What ChatGPT and the “golden age of AI” mean for investors

• In the field of generative AI, ChatGPT and other “large language models” have made giant leaps in how they learn and what they can create.
• This technology will be a game changer for many businesses, helping them answer questions, create content and even write code.
• We believe AI is an attractive investment theme that affects many sectors and industries — not just traditional technology stocks.

In what seems like the blink of an eye, artificial intelligence has taken a great leap forward — and it feels like everyone is now talking to robots. New “chatbot” technology, recently popularized by a tool called ChatGPT, helps us find instant, useful answers to questions both practical (“What’s the best TV to buy for a large living room?”) and prosaic (“Write a thank-you note in the form of a Shakespearean sonnet”). But it’s just the latest in a string of groundbreaking developments in “generative AI” — which we believe will have profound implications for companies and investors alike.

While OpenAI, the company behind ChatGPT, isn’t the only player in this space, it has certainly taken the world by storm. ChatGPT gained more than a million users in the first five days after its release in November 2022, then added 100 million more by February 2023.¹ To put this in context, it’s the fastest adoption of any consumer application in history.²

So just what is generative AI? It’s not a single program or tool — rather, it’s an umbrella term for a form of machine learning (ML) called “deep learning” that has been in use for some time, generating little public interest until now. Meta Platforms, Microsoft, Google and others started creating private and open-source advanced “generative pre-trained transformer” (GPT) models in 2018. Generative AI uses models trained on sets of data to perform certain tasks – including creating text or images – and/or make predictions, all without human direction. And as recent news headlines have shown, these systems recently made a big technological leap in how they learn and what they’re able to produce.

Most early ML models involved supervised learning, in which humans were needed to classify data — for example, identifying an image as a “dog” or a social media post as “political.” But the notable new developments in generative AI involved unsupervised learning, in which the model makes its own predictions and calculations based on the massive amounts of data fed into it. That means the technology has moved from being able to identify a dog in an image — the famous “chihuahua or blueberry muffin?” test — to creating an image of a dog or writing poetry about a dog (Exhibit 1).

This technology is a game changer for many businesses. ChatGPT and other large language models (LLMs) are trained on immense sets of text data to execute specific tasks, such as finishing a sentence or completing a line of code. LLMs have billions of variables (parameters) that the models can change as they learn, increasing their accuracy rate and expanding the business use cases. For example, an AI-driven chatbot might be able to answer 90% of questions from banking customers, freeing employees to spend more time selling services or provide a better in-person experience to the bank’s highest-value clients.

Imagine all the possibilities for this type of technology. Companies that need to create written materials — simple product descriptions, technical manuals or even answers to questions like “Why did my insurance rate go up?” — will be able to do it faster and cheaper than they could using human labor. For software companies, generative AI technologies can perform the grunt work of basic coding, allowing programmers to engage in more-important and higher-value-added programming tasks.

Generative AI can be used to generate text, images, voice and movies. This could aid or even replace time-intensive human tasks, such as designing logos, illustrating scenes and designing products. Generative AI models will be integral to the next generation of augmented and virtual reality–related technology in gaming; visual interactive entertainment; and business simulations, such as “digital twins.” Indeed, there are numerous immediate applications of the technology across a variety of sectors and use cases. The top right quadrant of Exhibit 2 shows that the types of tasks AI can perform has increased significantly in recent years (and so has the computing power required).

Generative AI can replace time-intensive human tasks, such as designing logos, illustrating scenes and designing products.

For investors, we believe the technology is worth a closer look. Generative AI has the potential to significantly enhance customer engagement, free employee resources for more value-added activities and improve company margins. At the same time, companies that do not have a well-thought-out or well-executed game plan around the technology may find themselves disrupted, rather than being the disruptors.

But generative AI isn’t all fun and games. The widespread adoption of the technology raises important ethical questions and legal issues, such as copyright and licensing issues for AI-created images. Chatbots can also produce incorrect, inconsistent or even inappropriate answers — a result of their using the entire internet as their training set.

For many companies applying the technology to specific use cases, this will be less of an issue, and they can further train the machine with more specific data sets and fine-tuning. Indeed, OpenAI is hoping that by initially opening ChatGPT for broad public use, it can better help train the model with real-time feedback from users.

For other companies — particularly those in the social media sphere — there could be problems, particularly if chatbots facilitate the dangerous spread of false information. As we have seen numerous times with the advent of a new technology, the applications and uses are introduced first, and then regulation has to catch up. This is why it’s important for investors to be selective in these early days, favoring a bit of caution as opposed to taking an “all-in” approach.

Now we have generative models for proteins. We have generative models for chemicals. We have generative models for language, for text. We have generative models for images and video. As you know, we’re working on generative models for 3D. You won’t be able to populate the world’s Omniverse, the metaverse, with human-engineered content. It has to either be perceived through computer vision, or generated, or a combination of both. – Jensen Huang, NVIDIA CEO³

We believe investors should not view the difficult performance environment for technology stocks in 2022 as an indication that AI-driven transformation is slowing or that the importance of AI-related technological advancements has decreased. Rather, we are witnessing increasing use of AI by a range of companies for cost reduction, revenue enhancement and improved customer engagement. The number and variety of use cases continues to grow rapidly (Exhibit 3) — perhaps evidence that we are entering a “golden age of AI.”

We expect to see improvements in the quality and diversity of generated content; new types of generative models; and broader application in a variety of industries, such as health care, finance and transportation. Additionally, AI should become more accessible to a wider range of users through user-friendly interfaces and tools. The precise future of generative AI is hard to predict, but we are optimistic it will be a key piece of AI-driven transformation and innovation. We advocate viewing AI as a high-conviction theme that crosses sectors and industries, as it is now being adopted well beyond traditional technology stocks.