Maybe cars can teach themselves to drive in the more structured states (the MANIAC book)

I recently finished The MANIAC, a concise novelized biography of John Von Neumann bizarrely bolted onto a history of computer programs that dominate chess and go. Somehow the combination works! What I hadn’t realized was how quickly programs that play chess and go can evolve when entirely freed from human guidance. Apparently, in a matter of just a few hours, a program can go from knowing almost nothing about chess other than the basic rules to being able to beat a grandmaster.

This kind of success has famously eluded those who promised us self-driving cars. We’ve gone from failing via humans encoding rules to failing via AI-style training sets of good driving and bad driving (coded by people in India? if you’ve ever been to Delhi or Mumbai maybe that explains the failure). Benjamin Labatut (the MANIAC author) reminds us that when the situation is sufficiently structured computers can learn very fast indeed.

Returning from a helicopter trip from Los Angeles to Great Barrington, Maskachusetts, my copilot commented on the chaos of road markings as we entered Cambridge. “Are there three lanes here or two?” he asked. This is a question that wouldn’t be posed in most parts of Texas or Florida, I’m pretty sure, and certainly not on the main roads of the Netherlands or Germany. Instead of the computer promising to handle all situations, I wonder if “full self-driving” should be targeted to the states where roads are clearly structured and marked. Instead of the computer telling the human to be ready to take over at any time for any reason, the computer could promise to notify in advance (via reference to a database, updated via crowd sourcing from all of the smart cars) that the road wasn’t sufficiently structured/marked and tell the human “I won’t be able to help starting in 30 seconds because your route goes through an unstructured zone.” The idea that a human will be vigilant for a few months or even years waiting for a self-driving disconnect that occurs randomly seems impractical. The MANIAC suggests that if we shift gears (so to speak) to redefining the problem to self-driving within a highly structured environment a computer could become a better driver than a human in a matter of weeks (it takes longer to look at videos than to look at a chess or go board, so it would be weeks and not hours). We might not be able to predict when there will be enough structure and enough of a data set and enough computer power for this breakthrough to occur, but maybe we can predict that it will be sudden and the self-driving program will work far better than we had dreamed. The AI-trained chess and go systems didn’t spend years working their way into being better than the best humans, but got there from scratch in just a few hours by playing games against themselves.

Regardless of your best estimate as to when we’ll get useful assistance from our AI overlords, I recommend The MANIAC (note that the author gives Von Neumann a little too much credit for the stored program computers that make the debate regarding self-driving possible).

Separately, based on a visit to the Harvard Book Store here’s what’s on the minds of the world’s smartest people (according to Harvard University research)..

Full post, including comments

Meet at the Miami Grand Prix (F1 race)?

Who wants to meet at the Miami F1 race, sort of, on Saturday, May 5? This is one day before the “big race” that rich people are desperate to see. There are a lot of events throughout the day, including a spring race, the F1 Academy (from which 73 genders recognized by Science are excluded and this exclusion is the epitome of social justice), and some Porsche racing.

The official F1 site sells only 3-day tickets. A lot of buyers, however, don’t want to show up on Friday and Saturday. Individual days thus show up as verified resale on Ticketmaster. A “campus pass” that lets you walk around is about $120 and a ticket in a grandstand is $200-300 (this also includes the right to wander).

Given the often-brutal Miami heat and sun, I picked tickets in the Turn 18 grandstand. This has its back to the sun and rows beyond about N are shaded (our tickets are in Q). [Post-race update: M is the best row! It is in front of the columns that hold up the shade roof but is still completely shaded. Some of the rows below M also are well-shaded, particularly on the west side of the Turn 18 grandstand, as the afternoon develops.] It has views of cars braking out of the longest/fastest straight and then navigating a couple of turns. The one knock against this grandstand is that it might be a long walk to the Fountains and Promenade areas. The Marina grandstands might be better for taking that one perfect photo of a race car in front of boats, but I don’t think the cars are moving as fast in this area.

It looks as though parking passes are sold on SeatGeek and VividSeats. I’m thinking that traffic won’t be terrible on Saturday because people will arrive gradually and also leave gradually depending on which of the events they’re interested in. On Sunday, by contrast, there is literally nothing on the schedule after the 6 pm finish of the Grand Prix per se (which I’d rather see on television so that I would have a chance of understanding it) and, therefore, there will be a mad rush for the exits.

