Artificial Intelligence (AI) had two big moments in the popular imagination over the last two years: the extremely successful rollout of ChatGPT in late 2022 and chipmaker Nvidia’s blockbuster earnings report in May 2023. While these moments captured the public imagination, AI has already been playing an important role for many businesses and in our home life for some time.
The “ChatGPT moment” does represent an important advance in AI , but it’s built on a foundation of decades of work in place. Here are just a few of the places where we already encounter AI every day.
- Advertising, advertising, and more advertising, on the web and in apps, but using highly detailed mass customization that is constantly testing what makes you, for your own special reasons, go click
- Facial recognition, for example, when you open your phone
- Search engines
- “AI” opponents in games
- Preference algorithms to make movie, music, or book recommendations
- Voice recognition and digital voice assistants
- Driving directions
- Spam filters (which science fiction writer Corey Doctorow made the first AI to achieve consciousness in his short story “I, Row-Boat”)
We’re getting a lot of questions about AI these days, including the nature of the investing opportunity. Here are some answers to some of the basic questions we get, with some general comments on investing opportunities and places to learn more at the end.
What is artificial intelligence?
There’s no clear definition of artificial intelligence. Generally, artificial Intelligence is the ability for a machine to interact with its environment in a way that appears to exhibit some characteristic of human intelligence. But looking like human intelligence isn’t the goal in most cases. Instead, it’s achieving a certain goal, which requires something that looks like intelligence. A couple of things to think about with that.
First, we don’t realize how incredibly smart we human beings are. Recognizing a cat or understanding a question in ordinary language, never mind answer it, are incredible accomplishments in the field of artificial intelligence. On the other hands, some things that we would consider very challenging are much easier for AI. IBM’s Deep Blue defeated world chess champion Gary Kasparov in an over-the-board match under tournament conditions in 1997. It took another 15 years before researchers at Stanford and Google had developed a program that could reliably recognize a cat. So “characteristic of human intelligence” is much broader than we usually think it is. Computers are really good at some tasks we would consider impossibly challenging; they are really bad (or at least have been) at some tasks we would consider very simple.
Second, even though we are defining artificial intelligence as doing things characteristic of human intelligence, what they can do with those abilities now goes well beyond human capabilities, whether in speed, outcomes, or both. AI can solve problems we just can’t solve. Of course, we have solved the problem of creating AI that solves those problems, so we do get some of the credit. (And this isn’t the place to get into the conversation about when AI will be able to do that itself.) But then cars can move faster than we can run and human beings can’t fly, so using machines to enhance our capabilities is nothing new.
Stay on Top of Market Trends
The Carson Investment Research newsletter offers up-to-date market news, analysis and insights. Subscribe today!
"*" indicates required fields
How long has AI been around?
The earliest programs generally recognized as AI were written in the early 1950s. Not surprisingly, there were focused on games (chess and checkers), which were well suited to initial research. Checkers has a clear set of rules, a manageably finite number of legal moves, and a clear goal, but is complicated enough that winning requires “thought.”
While there were early leaps in AI progress and AI continued to generate buzz, ultimately the lack of computing power limited how far research could go and the limited availability of computers kept research from reaching any kind of critical mass. That slowly began to change in the 1980s and accelerated in the 1990s, with AI increasingly finding broader uses. So AI has been around for a long time and has even been actively used for quite a while, but the widespread availability of ChatGPT and other generative AI models represents an important evolution.
What was so special about ChatGPT?
Even though AI was already doing some extraordinary things, the mass adoption of ChatGPT upon its release represented a critical point in the development of AI. First, ChatGPT could respond to a wide variety of prompts (questions) for which it was not specifically trained. This allowed it to try its hand at things like taking the bar exam (scoring in the 10th percentile (low) in the initial release but 90th percentile in the latest version) or the SAT (70th percentile in math on the initial release; 93rd percentile in the most current version). The challenge for AI isn’t answering the questions on these tests; it’s understanding what the question is. If the AI “understands” the question, answering it is usually easy to trivial.
Not only can ChatGPT address a wide variety of questions; it could provide answers in ordinary language, actually creating content. In fact, ChatGPT and other “generative” AI models can create content in response to general prompts that looked a lot like the result of human creativity, which includes written content, images, or even songs, depending on the AI.
In addition, although it may seem trivial, it can do all of these things with a simple user interface that makes it accessible to just about anyone. No programming experience required. Just give it the right prompt and off it goes, with perhaps just a little experience in writing effective prompts.
And while ChatGPT was the first generative AI to gain wide adoption, the playing field is extraordinarily wide, with hundreds of AI tools now available and thousands of startups developing more.
Will AI supercharge the economy?
No. Not supercharge. We believe it will have a meaningful impact on the economy, but it generally takes longer for new technologies to have a broad impact on economic growth than most people think, since it takes widespread adoption that enhances work productivity. (Sometimes technology instead enhances our non-work life, which has genuine value but can even be productivity limiting from a strict economic viewpoint.) However, we do believe the overall environment that made the development of generative AI possible may have a more immediate impact on the economy, fostering a period of productivity growth now similar to what we saw in the late 90s. We wrote about this in our Outlook 24, “Seeing Eye to Eye” (with a little pun on AI in there) and our global macro strategist Sonu Varghese provided more depth in his recent blog, “There’s Reasons to Be Optimistic About Productivity.”
But what about the investment case?
We won’t make a specific investment case here, but instead just highlight some broad principles.
- It’s hard to know who the winners and losers will be. Remember, Amazon started as an on-line bookseller and Netflix as a movie CD delivery service. Find a way to gain diversified exposure.
- You can’t just focus on industries or sectors that use AI, because that would basically be everyone right now. Especially in the early stages, we would emphasize businesses that enable AI and whose broad business could be considered “AI adjacent.”
- While some opportunities will be missed because there’s some great AI research being done abroad, we would emphasize US exposure. We believe the US is likely to maintain a regulatory environment more friendly to the development and deployment of AI, has the advantage of a strong market well suited to early growth, and has the research density and current leadership, in both industry and among universities, to maintain an edge.
How do I learn more?
Here are some non-investment-related resources for further general reading on AI from either top research or industry sources.
Stanford University’s 2023 Artificial Intelligence Index Report. For those who like visuals, an extensive collection of AI-related charts and graphs with accompanying analysis and insight. Lengthy at 384 pages but still digestible.
McKinsey’s The State of AI in 2023: Generative AI’s Breakout Year. A more modest 24 pages introduction to generative AI. McKinsey also has a short introduction to AI more broadly.
For those interested in policy implications, well-known think tanks Brookings and the American Enterprise Institute both have commentary. And, of course, the federal government also has its take.
There are plenty of brief introductions to AI from industry sources, such as IBM and Google.
For more of Barry’s thoughts click here.
02118200-0224-A
