Posts by Justthefacts

What do Uber drivers make of Waymo? 'We are cooked'

You’d think so, wouldn’t you? But you aren’t even close. A train driver earns *a lot* more per passenger than a bus driver.

A bus driver earns £14per hour. It’s such a casual job, that nobody even quotes an annual salary.

A train driver in 2023 earned £60k-87k *before* the recent 15% pay-increase (back-dated). So that’s £100k *base* salary. But the industry runs on overtime. Almost everybody works an additional one or two six-hour shifts per week, @£600 per shift. Yes, the average train driver in the U.K. does indeed take home £150k, and those who do a lot of overtime can pull £200k+. I find that middle-class software engineers with degrees tend to get quite upset when they discover this.

Now do you understand why storing the electricity in a battery vs bringing it in on a stupid metal rod, is just *irrelevant*? Social factors win against naive back-of-the-envelope theorising every time.

Sorry, this is just another bunch of special pleading to support the conclusion you want. A tram is a bus with an overhead transmission line, and metal wheels.

Now if you *want* to overdesign and overbuild it, by your own admission multiplying the weight by more than double, so that you can beat the hell out of it with a jackhammer and it still keep on going, you can. Whether it’s a bus or a tram. It will last perhaps three times as long, and cost twenty times as much. That’s why nobody builds them like that any more, unless they are spending somebody else’s budget.

I’m an old git. I’ve driven an old-style decommissioned and re-purposed school bus (USA) from the 1950s, way back in the 1980s for a few years, hacking back and forth from Iran through central Asia to Kazakh and Uzbekh. I added well over 200k miles on it myself, changed buckled axles and even a whole gearbox out on the high roads more than once. The logbook said it had over 5 million miles on it when I got it. And the company I sold it on to, was still running that same bus as recently as 2020. Covid did for that bus (company) otherwise it would still be running today. There are no trams on earth that have done the sort of starship mileage that is commonplace in those regions.

Please, for the love of god, take a step back. You are post-justifying a system which you truly desire in all your heart, and have already totally committed to emotionally. You are seriously suggesting building an entire transport system, complete with rails and electricity, as a substitute for a bit of street signage. If you are that desperate for location cues, you can buy a sign with a massive arrow and the word BUS and CITY CENTRE, and stick them every 100meters, for £50 each. As opposed to £200,000 for the street works.

This is not rational. Light rail is a religion.

“Anywhere you’d need a dedicated bus Lane”

And where is that, exactly? You are imagining a fixed built environment. But it isn’t.

Build a 500-place care home for dementia. If there isn’t a bus stop directly outside you have consigned those inside to never seeing family or friends again. Because family are too old to drive, that’s the reality. How many decades before a tramline stops there? The football ground that needs 50,000 people to get to it, and the Council builds a light rail station. Five years later, they fall out of Champions League, and attendance drops to 5000. Ten years later they go bankrupt. What happens to the £50M rail system? It’s a ghost town.

Your planned economy does not work. You cannot build fixed infrastructure for a living city.

You cannot forget the economies of scale. That’s the single most important fact about economics.

“start producing trams to the same standard build, like buses”: this will not happen. Ever. Every local tram system, everywhere, feels the need to specify a custom-build with some little twiddles their local mayor decided was necessary. This is a sociological problem, not a technology problem. It is driven by the incentives of decision-making bodies. Do not make the mistake of thinking that the laws of human nature are any weaker, or ignorable, than the Third Law of Thermodynamics.

The only way to break this link, is to design systems that are small enough, boring enough, or invisible enough, that nobody wants to fiddle around with it. Small systems are efficient. Large systems are broken by design.

Nonsense, for so many reasons. If it’s electric vs diesel that worries you…..the vast majority of buses being bought today, are electric, on a replacement basis. See how easy and flexible buses are? You don’t need to change the entire infrastructure, just to change the motor technology. What *possible* mechanism is there, that would make an electric motor in a bus require maintenance calls every other day; while an electric motor in a tram sized for the same load magically “just works” for 30 years?! This is magical thinking.

