Self-driving car talk – 2019

We’re a decade from self-driving cars. When I write self-driving I mean a car that drives you around while you sleep, which is level-4 and level-5. Since I think we have 10 years until that, in the meantime let’s talk about level 1, 2 and 3 in a granular way.

There are 5-levels on the way to self-driving nirvana.

  • 1995-2015 was about level-1
  • 2015-2025 will be about commercializing level 2 –– Tesla and Cadillac are already at level-2 using RADAR.
  • 2022-beyond will be about Level-3 using LIDAR, and it will be slow hill climbing from there to level-4 and level-5. Get comfy for the road-trip to self-driving.

Level 1 (1995-2015) = one of the following but not both: traffic aware cruise control–TAAC (e.g. cars with) or lane keeping (e.g. cars with)
Level 2 (2015-2025) = both TAAC and lane keeping with nags, we’ll be stuck here until 2025 because of the $50,000 incremental cost per car for LIDAR. Tesla and Cadillac have this. Eventually every car will.
Level 3 (2025-beyond) = both TAAC and lane keeping without nags. Tesla and Cadillac can do this now, but not without killing people. Thus they’ll probably not be granted this without LIDAR. Trouble is, LIDAR makes cars $50,000 more expensive. Many people disagree that LIDAR is a requirement for level 3, including Elon Musk and my brother in-law – both of whom think LIDAR is a crutch until machine learning algorithms can extract safe self-driving cues from RADAR hardware. In either case, it seems that the first cars with level-3 will be Lyft-like services like Cruise and Waymo, not cars you can purchase like Tesla.
Level 4 = both TAAC and lane keeping without nags and the driver can sleep, but only on certain roads (e.g. either divided-controlled-access-highways or urban roads, not both)
Level 5 = same as above, but for all roads

Below is a helpful diagram from the Society of Automotive Engineers.

Continue reading “Self-driving car talk – 2019”


In 2021 pieces of Mars will land safely on earth as part of Mars 2020. The Mars 2020 rover collects samples and leaves them in canisters on the surface. The lander deploys a fetch rover to collect the samples and deposit them in an ascent vehicle, which blasts into Mars orbit. There, a return orbiter collects the samples for transport back to Earth.


Mars Sample Return overview infographic (ESA)

SpaceX Falcon 9: How Elon Musk’s Rocket Is Winning the Reusability Race

This wonderful illustration from July 2019 National Geographic explains one of Elon Musk’s greatest space innovations – rocket reuse.

Adding reusable technology reduces the payload and cost. In order to make the Falcon 9 reusable and return to the launch site, extra propellant and landing gear must be carried on the first stage, requiring around a 30-percent reduction of the maximum payload to orbit in comparison with the expendable Falcon 9.

With full reusability on all three booster cores, the Falcon Heavy will lift approximately 18,000 lb to geosynchronous transfer orbit at a cost of $4200/pound. The ultimate goal with the development of SpaceX is to bring the cost down to $500/pound, which is believed to be possible only with rocket reuse.

Here’s another space race illustration from the paper edition of National Geographic, which you can get delivered to your door for $20/year.

In-depth– Spend analysis of 6 years 686 rides and $12,041 spent on Uber

What I learned from my Uber data

I save $703/month using Uber versus owning 2 cars (🚙🏎). I’ve spent $12,041 on 686 Ubers over 6 years. By comparison, car ownership cost $11,783 per year, most of it depreciation. What could $703/month buy you? Well, four years of $703/month and you’ll save enough to climb Mount Everest 🏔 unguided. Or you could buy a shiny new laptop every 3 months. More details below. Continue reading “In-depth– Spend analysis of 6 years 686 rides and $12,041 spent on Uber”

eBay Adds to Machine Learning Hype

Earlier this week I blogged about 5 simple questions you can ask to determine AI hype.

Today I saw something with more hype than I would expect from a respected tech company. “eBay determines this price through a machine learned model of the product’s sales prices within the last 90 days.”

In my opinion this price prediction is not machine learning. It’s just math. It’s not a machine learned anything. It fails 1-5 of Stephen’s tests.