Discover more from Fallacy Alarm
Real World AI Computing as the next big innovation platform
AWS meets Apple App Store
Innovation events as major wealth creation catalysts
These are challenging times for investors. The more stress we have in financial markets, the more short term trends overshadow long term trends. There are some market participants who thrive in this as they can consistently generate profits by anticipating short-term cyclical trends. I am not one of those. I can see the arguments for oil going to $200 and to $50 and I have no idea which side is more compelling. Likewise, I can see the argument for 10y treasury yields hitting 5% next year, while I can also see the case for <2%. Again, I have no conviction on either side.
What I think I can do, is to assess what happened in the past and make inferences to today. What looking at the past is telling us is that the really big wealth creation moves in history have come from big innovation events, most of which share two charming features:
You only have to identify it correctly once and then ride it for a decade and more, which makes it in my view much easier than assessing cyclical trends, which you have to get right again and again to stay ahead.
You usually have a lot of time jumping on the right train even after the innovation is unveiled and shows some initial commercial success, which makes it much easier as a retail investor to compete with the big Wall Street machines.
Let’s look at three examples from the past to set the foundation for today’s assessment:
Mobile Computing (aka the Smartphone)
Apple unveiled the iPhone in January 2007. Since then the stock price is up 50x. And the innovation cycle of the smartphone is still unfolding. Think for instance about recent developments into making it a payment terminal and the integration into car as recently presented in their developer conference. The essence of Apple is still the milking of that 2007 innovation. Everything that came thereafter is in some for related to the iPhone (iPad and wearables for instance).
Amazon developed AWS in the early 2000s. In 2002, Bezos wrote his famous email mandating the use of service interfaces for internal communication which laid the groundwork for their AWS ecosystem. Please see a summary of that email below sourced from here:
It is difficult to pick a date when the public finally became aware of the power of AWS. It was first publicly acknowledged in this blog post in 2004. The first features were publicly launched in 2006. Since then the stock has done a 60x.
Microsoft released Windows in 1985, which went on bringing computing into people’s homes over the coming decade.* Microsoft Office was introduced in 1990. Excel followed in 1992 and is probably the most powerful application Microsoft ever developed in terms of its impact on the business world. Check out this hilarious commercial from the Excel launch:
Since 1992, MSFT is up 100x.
From past innovations to tomorrow’s innovations
What the examples above have in common is that the reference dates (Apple in 2007, Amazon in 2006, Microsoft in 1992) marked points in time where it would have been theoretically possible for an outside investor to assess the innovation and see its long term potential. The earlier, the more challenging to actually make that bold long term call. But there was plenty of time along the way to jump on the train as the innovation cycle unfolded.
So, what are the big innovation leaps coming at this point in time? To me, Real World AI stands out.
What is RWAI Computing?
We have seen tremendous technological progress over the last three decades in information technology, be it mobile computing or cloud computing or personal computing. What all of these have in common is that most action happens digitally, i.e. we shipping bits and bytes around, which has created enormous economic impact and explains the performance difference between US stocks and European stocks over the past decades to a good extent.
One frontier that seems like an obvious next step that will come at some point is the liberation of artificial intelligence from the cyberspace to real space, i.e. make programs/applications move through our physical world like they are moving through the digital world.
The questions are a) by when will RWAI Computing become a thing and when will it become an investable theme and b) who will be the likely winner(s)? As I will elaborate on below, I believe the answers to these questions are ‘now’ and ‘Tesla’.
On September 30, 2022, Tesla will likely unveil the humanoid robot prototype:
While many people are excited for this event and some are scared it will not live up to its expectations (if Elon unveils something less exciting than the Boston Dynamics robots for instance), I am most interested in Tesla’s long term development plan beyond the initial prototype, similar to how they laid out their robotaxi business model as an ‘AirBnB meets Uber’ during their Autonomy Day in 2019.
