This article first appeared in Wealth, The Edge Malaysia Weekly on April 24, 2023 - April 30, 2023

ChatGPT is an open-source generative artificial intelligence (AI) foundation model by OpenAI. If you go to the website, you can ask this AI model questions and get a response generated from all the data on the internet. ChatGPT is currently in its fourth iteration, with processing power in the trillions of parameters.

ChatGPT has been trained using publicly available data on the internet. It does not search the internet and give you an answer based on a particular source on the internet but rather, it gives you an answer based on its “learning” or “training”.

Hence, it is a generative or predictive model and not a deterministic model, meaning its answers are not absolute truths. ChatGPT is a model that can even generate new content from its learnings. More on that later.

By my limited usage of ChatGPT, it seems intelligent and is able to answer follow-up questions on the same topic, much like a helpful sales representative or telephone helpline operator. Indeed, ChatGPT is estimated to have hit 100 million monthly active users just two months after launch, which is an achievement when you compare it to the 2½ years it took Instagram or the nine months it took TikTok to hit the same number.

The potential of ChatGPT lies in it being a foundation model that can be trained to become a specialist. For example, it can be trained to be a specialist model for the coding community or the vet community for your pets. Think of the medical student studying to be a general practitioner (GP) and then going on to specialise in fields such as the heart (cardiologist) or brain (neurologist). In this case, studying to be a GP is the foundation before going on to other specialisations in the medical field.

From this foundational model, ChatGPT can be used by many applications to perform various functions. For example, an application called Roam Around (www.roamaround.io) — this was making the rounds recently — uses ChatGPT to generate a travel itinerary for several days in a chosen destination. Instead of going to, for example, Tripadvisor and doing your own research, Roam Around has already learnt about the many things people have posted and is then able to generate a list that is suitable for your length of stay.

The ability to generate new content, be it text, video or pictures, that has meaning and context is what makes generative AI such a powerful development. It has the potential to be disruptive in areas of industry not previously seen, especially in white-collar ones.

The applicability of ChatGPT to different types of applications is thus wide. This is its potential as a foundation model. But it is not the only foundation model. There are others that are less known or that are not open-source models.

For investors, it is always about the financial return on your investment. The applications that can be built using a foundation model are the obvious ones. Search engines can be supplemented with this model to give better results, thus becoming more valuable to online advertisers. Social media providers can incorporate the model to connect people and provide ideas for people to enjoy activities together. Sales management software can integrate a data analysis assistant to generate sales leads based on criteria set by the regional sales manager.

However, if one looks deeper, the enablers of ChatGPT are also potential beneficiaries. To win this race, industry players need a few things. One, a big data set for the AI to learn from. Two, deep pockets for research and infrastructure investment. And three, the ability to deploy the technology quickly.

A set of high-quality data sets is a prerequisite for training an AI model to use various statistical applications to generate actionable insights. The internet provides a baseline for a generative AI model to learn from. But to develop into a specialist model, the AI will require more specialised data. Existing players with a big data set can leverage their data set to generate higher value for their business processes and actionable customer insights.

An AI’s learning ability is correlated to its processing and storage capacity, as well as the human engineering talents available to help it learn and then later generate new content. The hardware in this case is the semiconductor and cloud computing chips used to process and store content. The AI-related talents are limited as the segment is still in its infancy, thus the fight for talent will be brutal and unrelenting but highly rewarding.

Having the platform to scale any AI model will also be key, much like most technologies. The quicker one can scale and spread the use of the application, the higher the market share growth and profits that follow. There will be niches left for the rest of the pack. And while some will be lucrative, the competition is likely to be high as well.

It is estimated that generative AI can capture revenue worth an estimated US$900 billion (RM3.9 trillion). This estimate is made up of search engine advertising and other online advertising spending, software-as-a-service (SaaS) solution providers, cloud infrastructure providers and AI-related chip manufacturers.

For the casual observer, Microsoft’s search engine Bing is likely to disrupt Alphabet’s Google search engine using the new AI model. With a mere 3% share of the global market, Microsoft has the incentive to disrupt and capture the market, as even marginal gains are a significant contribution to the bottom line (a 1% increase is estimated to add US$2 billion to Microsoft’s revenue).

Microsoft Teams has launched a premium subscription version of itself that incorporates AI-assisted MS Teams activities. For first-time users, the monthly fee is US$7 per month and will be increased to US$10 per month later in 2023. Other SaaS companies will want to add an AI assistant to their offerings. This reminds me of Iron Man’s Jarvis. The ability to have an assistant for various tasks and professions will be a realistic vision for the future.

The major players in cloud computing infrastructure, Amazon, Alphabet and Microsoft — and for China, Alibaba, JD.com and Tencent — are all well positioned to capitalise on the AI-related investments to come. These players’ dominant market position and platform size are good bases from which to build to capture the opportunities in generative AI development.

And lastly, the semiconductor chips used for AI training — the graphic processing unit (GPU) and networking-related chips used for AI inference (communications in the form of user queries and machine-inferred replies) — will see incremental growth. Quality names in the value chain of chip design and manufacturing are likely to gain from having generative AI as a catalyst.

Michael Lai is executive director of wealth advisory (wealth management) at OCBC Bank (M) Bhd

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.