Posts by Steve Hersey 136 publicly visible posts • joined 11 Aug 2009
ISTR that several of the recent failed landings involved trouble with laser rangefinders. I wonder if they're underestimating the amount of dust kicked up in low gravity by the descent rocket's exhaust plume, or its optical properties? That could certainly interfere with a time-of-flight range measurement.
Seems too obvious to overlook, but who knows.
Decades ago, I worked on a truly awful project building an app to talk to a portable instrument.
The entire software spec was, "We want it to talk to the flowmeter." Predictably, it just got worse from there.
There was a series of internal-only test versions; predictably, sales, who were expressly forbidden to give them to customers, did exactly that, and demanded that I support these incomplete test apps.
I responded by incorporating a special splash screen in the next test version that read, more or less, "INTERNAL-ONLY <EXPLETIVE> TEST VERSION, NOT FOR CUSTOMER USE."
To the utterly predictable howls of outrage from sales conveyed to my manager that this was inappropriate language to expose to customers, I responded by by pointing out that internal-only test versions were never supposed to be given out in the first place.
Alas, as Corey Doctorow is in the habit of explaining, the corporate enshittification process inexorably pushes the company into ever more rapacious, exploitive behavior to feed shareholders' insatiable demand for more profits every quarter. And the regulatory guard rails that would stop that slide have incrementally been eliminated. The DMCA made it all much worse, as it empowers all sorts of anticompetitive crap; it's now a felony to bypass the toner-cartridge lock-in features, as that's a "protected work."
I, too, am happy to repay a manufacturer with a decent profit for a decent product, but not to be endlessly squeezed for more by increasingly abusive practices.
The point where the aircraft designers BUILT IN A WHOLE SYSTEM to detect that the gears were eating themselves up was the logical place to stop the whole program and say, "Bad idea, let's go back to something more reliable." The fact that this didn't happen says a lot about how organizations can lock themselves into pursuing a bad idea rather than just scrapping it.
Full agreement with the poster pointing out that a flock of A-10s is more cost-effective. From what I've read, the A-10 isn't glamorous enough for the Air Force, but the folks who really hate it are the ones in its gunsights.
I'm sure there is exactly that serious work underway at MS to enshittify W10.
It wouldn't surprise me in the least if one of MS' last updates to W10 contains a logic bomb that randomly crashes the machine after official support ends if the extended support hasn't been purchased. I don't think even Redmond would be QUITE dumb enough to just brick W10 machines, but they're obviously not far from that level of dumb.
It never ceases to amaze me that folks will install the proverbial Big Red Mushroom Button on a wall or a rack and NOT put a transparent safety cover over it.
Seriously, folks? If you really, really NEED to mash that button, you'll have that cover open before anyone can say "NOOOO!", and if you DON'T need the button being mashed, the cover will prevent lots of needless drama and expense. There is *no* downside to the safety cover, especially if you cover it with painter's tape when the wall is being repainted so it remains transparent afterwards. Apologies to those painting contractors who really aren't that dumb; I know you're out there somewhere, busy as hell.
These folks need to be punished.
And we need to vastly expand the ability of agencies to impose meaningful penalties on corporate violators, especially large ones. F'rinstance, statutes could set the maximum fine as a percentage of the offender's assets, with a floor for the maximum fine to deter gamesmanship.
One: Generative AI is irredeemably crap. There's no way LLM tools can possibly replicate human judgment, filter for truth in any reliable way, or stop parroting obvious BS because it's on the Internet. To an LLM, Donald Trump's statements are equally valid input to Kamala Harris', and Fox News stories are as valid a source as NPR. Expecting sense from these tools is a fool's errand.
Two: There are lots of people intent on making money off these things, and determined to convince us all that they can do what they clearly cannot. There are also people who want to (mis)use these tools to get rid of those pesky, expensive human employees and make their quarterly financials look better. AI chatbots instead of human tech support, f'rinstance.
Problem one is a technical question; problem two is a social and ethical one.
About 25 years ago, my US employer provided weather satellite instruments to the European Space Agency, and I did on site ground support, which included setting up the instrument test console; a Sun Ultra One if recollection serves me. The test control room was wired for 220V 50 Hz, and after procuring the relevant power cords with Schukosteckers, I connected the Sun monitor to the mains - having first verified that it was automatically dual-voltage agile, and didn't require a switch.
Alas, I then neglected to verify that the CPU was ALSO auto-adaptive. (The nameplate said 105-230V AC, 50/60 Hz - we checked that before shipping it - so we were good, right?). Plugged it in, flipped the switch, and POP! Oops. Dead, smelly PSU. The CPU DID have a selector switch. which was, of course, still set to 110V.
