When it comes to AI, you can’t forget these 11 words...
Over the past few weeks, I’ve shown you where we are with AI today, and where it’s going...
I showed you how to profit. And why you can’t ignore what’s happening, even if you don’t plan on buying an AI stock.
But here’s the overarching message I want you to remember if nothing else:
This is the WORST artificial intelligence is ever going to be.
There are all kinds of problems with ChatGPT, the new AI tool taking the world by storm. The chatbot, which can have human-like conversations, became the fastest-growing product in history... gaining 100 million users in just two months.
But it’s politically motivated. It talks like a San Francisco liberal... and is often flat-out wrong. To some, it may seem like a silly toy.
None of this matters. The important thing to understand is right now, ChatGPT is the worst it will ever be. These systems only improve over time. And the speed at which these advancements are happening right now is truly incredible...
We saw how rapidly AI systems are improving in the past few weeks.
OpenAI, the creator of ChatGPT, debuted GPT-4 last month. It blows away the first ChatGPT. Researchers put both systems through a gauntlet of real-world tests, including the bar, SAT math, and GRE business school exams.
ChatGPT barely passed most of these tests. GPT-4 blows them out of the water. It scored in the top 10% on the bar exam. That’s remarkable when you consider the chatbot didn’t undergo any specific training.
Bryan Caplan, professor of economics at George Mason University, put ChatGPT through its paces in December. It could only muster a D grade on his midterm.
Not only did GPT-4 score an A on Caplan’s new exam, but it achieved the highest grade of any student!
GPT-4 is the real deal. It’s not your father’s clunky AI.
OpenAI’s new system comes with new capabilities, too. For example, it can understand “context” in images, as shown in this example:
Source: OpenAI
Using GPT-3, UK-based startup DoNotPay built a chatbot to negotiate fines like unwarranted parking tickets for customers. It kind of worked. But the bot often needed manual intervention.
DoNotPay adopted GPT-4 and achieved a major milestone. The bot “talked” to Comcast’s online customer service and managed to save someone $120 on their broadband bill. According to the startup, it’s the first time a bill has been negotiated purely by AI.
How is GPT-4 so superior to its predecessor?
My colleague Chris Wood and I talked about the new “AI law” in this RiskHedge Report. In short, with every 10X increase in the amount of data you feed into an AI system, its quality doubles. So to boost AI performance, researchers are using ever-greater amounts of data.
When it comes to AI, bigger is better.
Researchers didn’t stuff 10X more data into GPT-4. GPT-3 has 175 billion parameters, while GPT-4 has 100 trillion parameters—a roughly 500X increase.
Other AI tools are advancing rapidly, too.
Take image generator Midjourney, for example. Midjourney V1 debuted less than a year ago. Look at how much better V5 is at creating photorealistic images of Mr. Bean:
Source: Wired
AI isn’t some far-off tech with the potential to change the world in 10 years. Real breakthroughs are happening now.
GitHub Copilot is an autocomplete tool for coders built on OpenAI’s tech. Instead of humans having to write every line of code themselves, the software predicts what they’re trying to code and offers suggestions.
Coders using Copilot finished tasks in 55% less time, on average. Imagine a tool which could boost your productivity by 50%. That’s a total game-changer.
I’m predicting there will soon be “Copilots” for lawyers, accountants, doctors, and everyone else. Vast fortunes will be created by companies harnessing this powerful new technology.
Here’s a useful analogy on how to profit from AI…
Refrigeration was one of the most disruptive inventions of the 20th century. Being able to preserve food totally changed how Americans ate and cooked.
But do you know who invented the refrigerator? It was a company out of Detroit, Michigan, called Kelvinator, which you’ve probably never heard of.
Investing in Kelvinator wasn’t the best way to make money from the rise of refrigeration.
Instead, you wanted to bet on upstarts harnessing this powerful new technology—like Coca-Cola (KO), which built a $270 billion empire making cold, refrigerated drinks.
In my Disruption Investor advisory, we’ll continue to invest in companies powering the AI revolution, like Nvidia (NVDA) with its specialized GPU computer chips. We’ll also invest in firms currently using this breakthrough tech to disrupt massive, important industries.
If you missed my talk with Chris Wood on this—“How to Act on the AI Revolution”—you can still watch it right here.
Stephen McBride
Chief Analyst, RiskHedge
In the mailbag…
Over the weekend, we heard from several readers about our recent RiskHedge Report, “How I learned to stop worrying and love AI.”
Here are what a few folks had to say:
I think the risk from AI is that it could lead to political dominance. Kinda like Google achieved in search. Musk knows all the players leading the AI research and probably also knows that a pause is impossible with so much money and power at stake.
At least by proposing the pause, he is positioned with the good guys if the outcome turns bad. All the good stuff you point out—and probably a lot more—is also true, but let's get to nuclear power without the bomb. —Ray
Alec wrote:
I agree with you and also eagerly await the advances AI is going to enable, from biotech to healthcare, software developer to educator, lawyer to executive assistant. And hopefully even fed fund rate setter.
However, I also agree with the pause idea after listening to Max Tegmark on a recent Lex Fridman podcast. The pause is to allow society to start figuring out what sort of “seatbelts and guardrails” this new technology requires to keep all safe.
If AI stays at this inert machine-learning level, then the pause is moot. The problem is IF it makes the jump to AGI and then we have a super-intelligent, conscious, non-human species on our hands. Then it becomes much more than just a better pattern-matcher than humans.
Maybe it will never make the jump to AGI, but six months is nothing in the scheme of things to pause to do some reflecting on the issue just in case we do happen to create an existential threat. Thank you for your optimistic articles. —Alec
Angela said:
Feels like the people asking for a pause have some selfish motive. They are late to the party and will appreciate a pause so that they can catch up. —Angela
Mark wrote:
If we don't use AI to help solve the problem of climate change, we won't be around long enough for AI to harm us—or perhaps AI will put us out of our misery. —Mark
And here’s Dwayne’s take:
AI can and will do great things, BUT we are creating a whole new sentient being that can be controlled by maniacal fanatics, and therein lies the danger. If all people were good, caring individuals, I would have very little concern.
Unfortunately, that’s not the case. It will impact so many things and blur the line between what is real and what isn’t. Its power to control people will be immense. Once the control is established, I see freedoms being violated and a world like 1984 [becoming] a real possibility. It needs immense oversight, and that may not even matter. —Dwayne