I’ve no idea why you think diesel buses only last for 5 years. They can (and do) last for thirty years. The reason why most of them are much newer than that, is that the internal refurbishment cost makes it often more economic to replace after 10-15 years. Whereas with a tram….the internal refurb wears at exactly the same rate, but it’s so expensive that there’s no money in the budget for a replacement. So you are stuck with the same geriatric broken seats for much longer. They. Are. The. Same. Object. They. Work. Identically.. The only difference is that one is coach-built in tiny quantities and the other is built using modern volume production techniques.

Trains and trams also cost much more than buses to service. Buses are serviced by mechanics you can hire in any town in the world. Servicing a locomotive or tram is a skill you to pay to the moon and back for, for a variety of reasons including unionisation. Train and tram parts cost 10x what the equivalent bus part would cost. As mentioned, the drivers are more expensive.

A bus stop not only costs a couple of thousand to stick up a pole and a bus-shelter, but have you noticed it has no maintenance staff? A train station not only costs tens of millions to build, it requires a continuously employed staff *just to clean it* let alone anything else.

*Everything* about the ongoing running costs scales brutally with rail, in a way it just doesn’t with bus network. And in the end, it comes down to one simple reason: once you have decided to build a train, you are locked in to a specialist system.

Funnily enough, that even extends to energy costs. If you look at what the electricity costs of the rail network actually *are*, which is publically available info, you will discover that they pay more per passenger-mile than a bus pays in petrol. Because, again, there’s speciality infrastructure that all needs to be maintained; its not free to provide on-demand high-availability 2 MW power per train for its peak demand during acceleration, not free at all.

And yet, the lighter tram costs £1.5M for the identical effective capacity to a bus which costs £300k. Those are the real numbers.

Then build in dedicated bus lanes / dedicated bus roads. Did I say you shouldn’t?

“if you're doing that, why not put in electric wires and power”. Because it’s just irrelevant. The key point for everything is cost, which is driven by *volume* and production-lines. Buses are commodity generic items, you can buy them from anywhere, including Mexico, and they cost £300k. A tram costs £1.5M. Exactly the same thing, same capacity, in fact it’s “simpler”, the only difference is production volume. That’s it. That’s all there is.

Similarly, train and tram drivers make over twice the hourly rate of bus drivers. Because it’s a specialty thing. Again, that’s it, that’s all there is. *You* can drive a light goods vehicle. You can’t drive a tram.

Why are you obsessed with putting metal rods on the ground? That’s the give-away that it’s a religion, not technically-based. We can guide vehicles with centimetre accuracy using radio or GPS, that’s just a solved problem. Public transport and rail/tram have nothing to do with each other. There’s a strong argument that the catastrophic economics of rail/tram have prevented public transport being rolled out more widely for over a century, because they suck all the funding from buses, which are 10x cheaper.

The only reason why rail/tram gets built, is the political one of “naming plaque fallacy”. If you build rail/tram, the infrastructure is a big visible project. This comes with lots of fanfare and election manifestos, loads of project manager jobs that can be given to your mates, plus the chance to have an opening ceremony for the tram and railway station, with a plaque you can put your name on. Buses are driven by working class people, don’t have project managers, rolling out a new route involves one procurement person to buy the bus off-the-shelf,

and a summer-student to print out some timetables, and there’s no vanity signage on bus-stops.

The other claim “trams use less energy than buses” is obvious rubbish. Rolling friction is not a significant part of the fuel usage of a city vehicle, for something that stops and starts every few hundred meters.

European chip lobby seeks more government cash and policy clout

Re: Shocked of course

But the U.K. doesn’t have any native fab capability *because* we have the experience of why it fails. We’ve *experienced* Silicon Glen and come through the other side. Remember,Thatcher ripped down our coal industry and invested the proceeds in Motorola and others at Silicon Glen. Anyone who tells you, that the Tories “don’t do industrial investment/picking winners” has very selective memory. We pissed a vast firehose of cash against the wall. Lost it all, and learned the bitter lesson of exactly why you don’t do that in silicon.

We in the UK learned our lessons thirty years ago, *that’s* what entitles us to say “I told you so”.

In fact, many of us have been around much longer than that. I was in at the death of VEB Dresden. There’s a very good argument that the subsidy required at Dresden was so large, tens of thousands for every East German citizen, that not only did it do for Honecker, it was partly responsible for the fall of the Berlin Wall.

