Charlie StrossFeb 27, 2024

Apple Cancels Work on Electric Car, Ending Decade long Effort

The most recent approach discussed internally was delaying a car release until 2028 and reducing self-driving specifications from Level 4 to Level 2+ technology.

Many employees on the car team — known as the Special Projects Group, or SPG — will be shifted to the artificial intelligence division

https://www.bloomberg.com/news/articles/2024-02-27/apple-cancels-work-on-electric-car-shifts-team-to-generative-ai

Apple to Wind Down Electric Car Effort After Decadelong Odyssey

Apple Inc. is canceling a decadelong effort to build an electric car, according to people with knowledge of the matter, abandoning one of the most ambitious projects in the history of the company.

Bloomberg
Show threadCharlie StrossFeb 27, 2024
It's a classic case of Christensen's "The Innovator's Dilemma". Also, it's REALLY HARD to make a breakout automobile. (Tesla almost did, but screwed the pooch with bad quality control, poor ergonomics, and toxic company culture.) And even if you do, Ford/GM/Toyota/VW will catch up within a decade and eat your lunch. It's why most auto startups seem to be hypercars with a 7-digit price tag—there's a ridiculous profit margin—but production volume is too tiny for Apple to touch.
Show threadTonyC
@cstross I wonder if they thought they could license the software to other auto makers? The car would be a proof of concept.
Feb 27, 2024 at 8:33pmWeb
Show threadCharlie StrossFeb 27, 2024
@phlebas That was my thought too. But scaling back from self-driving level 4 to 2+, when the likes of Mercedes and Tesla are at 2 already ... oof.
Show threadTonyCFeb 27, 2024

@cstross One thing I found out today is that running enough computers for level 4 self-driving *halves* the endurance of an electric car's battery.

https://www.bloomberg.com/news/newsletters/2024-02-27/autonomous-electric-vehicles-will-guzzle-power-instead-of-gas

Autonomous Electric Vehicles Will Guzzle Power Instead of Gas

Limiting the battery drain associated with highly automated computing hardware will be a key challenge for car engineers.

Bloomberg
Show threadQueFeb 27, 2024
@phlebas @cstross it’s not computers, its servers worth of capacity in true self driving solutions. The heat and power loads are material.
Show threadNemoFeb 27, 2024

@phlebas
Laughing like a drain.

Nope. This is perfect. There's just nothing to add to that.
@cstross

Show threadVery Human RobotFeb 27, 2024
@phlebas @cstross
Consider 100 kWh car, 300 mile range, 75 miles per hour.
If I'm using 25 kW on average for the motor (freeway driving) that's 25 kW for self driving.
That's a lot of compute and a lot of heat to somehow exhaust.
Using 50 kWh for two hours of compute?
That doesn't quite pass the sniff test.
Show threadJonFeb 28, 2024
@phlebas @cstross be cautious about that. It's "up to 46%". EV power use of ca. 0.15 kWh/km means that such a computer would have many times the power consumption of a high-end gaming rig (unless you're stuck in traffic for hours, not moving). The cite for that graph is just "Bloomberg NEF" which is a self-hosted "research" operation and I cannot find such a study on their website, so cannot check their assumptions.
Show threadronbrinkmannFeb 28, 2024
@phlebas @cstross yeah I’m VERY skeptical of this estimate. Feels like Bloomberg making shit up (again).
Show threadCharlie StrossFeb 28, 2024
@ronbrinkmann @phlebas Their estimate is probably missing a footnote: "* power consumption estimate based on in-vehicle use of 2004-vintage Intel Itanium CPUs, other platform power consumption may vary."
Show threadTonyCFeb 28, 2024
@ronbrinkmann @cstross It does seem high. I have no way of saying whether or not it is excessively so.
Show threadTonyCFeb 27, 2024
@cstross I wouldn't be at all surprised to find that Apple decided that level 4 wasn't attainable.
Show threadAngus McIntyreFeb 27, 2024
@phlebas @cstross All you need is a space under the hood large enough for a legless Polish master driver and a candle, and the problem of FSD is solved, Mechanical Turk-style!
Show threadTonyCFeb 27, 2024
@angusm @cstross I think it was @pluralistic who pointed out that one company's experiment with "driverless" taxis involved two people per vehicle at the monitoring station.
Show threadTonyCFeb 27, 2024

@angusm @cstross @pluralistic

Ah, it wasn't quite two:

"Cruise, GM's disgraced "robot taxi" company, had 1.5 remote operators for every one of the cars on the road. They used AI to replace a single, low-waged driver with 1.5 high-waged, specialized technicians. Truly, it was a marvel."

https://pluralistic.net/2024/01/31/neural-interface-beta-tester/#more-7904

Pluralistic: Three AI insights for hard-charging, future-oriented smartypantses (31 Jan 2024) – Pluralistic: Daily links from Cory Doctorow

Show threadSpeakerToManagersFeb 28, 2024
@phlebas @cstross
It’s been evident for awhile that none of the companies trying to do it have a development path to level 4. Even complete level 3 is uncertain. We’d have been a lot better off putting those billions of dollars into mass transit.
Show threadCharlie StrossFeb 28, 2024
@SpeakerToManagers @phlebas Yes, and also e-bikes and changing urban infrastructure to be more pedestrian/bike/e-bus friendly. Which means reducing car numbers and use. Cars are the problem: tweaking the problem merely gives you a different version of the problem.
Show threadFeralRobotsFeb 27, 2024
@cstross @phlebas
Before you mentioned it I hadn't heard that framing, & it made me think 'did they figure something out that all the other vendors haven't been willing to admit?'