Earlier this week I started seeing Waymo’s doing (manned) test drives around NY which prompted some reflection.
I was wrong about self driving.
For a long time I assumed self driving would never really “happen.” Even if the technology worked in controlled environments, it wouldn’t scale or really matter to our transit system. It would always be 5 years and one milestone away from broad deployment.
It’s well past time to admit I was wrong.
Looking back, I made three interrelated, cascading errors
The tech would asymptote too low to be near perfect in all conditions. Until very recently, it seemed like the edge cases would continue to pose insurmountable barriers. Even if the AVs worked great in environments with year round great weather and relatively minimal pedestrians (suburban Arizona, eg.) they wouldn’t get good enough to go fully driverless in real world, chaotic environments. Instead, I assumed we’d see some really-useful-but-not-earth-shattering deployments in places like retirement communities, college campuses, etc. At this point that is clearly wrong. The technology has gotten much better, much faster than I thought it would/than it seemed like a few years ago.
People would be intolerant of anything but perfection. Robots are scary and people have strong incumbency biases that make them blind to extant, omni-present problems. That is, when a robot has a breakdown or a collision its news even if it happens at much lower rates than human drivers. A single AV collision would be unacceptable, I reasoned. This was partially right (the mere presence of Waymo in a collision frames the whole accident as a Waymo accident) but the technology really is so good that it rarely happens. People might be intolerant of failures but the failures are already rarer than I would have thought possible.
The politics would be impossible for a wide rollout. I believed that the politics would be block by block with NIMBYism, protectionism, and anti-tech activists making the wide deployment functionally impossible. In reality, anti-AV activism has been so poorly executed that it has made itself a losing argument; no one wants to side with people burning cars in the street. Combined with the safety record and the great user experience, there is a growing constituency FOR AVs and limited organized resistance. Waymo in particular has been so slow and measured in their rollout for so long that they have earned the right to now go faster - the opposite of Uber which almost poisoned the well for everyone.
I’m not a car person and AVs are not without risk but electric, autonomous vehicles are so clearly better than our current system. They are demonstrably safer, have perfect compliance with traffic laws and speed limits, and don’t need street parking.
AVs are also the best, most realistic vehicle (pun intended) for rapid, wholesale vehicle electrification: high utilization and low maintenance costs justify the upfront infrastructure investments. Electric fleets are just much easier to manage.
There’s still some major legislative hurdles to clear in NY but it now seems inevitable it will happen in the next year or two.
Getting away from ICEs and towards electric AVs would be a huge win for traffic, air quality, and noise pollution. With Waymo set to launch in NY (I assumed we would be last, not among the first) I can confidently and gladly say I was wrong.
I Read
Vibe Shifts: How growing skepticism of AI can ripple into markets, and why I’m still optimistic. Looking at the bigger picture, the fact that we even have AI this advanced is incredible. I am firmly in the camp that even if AI progress halted with current models, society would have years of innovation and productivity gains as a result of embedding intelligence in every piece of software. But real or not - society’s perception that we are shifting from step-function improvements in model quality/performance, like the jump from GPT-3 to GPT-4, to incremental improvements should be concerning to anyone who works in tech. Why should it be concerning? Because we are in an AI-fueled bubble. And as is the case with all bubbles, the AI bubble is not just confined to the asset experiencing inflating valuations.
A combative case for liberalism and a $4mm investment. I’m excited about the launch of The Argument, which basically seems like The Atlantic but way more online which roughly describes my politics.
How do we live with each other? This is the launch article for The Argument and it’s basically a case for liberalism overall. My shit. The Argument was created to reimagine liberalism for the 21st century, but it was not created to endlessly lecture about political philosophy and the global world order. The battle for America will not be won through historical analogies and 5,000-word journal articles from tenured academics. It will be won if we can convince people that the issues they care about are best addressed through liberalism.
It seems like I was/am right about Britain's future as an English-speaking outsourcing destination for US companies!
This made me laugh and is also enough to make me root pretty hard against Meta’s and XAI’s efforts compared to the other research labs. You don’t have to be aggressively safety-pilled to think that accelerating goonbot waifu slop is bad!