Re: "more say over policy decisions"

But you were the one who supported giving all that money to them, and sneered at those who tried to explain to you. You and a dozen others on this forum, are on record telling everyone what excellent industrial strategy for the EU. Whereas UK people who actually worked in the industry making silicon, told you that they had seen how all this ends up, because we’ve seen the whole game play out many times before. I was there all the way back to VEB Dresden, and through Silicon Glen. Told. You. So.

Have we stopped to think about what LLMs actually model?

Re: are we sure what we do is more than autocomplete

“We don't need vast quantities of training data to do what we do”

Quite the opposite. Babies need *vast* amounts of training data to learn language. Babies hear between 13 million and 45 million words, by age 4, to gain a vocab of just 1500 words. That’s an average repetition factor of 10000-30000. Children with a repetition factor below 2000-5000 are literally considered deprived and at risk of being taken into care; and even at 10000 their educational outcome is noticeably depressed.

So, no, humans need *a lot* of training data. It gets less as they get older, as they learn to learn.

https://great-start.org/wp-content/uploads/2020/02/Understanding-the-30-Million-Word-Gap.pdf

Re: "Whether we burn $500 million a year or $5 billion – or $50 billion a year – I don't care"

You are being silly. There’s dozens of other “human physiology front-end” problems. The majority of optical illusions fall into this category. The retina does a bunch of front-end pre-processing, such that the true image doesn’t even *reach* the brain. You see a purple edge-image next to yellow patch, because that’s literally the set of neuron impulses going into the brain. Since the brain never “sees” the true original retinal activations, how could it possibly get the right answer to “what is the colour of that edge”?

If you know a bunch of human brain physiology, it’s easy to create visual, aural, and tactile illusions from scratch. Understanding how cochlear cells encode sound harmonics, allows you to predict the Shepard “infinite rising staircase” sound from first principles. These are white box hacks for the human brain. They tell you a lot of interesting academic detail about the underlying processor…..but nothing of interest about “intelligence”. If an LLM trained on auditory speech, either is or is not fooled by a Shepard tone in the same way that a human is……then so what. What conclusion do you draw? That LLMs “can’t understand music”?

Suno and Udio disagree. The music they generate may not be what you want to listen to, and you may not be happy that it exists at all, but burying your head in the sand and saying “I am defining this as not-music, therefore it doesn’t compete with real music, therefore it’s all just waste of time I can declare myself above” is not going to get you anywhere. You are just getting dragged into pointless debate about angels dancing on pinheads, while the technology does what it does.

Re: "Whether we burn $500 million a year or $5 billion – or $50 billion a year – I don't care"

“What does that mean? I can't pick them up?”

Yes, exactly. Sorting them physically into piles, is a known strategy to cope with human well-known cognitive lack.

There is an equivalent trivial strategy to get ChatGPT to count the R’s in strawberry flawlessly. Just prompt it to write a Python script to count the Rs in any word; that Python script will be correct. Then ask ChatGPT to run the Python script on strawberry. It will give the correct answer. Simple. You just need to understand the tool you have; and want to succeed at the task in hand, rather than trying to catch it out.

“Asking someone to count 237 items will take them some time. Asking them to count three items is much easier”

Wow, you’ve fallen right into the anthropocentric trap. Why is counting 3 items so much easier for a human? Because the human brain literally has specific circuits for counting in blocks of 2, 3, 4 and 5. Humans count numbers like 237, by using iterative pile-sorting plus the native small-number-recognition. Standard tallying on a bit of paper with four bars and a cross through for the fifth one, is literally that.

You are changing the task from one that demonstrates that *humans can’t count* to a different task that uses a different bodged strategy.

Re: "Whether we burn $500 million a year or $5 billion – or $50 billion a year – I don't care"

“Ask it how many r's are in strawberry”

That’s a very specific hack, based on a white-box understanding of how LLMs work. LLMs use tokenisation, seeing “strawberry” as a single token. Beyond the front-end, it doesn’t see the letters in “strawberry” at all. It’s exactly like asking a French guy how many r’s in strawberry, and he answers “one” because he translated to French using google translate, and there’s one R in “fraise”. Slow handclap for being a smartarse, but nul points for using it as a metric of intelligence.