Fortunately, I had good relations with the local techs, and even more fortunately, the PSU was a standard PC type. We made a quick trip to the local Mega store, where I bought a replacement and swapped it in. Mission rescued. (I think we first temporarily pillaged an idle Sun workstation for its PSU so I could get the console up and running, verifying that the rest of the station hadn't died as well.)
Never made THAT mistake again.
Ironically, some of the other instrument contractors were using 110V-only computing gear, and had a separate AC supply through a stepdown transformer. Which I didn't use, since the Sun station was dual-voltage capable.
I find it curious that the discussion of renewable energy generation doesn't seem to include one blazingly obvious aspect: The world, and in particular the rich, highly industrialized countries, are going to HAVE to become much more economical in our use of energy in general, and electricity in particular. We simply cannot endlessly consume ever-increasing amounts of power and expect to survive. Nor can we demand that developing nations economize where we do not; that's a recipe for failure, massive unrest, and mass migration.
The US in particular (and probably China, but I'm less sure of that) could do much better in energy efficiency. When I've been in Europe, folks there seem to be much more conscientious about energy use in general, as illustrated by simple things like not lighting unused spaces. I've seen more motion-sensor lights there than I ever saw in the US.
Human civilization will not survive climate change without getting serious about energy use, and this is going to mean making do with less, especially for rich and powerful nations. That is an uncomfortable truth as well as a very hard thing to sell politically, but it IS a truth. Fortunately, there are signs that people are becoming aware of that fact and are willing to make necessary sacrifices, IF they see that process as equitable and fair.
Tesla will have a hard time competing with BYD in the Chinese domestic market. BYD has better build quality, newer designs, and - critically - they are not a foreign company. That last factor will deliver both political and customer-preference advantages to BYD. The Chinese government is notorious for tipping the scales in favor of domestic producers over Western multinationals, and I doubt this will be an exception. (And one cannot really blame them for that.)
Tesla may be "successful" in China in a break-even sort of way, but it sure won't be a growth engine or a money-spinner.
We see a tracked vehicle tooling around in open country, over uneven ground that the vehicle chassis can obviously just bull through without navigation assistance. We do not see it maneuvering among anything I would call an obstacle, even to a vehicle that couldn't just drive right over medium-sized shrubbery.
There may well be high-performing autonomous systems here, but the video shown doesn't substantiate that proposition.
"Look! A monkey!"
Nothing, not even free rides, could persuade me to trust myself to one of this man's robocabs. That's even assuming they ever happen, which I think is highly unlikely.
NPR reports that he's planning to achieve low manufacturing cost by using a "revolutionary" manufacturing process where almost the whole body is cast in one shot. Surely I cannot be the only person thinking "This isn't going to be as quick or as cheap as he seems to think. Process revolutions never do."
I do the tech for the home business.
We got a Canon inkjet with the big tanks on my advice because it was well reviewed and the ink cost per page was low, and have had nothing but trouble. There's a leaky seal somewhere, and multiple ink hoses get air-bound every few weeks. Colors stop working. Deep clean cycle wastes a lot of ink. Partner gets real annoyed. Cannot seem to fix it, and no replacement parts for the affected items.
While this is probably a one-off quality issue no one else has been bitten by, and any identical replacement would likely perform perfectly, we're sufficiently cheesed off with it that ink-tank printers are no longer an option, and the next time this one borks it's going on the curb. One bad experience is enough.
Obvious, but also obviously NOT ENOUGH. You can reduce the incidence of your staff being phished, but you *cannot* totally eliminate it. Therefore, you cannot responsibly base your cybersecurity strategy on that assumption.
For those few of us readers who may not already understand that point, defense in depth and robust intrusion detection and response are essential.
Also essential, if perhaps a tad less obvious, is NOT shaming the phished victims into hiding the fact. That just helps the bad guys. Establish an infosec culture of *immediately* contacting Security, and of Security responding immediately in a supportive manner. No blaming, no shaming. THAT way you get the fastest possible notification that you've been attacked, and you stand the best chance of minimizing the damage.
Sure, you may get some false positives this way, but that is far outweighed by the benefit of quick and effective detection.
OK, so you MAY need to shame some C suite idiots into being more careful and more forthcoming, but that's a tool to be used sparingly and with great care.
The trouble with reference designs for crazily complex CPUs is that it's rather difficult for anyone but the chip maker to effectively optimize the bypassing scheme, as only the maker really knows where the critical points and needs are.
I'd be surprised if the motherboard makers' engineers get much leeway for comparative design testing when churning out the board for the Latest Sexy Chip. There must be incredible pressure to just use the reference design.
Pilotless electric-powered VTOL flight is HARD in a number of ways. Battery life, power-to-weight ratio, achievable range, VTOL takeoff/landing. Add in the regulatory issues of operating drones in airport-adjacent airspace and over populated areas, and the additional issues with getting certification to carry passengers, and you're looking at development, certification, and operating costs that utterly rule out taxi-scale costs. (Not to mention the steep challenge of getting actual humans to trust themselves to a tiny pilotless airplane.)
