tomo,

Thanks for the new link. Very interesting. Serious software failures at all levels from charge management to infotainment. It puts me in mind of Feynman's view on the testing of the shuttle rocket motor: no component-by-component tests, just assemble the whole thing, test to failure and say it's good for half that. Admittedly not a close parallel, but there's a similar air of slapdash about both setups.

All the BMW reset stuff is straight from that joke about car industry versus IT industry. Finally the car industry has caught up with try a reboot and see if it happens again. Yep: progress!

I like your idea of searching for notes of programmer disaffection in released software. Could be quite effective, but not once word got around. The toxic managers would be doing their own greps (once they found out what it was) since a key step in dysfunctional management is to hide the symptoms.

Shellenberger gets my tick of approval on that one. There's a lot of criticism of democracy these days; seems misguided, but how on earth *did* we end up with an Australian government that actively hates Australians?

On battery management, I got a sense of the complexity of it a couple of years ago writing some charge/discharge stuff. To begin with you're saying how hard can it be. After a while you find out. Batteries seem to be on the next big thing treadmill. Each chemistry has problems and, depending on which problems you focus on, you can direct people to the next magical solution. Much movement, little progess.

Somewhat interesting video on Musk's Boring Company. Too long, and too much of an advert, but watchable (or at least skimmable). Bit doubtful about scale and equipment failure, and it's not too attractive if it's Tesla only.

tomo,
That grok.com link tells me "This chat is private".

Interesting comparing the Apollo and FADEC software. Both in assembly language. Both on hardware that was relatively constrained (for the day). Google contrasts that Apollo had about 40 dedicated programmers for each major subsystem (lander, command, etc.) whereas FADEC had "a handful".

Google suggests this is at the heart of the problem. Doubtful. I'm a bit puzzled what 40 programmers could do on the tiny Apollo computers. I'd put far more blame on the high turnover in the FADEC team and the fact that the project dragged on for nearly a decade.

Quite possibly the *big* difference was motivation. I imagine the Apollo programmers sensed they were making history, and the passing parade of FADEC programmers were just there for the daily grind.

The sad thing is that things have probably gone downhill since FADEC. Vastly larger systems, far more people programming than before (inevitably getting further down the aptitude bell curve) and technology pervading nearly everything we do. Apollo -> FADEC -> Horizon -> Brazil.

Didn't know BMW went for WinCE. Blech. Probably involved some quid pro quo from Microsoft. Apparently it didn't last all that long and iDrive has been QNX/Linux/Android based for a fair while.

Svetlana piece sounds plausible enough. Thing is, the Russiagate stuff was never compelling to me. The question: why would the Russians have preferred Trump over Clinton? Without that being well established, the supposed controversy was just so much chattering amonst the talking heads.

tomo,
The Audi was indeed terrible. I was just wondering why they had the catch can (presumably because they anticipated seal failure), but had nothing in place to detect when coolant was getting into it.

A long time ago, car water pumps would fail badly after seal failure because coolant got past the seal and washed the grease out of the bearings and it wouldn't be long before it all fell to bits. Sorted out in more recent pumps by including a weep hole after the seal but before the bearings. The hope was that somebody would notice the loss of coolant and the rusty telltale below the weep hole. Meanwhile the pump continued to work fine. Kind of amazing if that lesson needs to be relearnt for EV motors.

I agree that things like oil pickups being made of plastic are crazy; the epitome of penny wise and pound foolish. There's some sort of sickness taken over in car manufacturers. Basic engine designs used to be very stable things, refined little by little with decades of experience. Nowadays it's a festival of daft new ideas (wet cam belts, concentric clutch slave cylinders, non-priming oil pumps). Any savings they make on fuel are completely lost in increased maintenance costs and/or shortened life.

Don't know what to make of the Widdecombe murderer. I'm pretty sure the internet has dramatically increased the number of delusional people getting around.

The Orlowski tweet was news on me too, though not a great surprise. Then again, a degree in creative writing is hardly a prerequisite to making a bunch of blarney about AI (or anything else you want to sell).

As for that other tweet with AI as a new foot-in-the-door for the nanny state: no thanks. An offbeat '80s song that I liked was O Superman, by Laurie Anderson. We seem to be getting closer to that comfortless future.

And from a less bleak past, I see one of the weekend's comments at Jo Nova's linked to a video on the Apollo 11 software. Interesting how they could do so much with so little.

tomo,
Thanks, I hadn't seen that Tele article by Orlowski before. Conclusion goes somewhat against the general drift of the article: it’s less an AI problem than an ethical one, but I think that's that's the right way to view it.

Computers are well suited to problems of search, and it seems this breakthrough on Navier-Stokes was joining up pieces that were already there. AI may mark the end for mathematicians like Ramanujan (who "merely" found pearls on the beach). Still plenty of scope for human invention. And as Grant Sanderson said in that interview (and as he practises in his YouTube channel) there's plenty to do in coming up with better ways of making results understandable.

As for misinformation crackdowns, they're on a hiding to nothing and it'll end up as a running joke in the mold of Pravda. The path to being believed is through an established record of truthfulness. Maybe they should give that a try.

Don't think Jean Philanthrope will be on my frequently viewed list. The things that go viral: very odd.

On Brazil: I've long thought we were already living in Brazil world in the details — all the little tendrils of bureaucracy reaching into everything we do — but as you say, the darker, big-picture parts of Brazil may be in play too.

Speaking of bureaucracy, had a refreshing change yesterday. My US bank has changed ownership and I'd lost access to online banking. Making it difficult, I had a friend's address registered with the old bank, so all the authentication stuff was being sent to her. Tried to get it sorted with a phone call to their support number. Two *amazing* things (to this inhabitant of Brazil world):

  o The phone was answered immediately by a human.
  o He bypassed a security check (authentication code SMS) so that he could update my phone and address details.

Dealing with a human who gives every indication he thinks you might be a human. Such a contrast with what we put up with here!

Very interesting on the EV coolant leaks. So "last indefinitely" becomes more like "last as long as an old-school water pump". Crazy that they let this completely predictable leak get into the electrics or the mechanicals. Ok, they'd have to swallow their pride and admit a (say) 100,000 mile service item to replace the seal. A lot cheaper than replacing the motor or the car.

Audi "catch can" seems a good first step. Lets you see that there's a problem. Seem to have failed on step 2: when coolant detected in the can, replace seal on motor.