An equivalent proof that *humans* aren’t intelligent is to show anybody a collection of 237 identical objects arranged randomly on a table, and get them to count them without any physical interaction. They’ll tell you a different number every time because they “lose count”. There’s no equivalent failure mode for LLMs, therefore humans aren’t intelligent.

“If anything, it proves how dumb the average human is”

I mean, yes. How does that show that LLMs aren’t *useful* as a tool? Given that dumb humans are doing those tasks already? Chatbots aren’t normally intended to replace what I would call humans. They replace something entirely different: “humans reading from scripts, written by companies”.

Did you actually properly read the article, or only extract your preconceptions? There’s a couple of really good points that Birhane made “ The idea is that cognition doesn't end at the brain and the person doesn't end at the the skin. Rather, cognition is extended. Personhood is messy, ambiguous, intertwined with the existence of others, and so on,". Spot on. And this “messiness of personhood” goes both ways. A human executing a business process robotically, is not really a human person. It’s something different, entangled with “the company”, as a persona. We even have literally different voices to speak: those with friends, those with parents, with colleagues, with customers/clients. It’s called code-switching.

So, can LLMs achieve the same accuracy as “person required to speak a business process, who gets fired if they don’t parrot the corporate speak and policy the way they’ve been told to?”. Probably yes, and possibly better than the human currently imitating that persona (not just cheaper). Is this “valuable”? That’s capitalism for you. If you disagree on the intrinsic value, then it’s capitalism you have a beef with, not LLMs.

EV sales hit speed bump as drivers unplug from the electric dream

It costs a thermodynamic minimum of 13% just to chill and compress the hydrogen, to the normal 700bar storage pressure. We get surprisingly close to that in practice, an astonishing 83% efficiency. Standard electrolysis is about 70% efficient. There’s a whole bunch more stuff too.

Hydrogen is environmental vandalism.

If I had some bacon, I could have bacon and egg, if I also had an egg.

France charges Telegram CEO with multiple crimes

Re: "It looks like he didn't comply"

I don’t dispute that Durov should be subject to French law….that’s a rod he made for his own back. Why would anyone give a shit about some random grifter running a dodgy platform, going to jail? This isn’t about Durov, this is about Macron. And also about how (if at all) one can still run a business in France.

#1 Quite clearly, Telegram has not magically changed into a cess-pool overnight. It has always been the preferred platform of Russian mafia and other unsavouries. Macron knew this three years ago, when he gifted the citizenship. At a minimum, this shows Macron’s hypocrisy. In a normal democracy, it is a *crime* to personally gift citizenship to people accused of running pedophile-enabling websites, whether or not Durov gave Macron a lot of money. In the U.K. and other democracies, Macron would offer his resignation immediately. In *any* democracy, Macron would be investigated by the police, and barred from the Elysee Palace until that investigation was complete.

#2 *Why* did Macron gift that citizenship? Just for money? Maybe. But Macron has access to vast coffers of money, he doesn’t need piffling bribes. It’s….a massive coincidence that he just happens to have called in the debt at the precise moment he has cancelled the democratic elections. And in addition, nobody could figure out why he called those elections, it looked like a massive mistake. Was it a mistake? Or has he planned this all along? Did this story actually start three years ago, and his plan was *always* to confiscate the media platform he needed, to operate the coup.

Maybe it was “only” bribery. Maybe. But it sure doesn’t look like it. It looks exactly like the planned conclusion of a long con.

Re: Viewed by some as an attack on free speech

The same one, yes. And Macron also personally intervened to give him French citizenship, in 2021. And then had him arrested.

Re: "It looks like he didn't comply"

He’s a French citizen? Is he really? Because it seems that this citizenship was due to the personal intervention of Macron himself in 2021…..

https://www.politico.eu/article/macron-finally-admits-he-gave-telegram-chief-pavel-durov-french-citizenship/

Re: "It looks like he didn't comply"

Is Durov really a “French citizen” though? Turns out that what *actually* happened is that Macron personally intervened to give him French citizenship. As recently as 2021.

https://www.politico.eu/article/macron-finally-admits-he-gave-telegram-chief-pavel-durov-french-citizenship/

Re: "French president Emannuel Macron used his X account"

President-for-Life Macron, who ignores election results? Are you seriously still going to pretend this isn’t a coup?