This one won't fly. Fuggedabouddit.
Upvote the above.
SpaceX doesn't just ignore Musk's loonier spur-of-the-moment notions; they have a dedicated team of Musk minders specifically tasked with steering him away from anything he might break, and mitigating his worst excesses. Let him push the shiny buttons, preside at major launch events, and name the drone ships, but keep him away from the actual engineering.
The dead bird site lacks any such mechanism, which goes some distance to explain how he managed to do so much damage to it so fast.
Self-crashing cars marketed as self-driving, terrible product quality built under dangerous, awful working conditions, hiring whole teams to dissuade the owners from using their warranty, selling domestic-use flamethrowers, touting mass-transit vaporware, the list goes on.
It's hardly surprising that he'd sell a flashy, expensive, basically useless truck that rusts in the rain.
I had a disagreement with a newspaper tech reporter who insists that Musk is an amazingly successful, innovative entrepreneur. You need a pretty narrow field of view to see him that way.
Sure, if you train a cloistered LLM on data certified to be pure, organic, pesticide-free, and free-range, you'll get something that only reflects YOUR unacknowledged biases, and that you can convince yourself is sorta almost error-free. But I bet it won't be very useful. Enlarge the training data to include stuff from the 'Net because you just need more training data, and you've introduced stuff that will skew the LLM. You CANNOT effectively vet a large enough training data set to make it really useful without an uneconomically large amount of tedious, soul-deadening human effort (which we all know those eager profit-hungry corporate types won't support), so the data will be poorly vetted (if at all!) and crap will get in. Game over, insert quarter. Your LLM is screwed.
Sounds a bit like GE's infamous 60-30-10 system (I may have the exact numbers wrong, blame it on traumatic amnesia), where each department was required to rank its staff into those three bins and terminate 10% of its staff as below expectation.
The problem my group encountered: Our 30-person space instruments group had been acquired by GE as a byproduct of buying an NDT (nondestructive testing) business that they ultimately weren't allowed to keep (another warring arm of GE had already bought another NDT operation, so antitrust got in the way). We HAD no underperformers, and there was no one we could spare to throw to the wolves of mindless corporate policy. Somehow our leader managed to fight off the corporate overlords, but it was a close thing.
Another notable craziness was that every department within GE was required to commit to double-digit revenue growth year-on-year, which the sane among us will know isn't possible in the real world. (How do you contrive to sell 10% more satellite gear to the government every year?)
One more GE insanity among a crap-ton. The business schools must be teaching this kind of evil crap, it's so common among corporations.
Perhaps the most useful comparison to human drivers is this one: If you have human driver who very rarely gets into a collision over many, many km driven, but still obstructs emergency vehicles, drives through crime scenes, blocks roads randomly, and tries to drive with a trapped pedestrian under the car, would you revoke their license? So would I...
I'll grant you that this may not objectively be the best measure of driver safety, but it's definitely the one we would apply to a human. And humans do reasonably well in those corner cases that AI cars just cannot manage.
I haven't seen confirmation of this, but one news source I've seen says that Cruise initially shared crash video with the DMV that showed the Cruise car striking the pedestrian (apparently unavoidably) and stopping, but withheld additional video that showed the vehicle then starting to move again and dragging the trapped victim under the car.
It's always the coverup that gets them...
Does anyone really believe that Windows 10 is materially deficient for ongoing use? <crickets> You there in the back, put your hand back down, we know you work in Marketing for Microsoft.
No one NEEDS Windows 11 except Microsoft's share price (and Dell's). Dropping Windows 10 support is purely a method of squeezing the marks for dollars.
With Windows 10's forced-update methodology, any "obsolescence" argument is moot. The OS can be kept current with little to no user intervention (or, alas, choice).
Now, if Microsoft needs an ongoing revenue source from this OS to fund ongoing improvement, an annual support subscription would certainly be possible. I'd probably even buy one, and I know that most corporate users would. But scrapping the OS and dropping support for existing hardware that still works perfectly well is needless, wasteful churn.
Perhaps we should build dedicated pathways for these vehicles; hey, here's an idea! Lay down steel tracks and move the freight on special vehicles built to run on them. Call'em "trains," perhaps.
Going out on a limb here, one could even use these "trains" for carrying passengers.
Silly idea, no one anywhere in the world has ever made that work. Nevermind.