Full post, including comments

Top Gun slows down to 25 mph (across Florida by EV)

“I feel the need for speed,” said my friend who flew the F/A-18 for the U.S. Navy to the Hertz folks in Tampa. They rented him a Volvo XC40 EV, purportedly capable of 293 miles of range. It’s 190 miles and 3 hours to get from Tampa to our neighborhood in Jupiter, according to The Google. It took him more than 7 hours, an average speed of 25 mph. He found various chargers along the way, e.g., from ChargePoint, but they often charged at only 6.5 kW, less than what you’d need to feed a climate-safe induction range.

Adding insult to injury, as he watched the divergence between miles to destination and miles to empty he was treated to an NPR broadcast about the irrationality of range anxiety. As with inflation, if Americans were only a little more intelligent they would see everything the same way that the D.C. ruling class does.

Perhaps there would have been some way for him to plan the trip around the availability of high-speed chargers, but that’s in a theoretical world, not our real world. His take-away: “I will never buy an electric car.”

Full post, including comments

Should Joe Biden fund an electric food truck startup?

Our current central planners love electric vehicles. “Slow Rollout of National Charging System Could Hinder E.V. Adoption” (New York Times):

More than two years ago, lawmakers approved billions of dollars [at least $5 billion] to build out a national electric vehicle charging network in the hopes of encouraging more drivers to switch to cleaner cars. The money, included in the bipartisan infrastructure law, was intended to help assure drivers they could reliably travel longer distances without running out of power.

But a robust federal charging network is still years away. Only two states — Ohio and New York — have opened any charging stations so far.

Central planners also love asylum-seekers and other immigrants without educational credentials. One popular job for low-skill immigrants is working in a food truck.

Our HOA sometimes hires a food truck for events on the common grass field. The sound of the truck’s generator is clearly audible and it would be much nicer if the truck were powered from a big EV battery.

What if the central planners in Washington, D.C. could be convinced to ladle out the taxpayer cash to a food truck EV startup?

What are the engineering requirements? A Google search shows that 5,000 watts (roughly 44 amps at 115V) is a good ballpark for the maximum draw of a typical food truck. Maybe the average load would be 2,500 watts and that needs to be supported for at least 6 hours, which implies a 15,000 watt-hour battery (15 kWh). The standard Tesla Cybertruck has a 123 kWh battery, so that should get the food truck to the site, run for two shifts if necessary, and get the food truck home. (The Ford F150 comes in 98 and 131 kWh “usable capacity” versions.)

In addition to saving the planet and providing jobs for asylum-seekers who wish to work, the Bidentrucks would reduce the ambient noise levels in our cities, which is an important equity issue. (“Noise pollution more common in communities of color and racially segregated cities” (Harvard 2017))

Maybe this all-electric truck wouldn’t help Greta Thunberg with her #FreePalestine goal, but it could save humanity from extinction.

Full post, including comments

Costco Auto Program: mostly a scam

Our beloved 2021 Honda Odyssey’s lease expires in January 2024. Due to the mostly peaceful inflation that the government says does not exist, leasing a replacement would cost nearly 2X what we’ve been paying on our lease whose built-in interest rate is close to 0%. I’m trying to decide whether to buy the car from Honda for about $25,000 or trade it in on a new one (where “new” means “identical to the 2021 version”). I requested a quote from the Costco Auto Program, figuring that the result would be a simple fraud-free number. Here’s what came back from the local dealer that is the Costco affiliate:

The dealer adds in about $4,000 in profit via some worthless accessories for $4,225 and then tacks on $999 as a straight-up “dealer fee” in addition to an “agency fee” of $99 (either of these could be $5,000, right?). On the third hand, there is a note about the accessories being discounted by 75 percent and a bizarre calculation that adds up to more than $73,000 (Cybertruck territory!).

What is the value of going through Costco if the result is having to sort through this multi-layered fraud and being delivered a car that has been disfigured by the dealer? It looks as though Costco negotiates a discount off MSRP and then the dealer is free to add back in as much profit as it wants to with accessories and fees. Costco could negotiate a price of $10 for the Odyssey and the dealer would still be able to charge $45,000 or $75,000 or whatever other price it felt like.

Full post, including comments

Unlimited EV charging for $372 per year

Florida Power & Light is our remarkably entrepreneurial local electric utility. They may have purchased a mailing list of “total douches very likely to own a Tesla” because they recently sent me an email trying to sell an unlimited charge-at-home plan for EVs:

If they don’t need to pull a permit and run wires inside the garage, in other words, it is $372 per year for all of the electricity that you can fit into your Cybertruck or F150 Lightning. That’s an interesting way to help sell the idea of EVs.