This is a carbon-copy re-run of the fall of 4th Republic, complete with patchwork three-party coalitions, arbitrary prime ministers being proposed by president.

None of this is new. Macron isn’t new. It’s a very French coup. And those in his patronage group will fete him as a hero, just like de Gaulle “defeated the Algiers coup”.

Re: "It looks like he didn't comply"

True. But when the target “just happens to be” Macron’s favourite mouthpiece platform (as Twitter was for Trump). And when, as is widely expected, the court judgement “just happens to be” a confiscation putting the platform in the direct control of the Elysee Palace.

And when simultaneously Macron has decided not to implement the results of the democratic election, refusing to accept any coalition that does not include his own party, despite that it was the smallest.

And when Macron arbitrarily decided not to even consider the election result for months “because of the Olympics”…..not a war, or a major natural catastrophe, just a pre-planned sporting event that the government isn’t even responsible for.

Then maybe it isn’t whatabouttery. Maybe this is exactly what it appears to be: seizing control of the media as the first step of a coup.

Telegram founder and CEO arrested in France

Re: "It is absurd to claim that a platform or its owner are responsible for abuse of that platform."

The CEO may well be responsible, but that doesn’t make it a criminal offence. It’s a core principle of modern societies that stuff which we do in the course of our job, is the responsibility of the company rather than individual employees being personally criminally liable. It protects you as an employee as much as the CEO. We couldn’t really have companies and paid employment at all without the protections of limited liability companies.

Having said that, I just checked something, and reverse my position…..seems like this Durov character has been hiding his “company” in British Virgin Islands, with a web of cross-holdings, such that it’s totally unclear who is the owner. He’s put a lot of effort into ensuring that there simply isn’t a company in France or anywhere else, such that government subpoenas simply can’t be applied. In which case, he has been hoist by his own petard. The man is the company, the company is the man, and the corporate veil has been pierced. In that case, any company penalties can be applied directly to him. Any CSAM material passed through his servers should be treated as if found on the hard drive at his home.

Durov did this to himself.

Unique to France and other fascist dictatorships, yes.

There are of course many other fascist dictatorships in the EU, as 14 out of the 27 are currently no longer true democracies with Rule of Law. You probably represent one of them, which is why you are making excuses for them.

Re: "It is absurd to claim that a platform or its owner are responsible for abuse of that platform."

Then you initiate court orders against the company. Or go in heavy-handed and physically confiscate data from their servers. Or if you want to get really tough, you confiscate the company and appoint administrators. But you don’t arrest the *director* of a limited liability company, who has not personally committed a *criminal offence*, and put them in custody. That’s when you stop being a state with Rule of Law, and it’s a simple dictatorship. That moment, right there. There’s a massive difference between doing things which are unlawful, and committing a criminal offence.

When an airline runs flights that you can buy tickets for, knowing that there are insufficient seats, that is unlawful. They have signed contracts that they have no intention to honour. It is unlawful. Now in France, the CEO of Air France can be arrested and go jail for fraudulent misrepresentation for overbooked flights? This is insane. You know as well as I do, that’s not going to happen. What that tells you is that the authorities are picking and choosing who to arrest, based on nationality, or whether they like them, or if they have a funny shape nose, or whether they’ve annoyed Macron in some way. That’s what we found out today, in public. There is no Rule of Law in France any more. Some people will get arrested, some won’t. Which group you are in has nothing at all to do with what crime you committed, and is determined only by who you are friends with.

Re: "It is absurd to claim that a platform or its owner are responsible for abuse of that platform."

No. It’s very sensible and normal to enforce court orders against companies, when the company is breaking the law. Civil penalties.

Putting the CEO in jail?! This is extraordinary. You know what a “limited company” is? How many CEOs have been put in prison for eg their building cladding being responsible for dozens of deaths? How many in jail for food poisoning deaths? Is it…zero?

This is the owner of a media company. He is in jail purely so that Macron can take over Telegram and use it for President for Life propaganda. Telegram may well be used for many criminal purposes, and even rise to the level of being a criminal enabler. But the reason why *that particular guy is in jail* is because President for Life is taking over the media outlets.

