« Be Joyful Though You Have Considered All the Facts | Main | And third on the list »

October 17, 2023


My first thought: Ammonia exhaust is surely incredibly toxic and dangerous, isn't it?

Not a great idea. In fact, a bad one.

So I went looking. Not in depth at all, but here is the first site I found that wasn't pout out by ammonia energy advocates:https://www.sciencedirect.com/science/article/pii/S1352231014005809

Here is the concluding paragraph:

"NH3 is classified under the dangerous substances directive (67/548/EEC) as: toxic, corrosive and dangerous for the environment. It is responsible for the formation of atmospheric secondary inorganic aerosols that are known for its adverse health and environmental effects. Although, NH3 is usually associated with the agriculture and the rural environment, it is also observed in urban areas and roadside locations due to vehicular exhaust emissions. Several factors contribute to the formation of NH3 in vehicle exhaust.

"NH3 emissions vary considerable and depend on the vehicle and its emission control technology. While the vehicular emissions of NOx and CO have substantially decreased over the past years, the NH3 emissions for Euro 5–6 vehicles reported here (4–62 mg/km) are comparable with those reported during the last decade (Durbin et al., 2002, Huai et al., 2003, Huai et al., 2005, Heeb et al., 2006, Heeb et al., 2008, Livingston et al., 2009).

"Moreover, the vehicle that represented the newest technology and that complied with Euro 6 standards showed the highest NH3 emission factors at the two studied temperatures (62 and 70 mg/km at 22 and −7 °C, respectively). The presented results show that all tested vehicles, with no exception, emit ammonia. When and how much NH3 is going to be emitted will depend on the vehicle engine and after-treatment technology, the ambient temperature, the driving style and the car manufacturer strategy for NOx emission control."

So... clever PR and carefully chosen language, but Toyota is on a very wrong pathway here.

the people who pop up and talk about how going all electric is a woke thing and good on Toyota for standing up to the EV mafia.

The fact is, the "silent majority" is voting with their wallets. (The market version of "voting with their feet.") Charging infrastructure is still very much a work in progress. Rather rapid progress in some places, but a long way to go yet. Which makes the EV uptake we are seeing all the more impressive.

Hybrid plus: range, less dependent on charging stations.

Hybrid minus: you get BOTH the complications of electric and ICE.

I agree that, right now, hybrid makes more sense. But maybe not in five years.

Toyota doesn't want to be stuck providing the next BetaMax, I'd guess.

Rudolf Diesel started with an ammonia engine before developing the engine named after himself.
I believe, a major problem with the former was that it always leaked. Bad enough with fuel oil but fatal with ammonia. The stuff also being corrosive does not help

Diesel experimented with an expansion engine using ammonia rather than steam. That's different from using ammonia as a fuel.

It's untrue to say Toyota is 'standing up to the EV mafia'.

In reality they've recognised that they misjudged the market, and are making strenuous efforts to catch up.

Sorry, I should have been more specific.
Early fridges also used it (and iirc industrial freezers even quite some time after home fridges switched to the - then seen as unproblematic - fluorinated/chlorinated hydrocarbons).
Using ammonia as a fuel is arguably even worse.
And why use huge amounts of energy to make ammonia from hydrogen when the hydrogen itself is a suitable fuel? Liquid ammonia does not need to be as cold as hydrogen but still the stuff is so nasty, that I can't see any advantages. For that matter, I'd feel not very comfortable with methanol as car fuel either (for fuel cells. A former colleague of mine worked on catalysts for that). Has anyone proposed liquified prussic acid as fuel yet [except some antisemitic a-holes who joke about this, I mean] or carbon disulphide?

I just read that there are options for storing ammonia solid (as chemical compounds that decompose when moderately heated). Still I am sceptic about the production. Replacing Haber-Bosch with some catalytic room temperature process has been announced as just around the corner for many decades now. Something of a chemical Friedman unit.

I'm not sure why "Science at the end of the World" isn't all about stuff from South Pole Station.

Because we live in the endtimes, of course.
And, if you pedantically insist that end is spatial and it would have to be ending for a temporal meaning, I shall insult you as a grammar [Godwin].
And we live in very pole-Arysing* cirumstances/tempral relationships.

*Pol-arisierenden Zeitverhältnissen (the lame double pun does not fully translate from German)

Please insert an 'o' behind the 4th letter of the 8th word in the fourth line.

Please insert an 'o' behind the 4th letter of the 8th word in the fourth line.

Always keep in mind that in HTML, layout-affecting decisions like font family and size, block width, and other stuff is a local decision :^) More so in my browser than most people's, since I go out of my way -- including a custom plugin -- to enforce my choices on those matters. For example, the spelling error is on the fifth line of my browser's rendering of the comment.

