Newsletter

Welcome to Conor, Dom, Diego, Josh, Aidan, Brian, and the rest of my new subscribers.

This post talks a bit about war and how it is bad, why I find Comma.ai’s business model so interesting, a reminder that governments are not doing a great job of protecting us against the biggest threats to humanity, why if you live until 2050 you might live forever, and ChatGPT, which is the coolest thing I’ve found in years.

Enjoy.

Despite the US spending well over a trillion dollars in Afghanistan, people are starving and literally selling organs to survive. Never believe leaders who claim that invasions are in the interest of the invaded.

https://www.bbc.com/news/world-asia-63733683

Really interesting blog from George Hotz comparing his start-up to other self-driving car start-ups. His company comma.ai made a device that you can plug into most cars to let them drive themselves. Comma’s competitors (Zoox, Cruise, Waymo) are building their own vehicles designed to be fully autonomous with no steering wheel. Comma isn’t concerned (yet) with building cars, and their manufacturing expertise is limited to a small box that plugs into customers’ cars. Comma’s competitors have undertaken a huge manufacturing task, building an entirely new type of car from scratch. Comma has raised <$20m, their competitors have all raised >$1b. Comma has <50 employees, their competitors have 1-3k. Comma is profitable and has thousands of paying customers, their competitors are burning billions every year.

In terms of performance, in most cars Comma’s technology is on par or better than their competitors, including Tesla, and much more scalable because it doesn’t rely on hardcoded features or local maps, just its neural net/ai.

Comma is making a bet that it can develop self-driving software that is so good other car companies need to make their cars compatible with Comma’s technology to attract customers, before these car companies can develop the technology themselves. It’s also betting that its software will be good enough to eventually operate fully autonomous cars with no drivers almost anywhere in the world, before their competitors take over the autonomous driving market.

I like Comma’s business model. It’s super lean, profitable, with huge potential. It reminds me of this list of small teams that have achieved huge things. It’s possible because software has 0 marginal cost and is infinitely scalable. A brilliant software engineering team is better than 1,000,000 mediocre software engineers. The easier it is to scale, the more the market becomes winner-takes-all. It’s pretty hard to scale a dentist office, so there are hundreds of them just in Ireland. It’s super easy to scale software, so there’s only a handful of competitive search engines, maps, social networks, or spreadsheet programs in the entire world. If George Hotz can pull this off, he will be a very wealthy man.

If you haven’t already, check out Patrick Collison’s website. I love it.

I have 2 new posts up summarizing a book I read about Amazon 1994-2013. Amazon is quite a cool company, and one of the best-run organizations in the world. Not particularly fun if you’re an employee or vendor, but great for making money and delighting customers. Read part one here or part two here.

We have known about the risk of pandemics for thousands of years, with the black death, Spanish flu, and more recently with Ebola and swine flu. Yet, when we were hit with COVID-19 we had pretty much no plan, no spare ventilators, no ready supply of face masks or hand sanitizer, and no plan to increase ICU capacity rapidly. Some of the countries who were hit badly by SARS, such as Taiwan and Singapore, were able to deal with the disease much better, which shows that we humans usually only learn the hard way, and our governments are reactive, not proactive. As technology advances, we might not survive long enough to learn the hard way. We have very few missile defence systems which nobody really cares about, but if the US was struck by some nuclear missiles they would probably react by spending a lot more on missile defences. Unfortunately, in the event of an all-out nuclear war society would probably collapse and we wouldn’t get the chance to build back better. The fact that we all exist on a single planet is a single point of failure for humanity and underscores the importance that humanity becomes a multi-planetary species with a self-sustaining society on Mars and other planets.

The concept of longevity escape velocity is really cool. It’s basically the idea that we will reach a point where new technologies are becoming available that extend our lifespan faster than we can age. If you want to live forever you just need to survive until we hit this point and pray you don’t get hit by a car in the meantime.

If you haven’t check out ChatGPT yet I highly recommend you do. I now use it almost as much as I use Google, it’s so useful. Below are a few tidbits of what it’s capable of:

I tested out this code and it wasn’t perfect, but after tweaking it a little bit I was able to make it work. This took 5 minutes total and is about 5x quicker than it would’ve taken me to build it from scratch.

I can imagine this working similarly for law, engineering, and a ton of other stuff.

It’s not great writing jokes though.

I even read a blog post about a guy who managed to basically build an entire virtual PC inside ChatGPT. It’s wild.