The AI Abyss Is Not Investable
Why rapid AI progress can be a distraction for our investment research
Time Investment
Wordcount: 1,025
Time: ~4 - 5 minutes
Intro
AI progress is so fascinating that it sucked me into an Abyss.
A dangerous abyss that is sometimes difficult to recognize.
With AI as an example, let me share with you my experience of how we investors should structure our reading habits. Let’s explore how you can be a more ‘efficient’ investor and through it achieve more intentional success.
Lend me your attention for a few minutes and let’s explore it together.
The AI Newsletter
I receive a daily AI newsletter called Superhuman. Like all media, what does a daily newsletter want to do? It wants to have our attention. It wants us to read it every day, click its ads, subscribe to a pro version or whatever the business model may be. But the goal is always attention.
As investors where we place our attention and our research focus is truly, the most important decision we can make. There are seemingly endless possibilities and our time is quite limited.
Back to the AI newsletter.
The incredible progress of AI is so fascinating, because I love intellectual challenges, learning and anything that broadens my knowledge. But this feeling can be a dangerous distraction from the ultimate goal. Especially when like me you wake up a week later you wake up reading four newsletters about AI.
AI Newsletter #1 - Superhuman
AI Newsletter #2 - AI Roundup (just in case there is something different)
Robots Newsletter - Cause, why not?
Tech Roundup - Just in case…. ARGH
The fascination of progress and of new information sucked me into its abyss, because it made me lose track of my actual goal: I want to find investments. To find investments we have to be able to determine what is knowable.
AI Progress Is So Rapid, It is Not Investable
After three weeks of reading I found so many incredible things:
Google developed an AI model to communicate with Dolphins
The Semantic Clinical Artificial Intelligence (SCAI) is outperforming most physicians and doctors. And it isn’t the only tool out there.
An AI trained on billions of protein sequences is developing new proteins that could revolutionize pharma, and science in general
And there is so much more.
Emotional intelligence research from OpenAI.
Autonomous CyberSecurity from Microsoft.
META has a mind reader
META is speeding up access to MRI’s for patients. (quite useful)
On top of that there are so many new generative AI tools that promise to automate your marketing, video creation for ads and your customer service.
We cannot lose sight of the all important question
How can we find a predictable, sustainable cash flow business in this fast paced environment?
The answer is only very few can.
For example, to know which company will end up winning the Protein Synthesis we need to know at least the following:
How does protein synthesis work on a scientific level?
What does AI do that humans do not?
How much is the AI hallucinating?
How can we verify it?
How do we then know which AI is better than another if we’re not doctors?
And even if we know all of this, and are deeply involved, how can we guarantee that our company has a defensible MOAT and another doesn’t come along in six months that is outperforming us?
This is the essence of having a circle of competence. We cannot be experts in everything, so we have to focus our attention to the areas where we can develop an edge.
Know your circle of competence, and stick within it. The size of that circle is not very important; knowing its boundaries, however, is vital. If you are only competent in spots and stay in those spots you can do just fine.
~Charlie Munger
Follow The Waves Of Capital Flows Until You Find The Knowable
To be efficient in our research we need to focus on what we can know. Large technological waves will always show us knowable areas to actually focus our attention. I call it looking behind the scenes of the wave.
Here is how I think about AI:
AI will require Datacenters. The build out has already begun. META, xAI, Microsoft, OpenAI have all pledged or already invested billions into building out the AI infrastructure.
AI will require a lot of power. Microsoft signed a deal to revive a Nuclear Power Plant. META is partnering with the local Louisiana grid to build out renewable energy to power their center.
Where within these larger waves lies sustainable cash flow?
Datacenters
How are datacenters maintained?
Do they require consistent replacement of parts, like racks, piping for liquid cooling? Or something else?
If yes, which companies service the centers?
How does a Datacenter run and operate?
If a Datacenter was a standalone company what are its expenses?
What of these expenses are recurring?
Which companies are the revenue side of the expenses
Energy
How much additional energy demand will there be in the next 5 years?
Which sources of energy are available now?
Nuclear is great, but takes forever
Coal is here now, but it is horrible for the environment
How does our grid handle the additional loads?
How much will natural gas and oil benefit?
How does that affect big companies?
How does that affect small companies?
This is the start of my research framework. Follow the capital flows of the waves of the future.
As we answer these questions we discover more questions. In the process we will find areas of research that we are drawn to and some we are not. That’s fine, the options are endless.
I returned from the Abyss only when I realized that my emotional wiring (drawn into the unknowable by curiosity) drove me to spend too much time on invaluable reading. So I adjusted my strategy. Now I read a weekly newsletter that feeds my general curiosity, while refocusing my energy into researching waves I have more confidence in predicting.
“Now he doesn't read the way we read, you know, he's skimming a lot and he's especially skimming a lot when...he finds the author rambling or whatever… he is trying to get the nuggets he wants
~ Mohnish Pabrai on Charlie Munger's reading habits
My Current Deep Dive List
Datacenters
Energy
Minerals in the Americas
Space
Automated Manufacturing
Stay tuned for posts on each of them. Cu on Tuesday.
Thank you for spending your valuable time reading Intellectual Wandering.
Reading List
Protein Synthesis
https://theaiinsider.tech/2025/01/31/ai-simulates-500-million-years-of-evolution-to-design-shiny-new-proteins/
The AI Scientist
https://www.futurehouse.org/research-announcements/launching-futurehouse-platform-ai-agents