Hyy, thw whole pedantic joke would be lost, if I had just written: I forgot the 'o' in 'temporal'.
I actually thought about the layout problem and then deliberately ignored it for that reason.
And I was too lazy to count the letters from the start.

On a different science (OK, engineering) topic. I'm listening to an online webinar at the moment. Subject: 3rd Generation AI.

A couple of fun factoids:

  • A claim that the new generation of AI will be able to take a mix of 4 sets of DNA, and separate out a single individual's DNA.
  • A note on AI-driven drones: a .0000001% failure rate on detecting obstacles equals a 100% crash rate within a day.
Just in case anyone didn't realize how far AI is from being ready for prime time.

I may already be sensitive to this because my boss has become enamored of ChatGPT for creating documents. (Not that she isn't perfectly capable of writing them herself. It's just she feels it's easier.) The number of iterations of editing required to get to something viable for outside-the-company use is pretty high. Not to mention the recurring need to check that the software hasn't just made stuff up.

A note on AI-driven drones: a .0000001% failure rate on detecting obstacles equals a 100% crash rate within a day.

wj, can you explain what this means? Does it mean there are crashes on 100% of days, or am I being particularly dim?

As I understand it it's that even with a failure rate as seemingly low a crash within a day of use is almost certain. That would mean that such a drone has to make millions if not billions of correct observations concerning obstacles each day to be successful and safe.

It's like with the first Mars landers: 3 million essential parts iirc. So, even if each part has a probability of failure of just 1 in a million, that would mean almost certain (95%) a fatal malfunction.
Throw an x-sided dice x times (with x very high) the probability that a pre-chosen number comes up not a single time is 1/e (about 36%)

Thanks, Hartmut.

Thanks, Hartmut. That's what I understood as well.

Not my kind of software, but I'm an old-time systems analyst and I know how I'd approach the problem. Missing an observation is not sufficient -- or certainly should not be sufficient -- to cause a collision. (Sanity checks: people blink while they drive and it doesn't cause accidents; most people can close one eye and drive for quite a while without any problems.)

The software is going to have to interpret millions of observations and partial observations per day. To cause a collision, it will have to make a bunch of wrong interpretations over a significant length of time. The probabilities for that are quite different than the probability of a single failed observation.

Hartmut's reasoning also covers one of my favorite bar bets. "Look, I'll bet you that if we go around the bar and ask 30 people their birthday, month and day, at least two of them will have the same birthday. Loser buys the next pitcher, okay?"

But the misinterpretations are not necessarily independent either. One could lead to the next and - once they accumulate - correct ones could be sorted out because they contradict the previous "consensus". I do not know whether modern AI can suffer from anankastic syndrome too. If so, that could lead to self-overloading through demanding more and more data from the sensors in case of critical situations until the system has no more capacity to interpret the data because it is completely occupied with just processing the sensory input.
(I know this only as a hypothetical scenario from older literature*, so it may not be of relevance anymore. Btw, Stanislaw Lem used it in the final Pilot Pirx story "Ananke" (1971). There the idea of a self-learning system with an external trainer is presented and it looks quite similar to how modern AI systems get trained.)

*something like that happened in reality during the Apollo 11 mission when the computer in the LEM (Ha!) got too much sensory input and lacked the capacity to deal with it needing a manual override.

I want to say 23 is the number of people where birthday sharing exceeds a 50/50 chance. If it's not 23, it's 27, but I'm leaning heavily toward 23.

It's 23. The nice part is that leap years don't change the 50/50 threshold. Whether you calculate it with 365 or 366 days, it's 23. I'm not sure how to accurately account for an extra day every 4 years (use 365.25?), but the actual probability would be somewhere in between simply using 365 and 366.

I too remember the 23 but I also remember that it was rather tedious to calculate.

It's not difficult, if one assumes that there are n days at the year and that everyone present is likely to be born on any given day with probability 1/n.

With 2 people, the likelihood that they're born on different days, call it p_2, is (n-1)/n

With 3 people, the likelihood that they're all born on different days, p_3, is p_2(n-2)/n

And so on. For n = 365 one finds p_23 = 0.4927

I too remember the 23 but I also remember that it was rather tedious to calculate.

I keep a toy* APL interpreter that runs in a tty window on my computer to use as a calculator. 17 characters to get a list of the 29 probability values corresponding to "ask 2" up to "ask 30".

