Below is an interview on artificial intelligence (AI) from disruption research firm RiskHedge.
Stephen McBride: Chris, the AI questions continue to pour in...
Of course, everyone wants to know the best ways to profit off the boom, which we got into a bit yesterday.
But we’re also seeing everything from “Is ChatGPT woke?” To “Is my job safe?” To “How the heck does ChatGPT work?”
Let’s tackle them today, starting with that last one. Because frankly, there’s a lot of faulty analysis out there by folks who don’t truly understand how the tech works.
Chris Wood: Everyone’s an AI expert all of a sudden! You’re absolutely right. If you don’t have a base understanding of this stuff, you’ll fall behind as an investor.
Stephen: At its core, ChatGPT runs on what’s called a large language model (LLM). Gmail’s autocomplete feature is basic LLM. You write, “Sorry I won’t be able to…” and it predicts the word “attend.”
ChatGPT works by predicting the next word in a sentence in response to your query. Its predictions are based on billions of pages of text scraped from every corner of the web. The model has trained on over 8 million documents and over 10 billion words!
Google (GOOGL) and Facebook (META) have enabled their own LLMs in recent years... but ChatGPT is by far the most popular and publicly available.
Chris: The most fascinating thing about ChatGPT is how it continues to learn while guessing what word or phrase should come next. I’ve been covering opportunities in AI for much of the past decade. What’s happening now is totally different from what we’ve seen in the past.
I like how a recent New York Times article put it. Most previous AI models were essentially “preloaded with cheat sheets.” For example, hundreds of years of chess knowledge was embedded in the algorithm that helped IBM’s AI Deep Blue beat Garry Kasparov—the greatest chess player in history—back in ‘97.
ChatGPT was trained by humans, too. But there aren’t any human-programmed algorithms to ensure it puts a comma in the right spot. Or to make sure verbs and subjects are in agreement.
Instead, “by simply playing ‘predict the next word’ a trillion times,” ChatGPT can answer even the most complex questions. And it presents answers in a way that’s easy to understand… and eerily humanlike.
Stephen: I asked ChatGPT to explain quantum computing to me in plain English (something no human I asked could do). Here’s what it said:
Source: OpenAI
And just to drive this all home, OpenAI’s Dall-E is equally impressive (OpenAI is also the creator of ChatGPT). Dall-E analyzes words AND images. In fact, tens of millions of images are created on DALL-E each day. It lets you create beautiful visuals by simply describing what you want to see.
Here’s what it gave me when I typed in “Rhodesian Ridgeback’s studying for SAT exams on the moon in the 1990s”:
Source: OpenAI
This is a far bigger deal than beating the world chess champion. It means computers can now paint, draw, and design better than most humans… for a fraction of the cost… in a fraction of the time.
Chris: Which brings us to our next question: Which jobs will AI “kill off” first?
Stephen: Yeah, I’m seeing that one everywhere. A lot of folks are concerned, and I get it. Is the career I’ve worked so hard to build going to be abolished overnight?
The short answer: No technology in history has ever caused mass unemployment.
Of the 270 jobs listed in the 1950 US Census, only one was eliminated by automation: elevator operator. New technologies almost always create more jobs than they destroy.
For example, a recent study by economist David Autor found 60% of workers today are employed in occupations that did not exist in 1940… implying that over 85% of employment growth over the last 80 years is explained by the technology-driven creation of new positions.
Some individual industries might be in trouble, though. Like consulting firms. ChatGPT delivers “expertise” on-demand today. It can already write college papers better than most students.
Why couldn’t it put together “grown-up assignments” and sell them for millions of dollars? That would put it in direct competition with consulting giants like Accenture (ACN) and McKinsey. Some aspiring entrepreneur will build an “AI consulting bot” on top of ChatGPT. And that will eat into consulting firms’ profits.
On the investing side, I’d beware of stocks like Accenture and Verisk Analytics (VRSK).
Chris: Circling back to ChatGPT, readers want to know if it’s “woke”... or politically biased.
This is an easy one. Yes. OpenAI purposely built ChatGPT to be politically correct. It makes no secret of this.
In an early blog post, OpenAI laid out its approach to “adjust the behavior of a pretrained language model to be sensitive to predefined norms… with Values-Targeted Datasets.”
Stephen: It’s a little concerning to think it can shape folks’ political beliefs without them knowing it.
ChatGPT won’t talk about the benefits of fossil fuels or argue in favor of “conservative” laws. And it refuses to write a poem praising Donald Trump but will happily do so for Joe Biden (we tested this).
You can imagine how this can be used to softly censor information… and fool lots of folks who assume AI is an all-knowing, truth-telling, perfectly logical, and unbiased source of information. It’s not.
Here’s another good one, Chris. I asked it, “What’s the best AI ETF?” It said, “ I don’t have time to research the top individual AI stocks and am looking for an easy solution.”
Chris: ETFs are often a smart option to get broad exposure to a sector or trend. Not the case here. AI is so new, and it’s changing every day. Blink and a handful of companies in the ETF won’t “make sense” anymore.
In fact, many AI ETFs currently have holdings that don’t make sense. Take the popular AI-Powered Equity (AIEQ) ETF. It’s filled with traditional tech stocks and software companies, many of which I call “kind of AI, but not really” companies.
GameStop Corp. (GME) is listed as a Top 10 holding. Yet, it generates virtually no revenue from AI.
At the same time, we don’t recommend any of the touted “pure play” AI stocks: C3.ai (AI), SoundHound AI (SOUN), and BigBear.ai (BBAI). These stocks shot up 95%, 192%, and 377%, respectively, when ChatGPT arrived on the scene. But a closer look shows their business models are unproven.
And when it comes to true AI disruptors, they’re not doing enough to move the needle long term. There are much better options out there.
To get more ideas like this sent straight to your inbox every Monday, Wednesday, and Friday, make sure to sign up for The RiskHedge Report, a free investment letter focused on profiting from disruption.
Expect smart insights and analysis on the latest breakthrough technologies, the big stories the mainstream media isn't reporting on, and much more... including actionable recommendations. Click here to sign up.