This article summarises a fireside chat with Garry Tan, President and CEO of Y Combinator, at GIC Insights 2023, which took place on 31 October 2023 in Singapore.

Moderated by Choo Yong Cheen, GIC’s Chief Investment Officer for Private Equity, the conversation explored key considerations for AI-focused companies and non-tech companies that are seeking to capitalise on the benefits of AI, by commercialising and scaling its usage. They also examined potential risks and discussed what the future might hold for generative AI.

2022 was a tipping point in the evolution of artificial intelligence (AI). The technology has been around for many years, but the widespread adoption of generative AI tools only occurred in recent years across different industries, academia, and the broader general public.

In particular, large language models (LLMs), a form of generative AI, have broken new ground on many fronts. Not only do they leverage algorithms that combine deep learning with huge data sets, enabling the technology to understand, analyse, and summarise unfamiliar content, LLMs are also able to both generate new content and accurately predict future events.

Tan shared that access is what differentiates LLMs from most other forms of generative AI. As many LLMs are open-source, data scientists and developers are able to build on advancements pioneered by others, and apply them to their own use cases.

Access to large volumes of data is critical to realising the potential of these new technologies. Tan highlighted that businesses that have been able to monetise the use of LLMs and AI all typically work with large data sets. This could be proprietary data that has been amassed over a long period of time, or data sets that have been acquired from third-party data providers.

For example, a legal information vendor that provides data on US case law to the country’s many law firms started out as a user-generated website for information on various areas of law. Having compiled valuable data for over a decade, they integrated the application of LLMs in their business to now allow users to access information that once required months of manual research and fact-finding to put together. Users are now able to access the same information within minutes and use it in court immediately.

Tan believes that there are two industries with the potential to benefit enormously from the effective use of generative AI. The first is education. By leveraging generative AI, educational providers will be able to deliver a personalised learning experience that suits the pace of each individual’s learning style. In some instances, AI-powered personalised tutors are believed to be more effective in achieving education outcomes than human personal tutors.

The healthcare industry could be another significant beneficiary. Administrative spending accounts for roughly one-quarter of total annual healthcare spending in the US, which amounted to nearly US$4 trillion in 20211. Tan noted that if generative AI is applied effectively to administrative work in the industry, healthcare costs can be lowered significantly. A simple example he shared is the use of AI for medical scribing, which could free up time for healthcare professionals to increase standard of care instead.

While the integration of AI has the potential to improve existing business processes, many companies are also seeking to create new, AI-powered products. Tan advised that these firms must first identify the type of work that the technology can add value to. This is because until recently, AI has been predominantly applied to mundane and low-value activities only, such as the use of robots as servers in food establishments. Today, AI increasingly facilitates complex, knowledge-based work that demands human analysis, synthesis, and contextualisation to arrive at a decision or choice of decisions. Companies must be strategic and consider the types of problems that AI can help to solve.

On top of that, businesses must think strategically about product commercialisation too. There must be customer demand for new AI offerings to begin with. These businesses must also ensure they can compete with other firms that already have extensive research and development, as well as marketing capabilities.

At the same time, the proliferation of AI continues to raise concerns across industries, governments, and the broader general public. AI is often regarded as a form of super-intelligence with the potential to exert control over humanity. However, the notion of a supercomputer plotting a nefarious act is no different to a human planning such an event through desktop research. In many parts of the world, regulators and security services already have measures in place to anticipate the potential time and location of such acts being planned. In fact, many of these preventive services also adopt various forms of generative AI to optimise their processes.

Similarly, many are worried about the unethical use of AI. Several countries around the world have strict data privacy and data security laws, yet very few have enacted legislation on the application of AI. At the time of writing, only a handful of jurisdictions, including Brazil, China, the European Union, and Israel, have laws governing the ethical use of the technology.

Another concern is that the use and development of AI will be limited to a small pool of tech companies only. Tan reminded the audience that certain LLMs, such as ChatGPT-4, are closed-source and only available with a paid subscription. Limiting access will stifle innovation, slow market development, and create monopolies.

Ultimately, regulators and businesses must ensure that existing and emerging AI technologies are safe, remain as accessible as possible, encourage innovation, and help advance economic activity.

Tan also highlighted the importance of embedding this new, AI-enabled way of working into today’s secondary and tertiary education systems, as well as vocational training programmes. While use of generative AI is growing rapidly, there remains a large proportion of the global workforce that is unfamiliar with its capabilities and applications. Thus, the accessible nature of open-source LLMs that allows users to experiment with the technology is critical in enabling more to become familiar with its potential and hence feel less apprehensive about it. This might even lead them to discover new ideas to integrate AI into their future work and workplaces.

According to the World Economic Forum, while on one hand, AI may replace 85 million jobs worldwide between 2020 and 2025, the technology is also expected to create 97 million new jobs2. Given its enormous value-add potential across a range of industries, markets, and business models – old and new – being able to work side-by-side with AI will be imperative for the success of the workforces of the future.