This is on top of the somewhat fake SolarTogether program for virtue signalers who want to say that they’re using all solar power at home.

(Of course, I immediately signed up for SolarTogether. It’s a small price to pay for the virtue points!)

If we did have a Tesla and if we could fit it into our 2-car garage (mostly filled with bicycles, shelves, boxes, an heirloom sports car), I would definitely sign up for this new FPL scheme.

Full post, including comments

Cybertruck history from the Elon Musk biography

The Cybertruck is officially launched. I wrote about this product in 2019:

With today’s Teslas, one expects the driver to emerge and deliver a lecture on climate change, the merits of Elizabeth Warren, unions (except at Tesla itself), and a larger government, etc. What would the image be of someone who drives what looks like a high school kid’s first SolidWorks project?

How did they pick the name? This is for people who used to say “I’m using the Interweb”?

How are they going to deliver this car at $40-50k profitably? Isn’t stainless steel expensive?

(The announced 2023 base price is 61,000 Bidies; it is fair to say that this is the same or less than 50,000 2019 dollars?)

What does the Elon Musk book describe? A 2018 design session:

The Chevy Silverado was still on the showroom floor for reference. In front of it were three large display boards with pictures of a wide variety of vehicles, including ones from video games and sci-fi movies. They ranged from retro to futuristic, sleek to jagged, curvaceous to jarring. With his hands casually in his pockets, von Holzhausen had the easygoing and loose-limbed manner of a surfer looking for the right wave. Musk, arms akimbo, was coiled like a bear searching for prey. After a while, Dave Morris and then a few other designers wandered in. As he looked at the pictures on the display boards, Musk gravitated to the ones that had a futuristic, cyber look. They had recently settled on the design for the Model Y, a crossover version of the Model 3, and Musk had been talked out of some of his more radical and unconventional suggestions. Having played it safe with the Model Y, he did not want that to happen with the design of the pickup truck. “Let’s be bold,” he said. “Let’s surprise people.”

Every time someone would point to a picture that was more conventional, Musk would push back and point to the car from the video game Halo or in the trailer for the forthcoming game Cyberpunk 2077 or from Ridley Scott’s movie Blade Runner. His son Saxon, who is autistic, had recently asked an offbeat question that resonated: “Why doesn’t the future look like the future?” Musk would quote Saxon’s question repeatedly. As he said to the design team that Friday, “I want the future to look like the future.” There were a few dissenting voices suggesting that something too futuristic would not sell. After all, this was a pickup truck. “I don’t care if no one buys it,” he said at the end of the session. “We’re not doing a traditional boring truck. We can always do that later. I want to build something that’s cool. Like, don’t resist me.”

(maybe Musk was right on the highlighted point; Ford and GM are selling traditional-looking electric pickups and demand is weak)

What did the in-house design nerds think in 2019?

When Musk walked in the door leading from the SpaceX factory, his reaction was instantaneous. “That’s it!” he exclaimed. “I love it. We are doing that. Yes, this is what we are going to do! Yes, okay, done.” It became known as the Cybertruck. “A majority of people in this studio hated it,” says von Holzhausen. “They were like, ‘You can’t be serious.’ They didn’t want to have anything to do with it. It was just too weird.” Some of the engineers started working secretly on an alternative version. Von Holzhausen, who is as gentle as Musk is brusque, spent time listening carefully to their concerns. “If you don’t have buy-in from the people around you, it’s hard to get things done,” he says. Musk was less patient. When some designers pushed him to at least do some market testing, Musk replied, “I don’t do focus groups.”

An Australian hater:

Full post, including comments

Shifting gears: Why Tesla’s previous failures in Full Self-Driving might not predict future failure

From Elon Musk, the book:

Almost every year, Musk would make another prediction that Full Self-Driving was just a year or two away. “When will someone be able to buy one of your cars and literally just take the hands off the wheel and go to sleep and wake up and find that they’ve arrived?” Chris Anderson asked him at a TED Talk in May 2017. “That’s about two years,” Musk replied. In an interview with Kara Swisher at a Code Conference at the end of 2018, he said Tesla was “on track to do it next year.” In early 2019, he doubled down. “I think we will be feature complete, Full Self-Driving, this year,” he declared on a podcast with ARK Invest. “I would say I am certain of that. That is not a question mark.”

So they’ll fail again in 2024? Maybe not.

