie Joseph, principal analyst at research firm Forrester, said demand for prompt engineers here has not ballooned, likely because the technology is still growing.

But interest in generative AI has spawned a growing pool of prompt engineering freelancers in online marketplaces such as PromptBase and Gumtree, he added.

Firms keen on a quick fix for AI issues can turn to thousands of listings on these platforms.

Prompts for generic artwork for Midjourney or instructions for ChatGPT can start at several dollars, but bespoke prompts can fetch top dollar depending on the complexity of the task.

Despite the interest in prompt engineering, AISG’s Dr Teo said he is wary of overselling the hype.

“Prompt engineering is temporary because over time, we will learn how best to talk to a bot. But we will always need people who can translate bot behaviour and create solutions with coding and emotional intelligence.”

He added: “The human aspect is always going to be valuable.”

Mr Raju Chellam, executive committee member of trade association SGTech’s Digital Trust and Cloud & Data Chapters, said basic prompting could become a thing of the past as AI systems learn to automate its own prompts.

Instead, prompt engineers will need to focus on the ethics of bot behaviour.

He said: “Prompt engineers will need to design prompts that can provide clear explanations of how the AI model arrives at its conclusions, especially in healthcare and finance, where the consequences of incorrect decisions can be significant.”