Years ago, when I worked on weather satellite instruments, I was installing a ground test system, made in the US, in the satellite contractor's test control room in Germany. Our system, which was based on a Sun Ultra 5 (or possibly an Ultra 1; it's been a while), was intended to read and interpret instrument data. I carefully checked the display monitor - yup, it was 110/240 V, with auto-switching. Plugged it into the native 220V outlet and cabled it to its system unit - yup, also marked 110/240V. Plugged that into the 220V outlet, too. Pressed the power button: CRACK. Wisp of smoke. Exclamations of dismay. Turns out the system unit, unlike its monitor, was NOT auto-switching!
Fortunately, that particular Ultra chassis used a standard PC-compatible power supply. A quick trip to the local Mega Store later, and we swapped in the replacement supply. Even MORE fortunately, the damage was limited to the power supply itself - which we then buried deep in a cabinet and never spoke of again. Well, I didn't, though there was some light-hearted fun at my expense, which I do suppose I earned. I blame jet lag.
The most surprising thing about this story is that Amazon appeared to have ANY interest in the well-being of their drivers. There are many horror stories about their drivers having to pee in bottles and suchlike, and I don't think most of those were fabricated.
As for the rest, the lesson is clear: Cloud-connected devices, like cloud storage and servers, are not under your control. They are controlled by avaricious, amoral corporations who fundamentally have no fucks to give about their obligations or your needs, interests, or the survival of your company. Trust them at your peril.
I encountered the gotcha once when I set up a piece of test equipment based on a Sun Ultra 5, sent from my US office, in the control center of the satellite integrator in Germany. Checked the monitor, everything fine, it was 110/220V autoswitching. Plugged it into the 220V mains. Plugged the CPU in, switched it on, and BANG! Turns out that while the monitor was autoswitching, the CPU was NOT. Fortunately, I was able to get a compatible replacement supply (I think it came from the local Mega Store) and bring it back to life.
Any experienced engineer knows that the junk box is, to misquote Dune, "a wellspring of cunning and resourcefulness." Need a weird connector to build an adapter? A custom cable to defeat the hardware interlocks? A quick and dirty test fixture? Parts to fix that device that brought the production line down? A well-stocked junk box makes all these things easier, and some of them possible in the first place. And it's a frequent source of ideas, as well.
I regularly replenish mine by cruising the manufacturing floor and dumpster-diving their rejected material (often helpfully red-tagged with what bits to avoid 'cause they're the broken parts). Not to mention the repairable equipment they couldn't repair or couldn't be bothered to. (Seriously, how hard can it BE to replace the reverse-protection diodes on that nifty programmable bench power supply? Or, just spitballing here, to find a way to make it harder to plug the thing in backwards in the first place so you don't fry it AND its replacement?)
At a prior employer, a new engineering manager insisted that we clean up our cubicles and get rid of all that junk. We nodded, agreed, and stowed it out of sight until he stopped being our engineering manager, which didn't take all that long. He wasn't a bad person, he simply didn't understand how engineers work or why that "clutter" is productive. An engineer's work space is NOT a good place to go all Marie Kondo, trust me.
I've been a bystander for the situation. Sure, telling the ex-boss to shove it gives you brief satisfaction, but having them pay you stupidly large sums of money - on your terms - is a much more satisfying experience. Yes, the pain will be felt by the shareholders. Not your problem, the gain is yours.
If your focus is on really wanting them to suffer, find some way to get past that, because that hurts you more than it hurts them. (And if "some way" means taking a baseball bat to their Beemer - which is really NOT a good idea - then for heaven's sake make sure it doesn't get caught on security video. And make sure your consulting fee greatly exceeds the fine, just in case.)
Oh, and if they try to cheap their way out of their self-made mess by offering you part payment in used lab equipment, and you happen to choose equipment that's far more valuable then the value they agree to assign to it, make sure the transactions are all complete and the gear safely carted off before you mention that fact. I'm told their facial expressions are quite a treat.
"Tesla owners would be able to add their cars to a shared Tesla fleet that would allow them to be used as self-driving taxis when not in use."
So I can have persons unknown put ciggy burns and unmentionable stains on my expensive car seats while exposing my vehicle to additional road hazards and buggy FSD software? What a concept. And if the FSD decides to prang it, whose auto insurance rates will go up? Oh, wait, my insurance won't cover driving-for-hire at all, and Tesla won't pay either.
"What else would ChatGPT do to protect itself from being discovered as a liar?" NOTHING. It isn't intelligent. It cannot care or not-care.
Wow. These AI models are meta-creepy. Even the computer-savvy writer got sucked into thinking that there was something intelligent in that overrated database engine. ChatGPT would do nothing to prevent its discovery as a liar, because it has no intelligence, no concept of discovery or truth; in fact, no concepts AT ALL. There's nothing in there to have views or opinions.
All of which just goes to strengthen his argument that we should kill it with fire and salt the earth. Pension off the developers on condition that that they never do anything remotely like this ever again.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
