In Silico: The Growth of Artificial Intelligence (AI)

At its core, AI is about mimicking some kind of human intelligence or decision-making. It can help us process and categorise data, make decisions based on available data, or even create new data based on a prompt.

AI jumped into the public consciousness with the release of OpenAI’s ChatGPT in November 2022. ChatGPT is a generative AI capable of understanding and replying in natural language and outperforming similar tools we’ve seen over recent years. Since then, commentators, financial media, and tech companies have hyped how generative AI may evolve, and a variety of new models have been released in the first half of 2023. Today’s generative AI lets users create text content, images, and sound with simple prompts, igniting optimism—and pessimism—about what a future filled with AI may look like. [1]

For well over a decade before the launch of ChatGPT, AI has been deployed in commercial use cases across almost every sector and industry. AI is used for content recommendation, navigation, virtual assistants, song recognition, image enhancement, signal generation, and more. [2] With the release of highly capable generative AI models, businesses have also begun to integrate automated customer service chatbots, coding assistants, and more capabilities. We believe we’re at the beginning of what we may see from generative AI integration in business.

Generative AI is a technology that learns the relationships between data points and creates similar data from it. For example: generative AI can be used to understand how words are used in sentences and then create new sentences based on that understanding. Scaled up and made more powerful, this technology can be used in chatbots, to write emails, respond to questions, and summarise texts. This technology can also be applied to code, images, sound, and video.[3]

Generative AI is essentially a prediction model. Using the example of text, if we ask the model a question, it’ll give us the most likely series of words that follow based on its understanding of the data initially used to train the model. The latest models also understand and integrate context, topics, and cues. These models are probabilistic—there’s a dice roll that happens with every word to decide what word should follow next. This gives us the impression of creativity. [4]

Generative AI can be applied to a range of modalities, including text, images, sound, and even videos. When working with text, generative AI can be used to summarize documents, produce scripts, write emails, and more. It may also be particularly useful in software engineering by helping to understand, document, and write code. Similarly, it’s now seeing early use in everyday data analysis as well. When working with images, sound, and video, generative AI has also shown impressive capabilities in editing and manipulating content.

Maybe tomorrow, maybe never. We think it’s important to remember that AI has been an increasing part of our lives for the last 15 years. While some job tasks have been automated, few jobs have been entirely replaced and many new jobs have been created. It’s also important to remember that whether AI will replace people is as much to do with whether AI can replace a job as it is to do with whether and how legislators, regulators, companies, or consumers embrace new technology.

If AI can increase the productivity of workers, it may have a deflationary effect as businesses could produce more with less. Indeed, many technologies throughout history have helped grow real incomes, and recent technological change has also tended to have a deflationary effect on the broader economy. In the short-term, AI may have more of a wow factor than real impact. But like major technological innovations that have come before it, we believe AI has significant scope for compounding improvement as economies adjust to accommodate this new technology. [5]

We see AI affecting three broad categories of companies—enablers, adopters, and responders. Enablers of AI are those companies that are building and powering the AI systems. These include semiconductors manufacturers, deep-pocketed large tech companies, and those companies that have large amounts of proprietary data that may help train better AI systems. Adopters of generative AI may benefit from the technology through improved product offerings or new efficiencies in internal processes. Finally, responders are those companies that rise to the new and unique challenges that AI may bring. [6]

Considering the broader market environment, we’re concerned that the recent run-up in technology stocks is looking overdone. Valuations, both on a trailing and forward basis, have climbed markedly while earnings expectations have already risen for major tech names that are involved in AI. We suspect that if monetary policy and credit conditions were looser, valuations may be even more stretched and bubble-like. We note that it’s possible to be right about the impact of a new technology and still invest long-before its effects are realised.