* Ken Thompson wrote the first version at Bell Labs around 1974. The source code went to Berkeley, then to Yale, then to Purdue where they added a lot. I got it back from Purdue while I was at Bell Labs around 1981 and tidied it up. Took it to Bellcore with me under the terms of the 1984 AT&T breakup, then to USWest in 1988. Spent six months in 1992 or so brow-beating the lawyers at Bell Labs and Bellcore to get permission to open-source it. I believe that if you google "michael cain apl11" you can still find copies of it.

I once told five coworkers that there was a good chance that two of us had the same birthday. They poo-pooed the assertion. Turned out that one of them and I did have the same birthday. Also turns out that there's only about a 4% chance of that happening.

Wow, APL is my go-to example of "obscure computer language".

As in, "implement an HP calculator on a supercomputer, where a single keystroke can invert a 100x100 complex matrix".

Someone should ask ChatGPT to write a quicksort program in APL.

arr ← 2 4 1 3 5 1 0
sorted_arr ← quicksort arr

quicksort ← {
⍵ {
⍵≤1: ⍵
pivot ← ⍵⌷⍨?⍨⍵
left ← ⍵/⍨⍵ mid ← ⍵/⍨⍵=pivot
right ← ⍵/⍨pivot<⍵
} ⍺

—OpenAI ChatGPT 3.5

I tried several other LLMs and got all different and strange results. The above looks most like what I remember of APL.

I believe that if you google "michael cain apl11" you can still find copies of it.


I would assume (maybe incorrectly) that a big problem for Toyota would be scaling up EV sales to the level a company their size needs to make. There are plug-in EV Priuses (Preii?) so they aren't ignoring the market. The issue of the availability of charging stations is a big problem for long-term adoption of EVs, with range and charge time secondary issues. Those have to be addressed to eliminate ICE.

Priuses (Preii?)

If one wanted to use a Latin plural, it would be 'Priora'. However, Toyota conducted an absurd poll, and declared 'Prii' to be the winner.

Unless there's a true technology breakthrough so that the battery is not the most expensive/important component in the EV, the former car companies are in the position of either becoming a battery company (eg, Tesla) or becoming critically dependent on an outside supplier. I think Toyota has been stalling over making that decision.

They seem to be considerably behind, now betting their future on solid electrolyte technology they own, but that won't be available at commercial scale until 2027 or so.

Prius is the neutral form of the male/female prior, so Pro Bono is correct with priora at least as Latin is concerned.
But Italian turned tempus into tempo and the plural from tempora to tempi, so Prii would at least make some sense. Still it sounds horrible and Priuses would indeed be the better modern option.
Not to be confused with Priapus, although Musk is a bid D*** arguably (an ox would be preferable).
While we are at lame punning: Prius has no positive.

Here's an Oxford-style debate on the resolution, "Between now and 2035, electric vehicles in the consumer market will disappoint environmentalists by remaining a product bought mainly by the well-heeled minority."

The debaters don't present a complete picture of EV adoption. And what they do present may be skewed a bit by their biases. But they cover a large part of the arguments for and against EVs.

"The Manhattan Institute's Mark Mills and InOrbis CEO Rosario Fortugno debate the resolution, "Between now and 2035, electric vehicles in the consumer market will disappoint environmentalists by remaining a product bought mainly by the well-heeled minority."

Taking the affirmative is Mills, a Manhattan Institute senior fellow, a faculty fellow at Northwestern University's engineering school, and a partner in Montrose Lane, an energy-tech venture fund. He is author of the book The Cloud Revolution: How the Convergence of New Technologies Will Unleash the Next Economic Boom and a Roaring 2020s.

Taking the negative is Fortugno, the CEO of InOrbis, a company that works to develop technologies for electric vehicle fleet management, autonomous vehicles, and machine learning. He blogs at ApplyingAI.com on the topics of free markets, electric vehicle adoption, and the benefits of artificial intelligence."
Will electric cars disappoint environmentalists? A Soho Forum debate (YouTube)

If you don't want to watch/listen to an hour-and-a-half YouTube video, here are most of the arguments they made:

Here are the key arguments made in favor of the resolution:

• Raw Material Constraints: Mills argued there are insufficient supplies of battery minerals like lithium, cobalt, and nickel to support mass EV adoption. Expanding mining capacity takes over a decade. This will constrain affordability and access.

• High Battery Costs: Batteries are the most expensive part of EVs. Their cost is dominated by raw materials, which Mills says will face increasing price pressures. This will keep EVs unaffordable for average consumers.

• Infrastructure Bottlenecks: Mills cited the need for $500 billion+ in charging infrastructure upgrades to support widespread EV use. Home charging also requires grid upgrades. Lack of convenient charging access will limit mass adoption.

• Marginal Emissions Benefits: Mills disputed studies showing EVs are cleaner, citing uncertainties and flaws in emissions analyses. Upstream mining emissions may offset gains from not burning gasoline. This undermines a major motivation for EVs.