RWAI Computing will unleash an enormous innovation cycle that will quickly outclass the impacts of the personal/mobile/cloud computing revolutions. In addition, its scope will be so vast, that it will be an exercise for many parties, not just the company (or companies) developing this new computing platform. Think about: Virtually any economic activity will over time become an application, whether it is digitally or physically. Just like streaming and banking apps have disrupted cinemas and bank branches, RWAI applications will soon be available for waitressing, cashiering, cleaning or managing warehouses.
What I envision is a mix of AWS and the Apple App Store on steroids. There will be companies providing the hardware and a generalized software platform to operate said hardware based on which third parties can develop new RWAI applications. In Tesla’s case, this will happen with its Dojo supercomputer (Dojo as a Service).
Why is Tesla in the pole position for RWAI Computing?
Whoever has tried to learn something new will confirm that the fastest progress is made when they actually work on a problem they are trying to solve. Developing beats researching. Practice exams beat textbook reading. Nobody (not even Google) is working on hands-on RWAI applications to the extent Tesla does. Nobody tackles a bigger problem. And nobody has the resources (i.e. data) that Tesla is able to leverage.
As they are working their way through this challenge, they are making ground-breaking discoveries and developing best practices neural net architecture. This is evident when Elon talks about the FSD program, for instance when he mentions all the new terms they introduce when they describe the processes that they are building. Check out Elon’s comments in the interview below between 1:02h and 1:36h:
Here are some of the highlights:
“We've re-architected the neural nets in the cars so many times, it's crazy. We started off with simple neural nets that were basically image recognition on a single frame from a single camera, and then trying to knit those together […].
In fact, we've just recently done a new rev on the C compiler that will compile directly to our autopilot hardware. […] There's a lot of hardcore software engineering at a very sort of bare metal level. 'Cause we're trying to do a lot of compute that's constrained to the our full self-driving computer. And we wanna try to have the highest frames per second possible in a sort of very finite amount of compute and power. We really put a lot of effort into the efficiency of our compute. So there's actually a lot of work done by some very talented software engineers at Tesla that at a very foundational level to improve the efficiency of compute and how we use the trip accelerators, which are basically doing matrix math, dot products, like a bazillion dot products. And it's like, one of our neural nets is like, compute wise, like 99% dot products.
Over time there's less and less conventional software, more and more neural net. […] One of the big changes will be, right now the neural nets will deliver a giant bag of points to the C++, or C and C++ code. We call it the giant bag of points. Then you've got to assemble this giant bag of points in the C code and turn it into vectors. And it does a pretty good job of it, but it's, we wanna just, we need another layer of neural nets on top of that to take the giant bag of points and distill that down to a vector space in the neural net part of the software, as opposed to the heuristics part of the software. This is a big improvement.”
Elon Musk to Lex Fridman
Tesla in the early 2020s is building a new type of RWAI development platform which will be the basis and the gold standard for many subsequent projects. Tesla has important technology in everything needed to develop RWAI applications (battery technology, electric powertrains and actuators, visual neural net systems) on top of their current FSD program. Most importantly the humanoid bot. But they will eventually license their neural net training program to other companies. What AWS is for Internet AI, Dojo will be for RWAI.
AI/Deep Learning will be the largest driver for value creation in the 2020s and 2030s. It will dwarf what the Internet has done, both in terms of time and extent of its impact. The biggest AI/Deep Learning companies of today are sitting on top of the Internet as the base technology. Tesla’s FSD program is the first push to bring AI from the Internet into the Real World. It is a gate opener. The first instance of the liberation of AI from cyberspace to real space. It is a new frontier of digitization.
If the entire Big Tech have been built on the Internet and if RWAI will be multiple times bigger than the Internet and if Tesla captures a disproportionate share of the value creation, how freaking big can Tesla become?
*To be fair, Apple bet them in a narrow race when they launched the Mac in 1984, but I believe it is fair to use Microsoft here as the example for the personal computing revolution given they reached much greater scale subsequently and Apple’s 2nd big innovation event, mobile computing, today accounts for most of the company’s market cap today.