The Game-Changing Role of ChatGPT on Boards
By WILLIE CHENG
The Business Times, 6 October 2023
Since its prototype was unveiled on 30 November 2022, ChatGPT is now the fastest-growing consumer application ever. Within less than a year, it has amassed over 1.5 billion monthly visits and 300 million active users worldwide. Its parent company, OpenAI, is currently valued at US$90 billion (S$123 billion).
As artificial intelligence (AI) continues to revolutionise industries, ChatGPT stands out as a pivotal innovation. Given the velocity and scale of this transformation, boards can ill afford to be passive observers. So, how should directors and their organisations engage in this game-changing technology?
Understanding the technology
First and foremost, directors should personally acquaint themselves with ChatGPT. The registration process is straightforward and cost-free (for now). Once in, the user can pose questions, generate content, analyse data, assess sentiments, and even code.
In essence, ChatGPT is an AI chatbot built on Generative Pre-trained Transformer (GPT) architecture. GPT is a machine learning model designed to understand and generate natural language. ChatGPT is pre-trained on a diverse dataset that includes text from the internet, books and articles.
Of course, ChatGPT is only one example of Large Language Models (LLMs), software programmes trained to understand and generate human-like language. Others include Google’s Bard, Amazon’s Claude and Baidu’s Ernie Bot.
LLMs are, in turn, a subset of Generative AI technologies that can create new data resembling existing datasets not just in text, but also images (e.g. Midjourney), videos (e.g. Synthesia) and even music (e.g. Jukedeck).
Understanding AI should be part of a director’s ongoing educational journey and a collective effort undertaken by the entire board. Several institutions offer courses aimed at enhancing AI literacy. If in-house expertise is available, training programmes can be customised to explore the specific impacts of AI on the organisation.
The applications for Generative AI are still explorative. Early uses showing promise include:
- Conversational data mining: The AI model can be trained on extensive corporate datasets to be queried by humans for specific insights. For instance, SID is ingesting our rich repository of board guides, case studies, thought leadership articles and relevant regulations into a Generative AI model named SIDgpt. This contextual search engine would then provide SID members with pertinent answers to their questions on corporate governance and directorship matters.
- Document drafting: Generative AI can create first-cut business documents (e.g. meeting minutes or discussion papers) for administrative efficiency. Even transcripts of conversations or meetings can be input into ChatGPT to generate assessments or summary reports.
- Customer engagement: The technology can be used to handle routine customer queries autonomously. This would free up the staff to focus on more intricate and nuanced tasks.
- Idea generation: Generative AI can be used to facilitate brainstorming sessions. It has the potential to offer fresh perspectives that might not be readily evident.
The list can be endless. The board should task management to identify potential use cases for Generative AI in the company and initiate pilot programmes. Once validated, these can be implemented at scale.
Management should regularly report to the board on the progress of these applications, given their newness and impact. Some boards have even made AI a standing agenda item in every board meeting to emphasise its significance.
While Generative AI offers transformative potential, they are not without risks. As businesses begin to explore various use cases, they are concurrently uncovering associated risks:
- Data leaks: When users share data with ChatGPT, that information becomes part of the training data for the machine learning model. For example, staff at Samsung inadvertently leaked sensitive information via ChatGPT in three separate incidents.
- “Hallucination”: When it lacks the necessary data, generative AI can fabricate responses that may sound plausible but are incorrect. In a New York federal court case, an attorney faced sanctions for citing case law that was “hallucinated” by the AI model.
- Questionable sources: ChatGPT has been trained on a wide array of internet content, much of which would contain bias and even misinformation. This raises concerns about the accuracy and ethical implications of the information generated by the model.
Boards should evaluate such risks and implement a robust governance framework to align the use of the technology with the company’s security, confidentiality and ethical guidelines.
The board itself should use ChatGPT and Generative AI. It would not only benefit from doing so but would also be setting an example for management.
To begin with, ChatGPT can be an effective tool to disseminate information and share knowledge among the board members. It can help produce succinct briefings on business trends, best practices and emerging risks to facilitate informed discussions during board meetings. Additionally, it can assist in summarising meeting discussions, as described earlier.
ChatGPT can enhance board decision-making. It can generate business scenarios based on existing data and market trends, provide insights into competitors’ strategies and analyse customer feedback at scale. The technology can also be a brainstorming catalyst, generating innovative solutions to complex challenges.
For regulatory compliance, ChatGPT can be configured to stay abreast of legislative changes, thereby alerting the board to any compliance matters requiring immediate action.
As we enter a new era of AI-driven advancements, the question for boards is not whether to embrace technologies like ChatGPT but how quickly and deeply to do so.
The writer is a former Chairman of the Singapore Institute of Directors.