For years, Tesla’s Autopilot system relied on a rules-based approach. It took visual data from a car’s cameras and identified such things as lane markings, pedestrians, vehicles, traffic signals, and anything else in range of the eight cameras. Then the software applied a set of rules, such as Stop when the light is red; Go when it’s green; Stay in the middle of the lane markers; Don’t cross double-yellow lines into incoming traffic; Proceed through an intersection only when there are no cars coming fast enough to hit you; and so on. Tesla’s engineers manually wrote and updated hundreds of thousands of lines of C++ code to apply these rules to complex situations.

C++?!?! Seriously?

According to the book, Tesla is shifting to a ChatGPT-style machine learning approach:

“Instead of determining the proper path of the car based only on rules,” Shroff says, “we determine the car’s proper path by also relying on a neural network that learns from millions of examples of what humans have done.” In other words, it’s human imitation. Faced with a situation, the neural network chooses a path based on what humans have done in thousands of similar situations. It’s like the way humans learn to speak and drive and play chess and eat spaghetti and do almost everything else; we might be given a set of rules to follow, but mainly we pick up the skills by observing how other people do them. It was the approach to machine learning envisioned by Alan Turing in his 1950 paper, “Computing Machinery and Intelligence.”

By early 2023, the neural network planner project had analyzed 10 million frames of video collected from the cars of Tesla customers. Does that mean it would merely be as good as the average of human drivers? “No, because we only use data from humans when they handled a situation well,” Shroff explains. Human labelers, many of them based in Buffalo, New York, assessed the videos and gave them grades. Musk told them to look for things “a five-star Uber driver would do,” and those were the videos used to train the computer.

During the discussion, Musk latched on to a key fact the team had discovered: the neural network did not work well until it had been trained on at least a million video clips, and it started getting really good after one-and-a-half million clips. This gave Tesla a huge advantage over other car and AI companies. It had a fleet of almost two million Teslas around the world collecting billions of video frames per day. “We are uniquely positioned to do this,” Elluswamy said at the meeting.

Despite grand claims by academics seeking funding, rules-based AI generally failed to do anything interesting or practical from 1970-2010 (see MYCIN and CADUCEUS, for example). Statistical approaches to AI, however, began to deliver useful systems, e.g., for speech recognition, starting around 2010.

How Tesla describes the future:

FSD would provide a huge lifestyle boost here in South Florida where there are a lot of 1- and 2-hour drives that lead to interesting places, such as parks, cultural events, theme parks, etc. The drives themselves, however, are boring: straight highways, a lot of traffic close to Miami and Orlando. FSD should work quite well. FSD would also be good for getting to/from international airports. There are a lot more flights from FLL and MIA than from PBI, which is closer to our house, but with a self-driving car it might become more sensible to fly out of farther-away airports.

Full post, including comments

How is Rivian doing?

Back in November 2021, I asked “What edge does Rivian have in the truck or EV market?” and questioned the company’s stratospheric market cap. It has been two years. How is the company doing and how is the stock doing?

Given the calculation that working class subsidies to elite owners of EVs are $50,000 per vehicle (direct tax credits, higher costs for gas-powered cars due to EV percentage sales requirements, subsidized electricity), the company itself should be profitable. MotorTrend says otherwise: “Rivian Loses a Huge Amount on Every Vehicle It Sells” (October 5, 2023).

From May 2023, in the lower Manhattan neighborhood favored by elites (Chelsea):

Full post, including comments

Tesla road trip from New Jersey to Boston

Messages to a group chat from a friend whose daughter is a high school athlete:

  • So my daughter caught a ride today from NJ to Boston with a girl whose father had a Tesla
  • took six hours to get here
  • because they had to stop three times to charge his thing
  • “we went to get sandwiches and the dad still sat in line waiting for the charger”
  • “we had to drive sideways to some mall to the charger”
  • “the GPS said 3h50m to get to Boston, it took us six”
  • “I wanted to be funny so I asked the dad if he would recommend an electric car to us because we have old cars and it’s time to upgrade. He enthusiastically said yes”

This was not a weekend with exceptional travel demand. This was not a trip through a sparsely populated state. This was not a trip on back roads.

I’m still a Tesla fan because it is the only company with Dog Mode (see Car/Kennel from this blog in 2003). The latest Tesla 3 seems to have been restyled slightly. Here’s the 2019 version from Car and Driver:

This the latest version, perhaps available in the U.S. in 2024:

I hate to give up the space and sliding doors of the minivan, though, and I’m not sure how we would charge an electric car. We don’t keep our Honda Odyssey in the garage. We would need HOA approval to install a car charger on the exterior and I am not aware of any chargers that fit into a Spanish Colonial Revival style.

Full post, including comments