Re: I hope Musk travels to France

France has a president who refuses to appoint the democratically elected prime minister, of the party that opposes him, despite his landslide election defeat. He “deferred the transfer of power during the Olympics”; and now it will not take place at all. The same president is also now imprisoning heads of media companies, for crimes they didn’t personally commit.

It’s not a strawman. This is a coup in progress. Dictators are always elected, but ends in President for Life.

If this were sanctions against the company, that’s one thing. Arresting the CEO? This is a fascist coup.

Boom Supersonic takes baby steps toward breaking the sound barrier

Amdahl’s Law applies

There’s very little point reducing a six hour flight to three hours…..one you add a three hour check-in, plus an hour to get out of the destination airport. Plus probably an hour and half on each end travel-time to the airport. Then it only reduces 13 hours down to 10 hours. This is exactly why the rich now travel first-class, rather than caring about travel-time. And in addition, you can now remain connected email and phone on the plane, which one couldn’t in the sixties and seventies.

So, what’s the point of supersonic? Makes no sense.

To crew, or not to crew – that is the question facing Boeing's stricken Starliner

Re: Lesser evil

“I would like a second safe option for crew transport”.

Hard disagree. Duplicating capability is reactionary, and a waste of human intelligence, skill and labour.. A safe effective solution exists for Crew Transport: Dragon. If somebody else wants to spend their own R&D making a competitor project, and thinks they can do it commercially at even lower price/higher capability than SpaceX, then absolutely they should go and do it, but without subsidy. NASA should be spending the budget saved on developing some new, different capability that is needed, but that we don’t have.

An obvious alternative project: most astronauts spend their time when they’re up there, either fixing things, or “doing experiments”. Things that could in principle be done by remote robot *from ground* with sensitive accurate manipulation. The distance is irrelevant: 460km round-trip light flight time is 3 milliseconds latency. I’m well aware that remotely operating is easier said than done. It’s not just about the manipulator, but how a human operator can get the correct “feel”. Great. It’s a hard problem, and a useful one to solve. A great use of $3bn of development money then. I’m not tied to this idea, I’m sure that there are half a dozen other genuine new frontiers to be tackled, for that money. Pick your own goal and solve it. My point is, duplication is lazy and pointless.

For example, Galileo. Galileo is a simple copy of GPS constellation. Do you really imagine that the universe has decreed as a fundamental law of physics that the *only* way to navigate, is to transmit a spread spectrum signal from satellites? Take a step back, how unlikely is that? Do you really think that a century from now, that is how we are going to be doing it? No? Then why not spend the *$10billion* that Galileo cost, and will continue costing every year, on developing the new nav technology whatever it is. $10billion is a lot of money, it buys a lot of science and engineering. Building cargo cult is such a waste. Do I know what the next Gen nav will look like….not exactly. But here’s one possibility: improved quantum-based accelerometers and gyros; chip-scale atomic clocks. Does it work yet, no, but eventually something like it will.

https://www.imperial.ac.uk/news/245114/quantum-sensor-future-navigation-system-tested/

https://www.usni.org/magazines/proceedings/sponsored/quantum-sensing-new-approach-maintaining-pnt-gps-denied

https://www.nist.gov/system/files/documents/2017/05/09/VCAT-NIST-CSADemo.pdf

Adam from Mythbusters

If you haven’t seen it, can I highly recommend Adam Savages YouTube channel “Adam Savage’s Tested”

https://youtu.be/MxyqMw-IDq8

He’s an extraordinary Maker, more so than you get to see on Mythbusters. He also talks about the practicalities of how running such a making business actually works, which inevitably touches on a philosophy of life, in a “Zen and the Art of Motorcycle Maintenance” sort of way.

NASA pushes decision on bringing crew back in Starliner to the end of August

Re: Why the delay?

It’s well-known that the Space Shuttle design targeted in 100,000 risk. By the time the first one launched, they estimated 1 in 280. Actual outcomes were: lost 2 in 135. Re-analysis *after the fact* by NASA was *1 in 9* for the first 9 flights, and 1 in 90 longer term. Risk estimates are always delusional. Decades of experience shows that the only way to do this safely, is to fly a lot, very frequently and cheaply, and fix the failures as they occur. Anyone who tells you different is on cost-plus contract.