• Slow Urban Transition: With cities already lacking parking/charging, and apartments without home charging access, urban areas will be slow to adopt EVs. This limits their market potential.

• Luxury Niche for Now: Given high costs and obstacles to mass infrastructure, Mills sees EVs remaining a luxury niche product affordable mainly to the wealthy until at least 2035. This will disappoint environmentalists hoping for broader access.

In summary, Mills argued material constraints, battery costs, infrastructure needs, questionable emissions benefits, and urban charging challenges will limit EVs to an expensive luxury niche disappointing environmentalists.

Here are the key arguments made against the resolution:

• Falling Battery Costs: Fortugno cited forecasts of 50%+ reductions in EV battery costs by 2025-2027, driven by manufacturing improvements. This will make EVs cost-competitive with gas cars.

• Expanding Charging Network: With 50,000 Superchargers worldwide, Tesla is rapidly expanding fast charging access. This removes infrastructure limitations.

• Convenience Factors: At-home overnight charging eliminates gas station trips for most driving. Autonomous driving gives free time. These are compelling lifestyle benefits.

• Performance Advantages: EVs offer faster acceleration and high-tech features like advanced driver assist, which is not available in gas cars. This performance edge appeals to consumers.

• Lower Lifetime Emissions: Fortugno disputed Mills' claims, citing multiple studies showing EVs have lower emissions over their lifespan, especially as grids add renewables. This helps environmental goals.

• Market-Driven Adoption: Fortugno argued consumer preferences, not just policy or subsidies, are driving EV adoption based on performance, features, and falling costs. This trajectory will continue.

• Tesla Leading Transition: With top safety ratings, low maintenance costs, and market-leading technology like self-driving, Tesla is poised to dominate the transition to EVs across income levels.

In summary, Fortugno argued falling costs, expanding infrastructure, compelling features and performance benefits, plus environmental advantages will drive rapid mainstream EV adoption, disappointing pessimists expecting a niche luxury market by 2035.

—Anthropic Claude-2-100k

Not to be confused with Priapus, although Musk is a bid D*** arguably (an ox would be preferable).

The one case I can think of when I sympathize deeply with the iconoclasts.

Autonomous driving gives free time.

I think that is going to be a particularly hard sell. I know the enthusiasts love it. But the average consumer is going to be wary. And anybody in the IT field is going to be more wary still. Nobody wants to be a beta test site when it is their life and limb on the line.

At least in Germany car owners want to drive themselves. I guess that's also the reason why over here manual transmission is the preferred option and automatic a bit of a niche. Even people that could employ a driver often prefer to sit behind the wheel themselves. So, even extremly reliable autonomous cars would have some hurdles to cross to gain acceptance.
German obsession with freedom of the road has some similarity with 2nd amendment fetishism in the US. Speed limits are very unpopular for example and right wing parties had abolishing them as part of their party platforms, in particular 30km/h zones were targeted. If one knows the typical side streets of German towns and cities, one can see that this is extremism for extremism's sake in the name of unalienable rights, no matter how many kids get run over.

But the most reckless drivers around here by far are people connected to Arab embassies. They collect tickets by the thousands each year but diplomatic cover saves them from having to bear the consequences. In case of fatalities these guys get simply recalled.

But the most reckless drivers around here by far are people connected to Arab embassies. They collect tickets by the thousands each year but diplomatic cover saves them from having to bear the consequences.

I can't speak to Arab countries in general. But my experience in Saudi Arabia was that traffic rules were effectively nonexistent. For example lane markings were considered entirely decorative. Not even suggestion, but absolutely irrelevant to operators of vehicles.

Which makes me think that they are just driving abroad the way they drive at home. And, as diplomats, suffer exactly the same (total lack of) consequences for ignoring petty legal requirements that they would at home.

my experience in Saudi Arabia was that traffic rules were effectively nonexistent. For example lane markings were considered entirely decorative. Not even suggestion, but absolutely irrelevant to operators of vehicles.

It's the same in Tehran. Also, sometimes people just start driving in reverse gear if they miss an exit. And the smell is awful, because petrol is not as refined as over here (I think it's due to sanctions on technology or something).

It's the same in Tehran

The reaction of an Iranian to being considered the same as Arabs, on this or pretty much anything, can only be imagined. "Going ballistic" is one term that comes to mind.

Oh, I'm fully aware of that and certainly didn't mean to imply otherwise, lol.

Embassy personnel seems to have a general tendency towards that, the guys from the Gulf region are just the most extreme there.
Immunity is obviously character spoiler.

The comments to this entry are closed.