Session Summary

Generative AI has captured the imagination of the public since ChatGPT’s release in November 2022.

Generative AI is the first technological development that surpassed the Turing Test. Generative AI is able to interact in a way that seems human and has had some high-profile achievements over the past year, including winning an art prize, passing the US law exam, and passing Google’s coding test.

While the technology behind Generative AI (also known as foundation models) has been around since 2017, ChatGPT’s visual interface allowed a wide audience to interact with the AI models. This is partly why ChatGPT managed to scale faster than any previous technology platform, amassing 1 million users in just 5 days. This may also represent a paradigm shift in the way we interact with technology, from visual/ touch-based graphical user interfaces (GUIs) to natural language (writing or speaking in human languages to a machine).

Generative AI differs from machine learning in its ability to understand the context and components of a sentence or image, and to draw relationships between them. Machine learning is trained on specific data and needs to be fine-tuned in order to predict predefined metrics like credit scores. In contrast, foundation models are self-trained using broad, unstructured data. The prediction process is more of a “black box”, with outcomes that are less predictable than in traditional machine learning.

 

The impact of Generative AI on different sectors and companies will be unequal…

Tens of billions of investment dollars are flowing into the Generative AI ecosystem. However, contrarian to the media narrative, certain sectors may experience limited disruption from Generative AI. For example, sectors like Metals & Mining or Agriculture may not have as many use cases for Generative AI, although Agriculture has already witnessed disruption from other forms of AI, such as in crop forecasting.

Other sectors will be significantly impacted by Generative AI. Media & Entertainment will be heavily impacted, as will Software & Technology, where AI coding assistants will greatly improve productivity. While the heavily regulated core functions of Banking and Insurance (actuarial statistics, loan decision-making) will not be heavily impacted, ancillary functions like customer interaction could benefit significantly from Generative AI.

Job displacement is inevitable, but AI will also pave the way for the emergence of new employment opportunities. According to the World Economic Forum, 65% of children entering primary school today will end up working in jobs that do not yet exist today, and many of these jobs will be just as interesting as the ones that disappear.

 

…But everyone will stand to gain from understanding and learning how to use AI.

AI can augment jobs, allowing us to work at scale and improve efficiency. AI has played a significant role in enabling advancements in Healthcare, particularly in patient data management. To effectively utilise these applications, doctors would benefit from understanding some of the underlying logic. While AI can generate vast amounts of artwork, it is worth noting that behind the art prizes won by Generative AI, there was a human directing the composition, analogous to James Cameron utilising novel technologies in filmmaking. One does not need to be an expert in AI to utilise it effectively, but being AI-fluent is helpful.

Inability to utilise AI may widen inequalities. If MSMEs do not have the resources to utilise AI, they may fall further behind capital-rich large multinationals or conglomerates. However, the cost of deploying AI is likely to fall over time, and smaller companies may be able to better utilise AI if they are more nimble and agile.

Understanding the risks around AI will allow us to take mitigation measures. Some of the key debates around Generative AI include Intellectual Property (IP) infringement (who owns the data and the output?) and bias (what are you training your model on?), and hallucination (what is the model outputting?). Regulations around AI are still evolving, and there will likely be guardrails around AI detection tools (especially for image generation) and what is considered human- or AI- generated. Companies will also put in mitigation measures such as screening Generative AI output manually or using machine learning.