By the way, I see nothing wrong in 10% risk for first 10 flights; 1% first 100 flights, 0.1% first 1000 etc. There are tens of thousands of intelligent, motivated, successful adults who are willing to take such risks, to meet their personal goals. Those people are perfectly capable of both understanding their risk, and deciding rationally on their own goals. Many people do free solo climbing for example. *Pretty much everybody* who free solos routinely, eventually dies doing it. The very best may survive 20 years; the second tier may only last 2-5 years. But almost nobody makes it out alive.

If this were a reusable vehicle……

Then you would just fly it unmanned down to Earth, check the thrusters, replace them if necessary, turn it round in about three weeks, fly it back up.

In fact, you’d have done that over two months ago, as soon as you knew there was a problem. For the price of the petrol, when lives are at risk, of course you’d do that. Falcon9 turns around in about two weeks nowadays. But Starliner is sold as “reusable” with a turnaround time of…..six months. I beg your pardon, WTAF? Plus a “service module” which isn’t reusable and contains the thrusters and power, ie all the stuff that makes it a spacecraft, and not just a tin can with seats.None of this should have ever been approved, for exactly the reasons we are seeing today. This just isn’t the way to do engineering.

SiFive offers potential Neoverse N2 rival – the P870-D RISC-V core for datacenters

Re: China

Sigh.

“SiFive has modeled out the P870-D characteristics on TSMC's 3nm”. That’s so cute, LOL. I’ve “modelled” my roll-your-own core on 0.1nm technology; not only does it beat everyone else in power and performance, but it has 100% yield, and works on first spin with zero IR droop on all process corners. Oh, you were expecting *measured proof results* on real silicon….me myself having paid upwards of $2bn of my own money for the mask-sets first four silicon spins, such that *you* will have perfect silicon first time. Ummm, no, I was hoping you’d do that? What’s that, you don’t want to pay me for the privilege of debugging $2bn worth of development costs on my core? Why not?

What a load of total muppets.

Akeana debuts RISC-V CPU designs on $100M budget, longs for an Arm wrestle

So….if they are only selling cores (ie licenses\royalties) what’s their license fee going to be? If it’s 2%, to even match ARM, they’ll need $5bn in chip sales to even break even on the $100M already spent. That’s 10x what SiFive managed before going broke (sorry, “pivoting”). Not going to happen.

The truth is, ARM charge a suicidally *low* license fee, which only works because they make it up in staggering volume. Nobody can compete with that, and remain in business.

But it’s much much worse than that for a newcomer. If Akeana think they can charge even 2% for an unproven core, they’re about to get a rude shock. It’s not about whether the data sheet shows it has the performance, or even whether it “works”. The main question, and economically almost the only one that matters is - what is the yield of this core on TSMC 3nm. If it’s even 85% yield, while ARM are getting 95%, then the silicon manufacturer is paying a 10% tax just from using the core there at all! Akeana are about to discover that their “customers” expect *Akeana* to pay *them* money to use customer silicon as trial yield proving vehicles. Or at very best, an equal “partnership” where Akeana get no money, just “paid in exposure” by Marvell or Qualcomm or somebody. Whatever happened to those Qualcomm and Samsung RISCV cores from three years ago…dropped without trace after a few PR puff pieces…..yeah, that exactly.

Ah yes, that “paid in exposure”, so un-beloved of unpaid interns everywhere.As others have found out, there’s just no business model here.

Faulty instructions in Alibaba's T-Head C910 RISC-V CPUs blow away all security

Re: Obit

No there aren’t.

T-Head SOC is by far the * highest quality* RISCV on the market with the *most* testing. Few RISCV implementations do anything but emit smoke out of the box. TBH, few RISCV boxes even have anything in them at all, other than a note saying “To Follow when Parts are Available in 5 months time”.

Re: This is likely just the beginning

“The problem with microcoding a processor like RISC-V is that while it may make problems like this easier to fix, it adds a whole load of complexity to the chip design, requiring bigger dies and more power, and this is the antithesis of a pure RISC design.”

Which is exactly why “pure RISC designs” don’t exist, and everybody has been telling the RISCV evangelist morons that for over a decade. The additional complexity you describe exists, but there are payback advantages in practice, which you don’t appreciate until you’ve been doing this as a day-job for 20 years. And instruction decoding *sounds* complex to noobs, but it’s actually tiny relative to everything else. Really tiny. Honestly, so small it’s barely measurable in practice. The complexity of the entire instruction decoder circuitry is maybe 2% of the cost of the branch predictors. Which is only a fraction of the whole CPU core. And the whole chip area is dominated by *cache area*, with the entire CPU cores relegated to finding bits of spare silicon to fit nicely around the caches to minimise track lengths. Making a CPU without microcode is just *dumb* from a modern perspective.

Re: Only on T-Head's CPUs

But that’s *exactly* the point, why RISC-V principle is broken by design, not just in this.

Any of the other CPUs, you know who you’re getting it from. If it’s Intel x86 CPU, and it’s broken, Intel are both your supplier, implementer and responsible. If it’s ARM CPU, and it’s broken, it’s ARM responsibility to fix the core.

But these are *RISC-V* chips. They are generic. It’s just not possible to say “RISC-V chips are safe”. It’s a meaningless statement. There are a thousand implementations, some of which are broken, some not. You almost certainly do not know what the device you have on your desk is, let alone whether it is vulnerable, and no way to check it. If, for example, you are running with these Scaleway cloud people, how would you know? How could you know if it were newly fixed silicon or not - because even *they* don’t know.

I’m glad the penny is finally starting to drop. But I’ll just add another point: “The RISC-V oversight body”. There is none. Doesn’t exist. RISCV org writes the paper ISA description, sure. But there’s no conformance spec, no conformance testing, no certification. It’s totally unlike professional engineering specs like Ethernet or Bluetooth.

Infineon announces layoffs as Q3 results disappoint

Re: "weaker-than-expected economic performance"

Infineon management and board are perfectly capable of forecasting, and steering a sensible strategy. There are plenty of white papers online, laying out what management would have done, and that they opposed the decisions that are now causing job losses. However, like most European large caps, they don’t control anything any more. The EU Commission tells them what “European industrial policy” is, and management implement it. If the management *didn’t* JFDI as ordered, they’d be replaced in a matter of months.

Cf. Siemens ludicrous “decision” to buy Gamesa. That one was directly instructed by none other than a certain Thierry Breton. His “judgement” was so good, that the German government ended up bailing out Siemens for €8bn just a decade later. Speaking of which, how’s Breton looking for French Prime Minister these days? Given that his party wasn’t even in third place, but that doesn’t seem to matter in French “democracy” these days.

EU AI Act still in infancy, but those with 'intelligent' HR apps better watch out

I’m sorry that you had a tough time of it. But the reality is that many people *can* do those jobs, which means that there’s really no good means of selection. Seems you would have preferred a human face to interview you….but the evidence is that interviews by humans are also little better than chance at predicting “success in the role”.

The job of interviewing should be targeted at one thing and one thing only: making sure you don’t pick anyone in the *worst half*, because picking the actual winner is impossible. And that’s why the advice for interviewers always centres around “red flags”. A single red flag is a no-go, even though the chance of them being the actual best is not so very different from the others; but you want to avoid any in the worst half, and that’s difficult enough.

The evidence base is that the only thing that predicts success in the role are either: objective exam skills-based on task (which I’m afraid doesn’t really apply to B&Q). Or “has done good work for me in the past; has worked for my friend who I trust and gets personal recommendation”. Jobs for the boys, basically, with all the problems that brings.

So, the honest thing for B&Q to do, is just roll a literal dice and anybody who gets a double 6 gets hired. This doesn’t make anybody feel good.

Re: Simple idea

Alphafold, for prediction of protein structure and therefore binding site structure

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10136805

You’re right, though, Alphafold has been used for “years”. My company designed and uses what is essentially Alphafold in *non*biological engineering production, has been doing so for nearly two years. So I guess that’s “years”

If that’s all you got from your reading of the AI Act, then you haven’t read it at all. If this really affects you, and you seem to think it does, get a legal firm, a good one, and get them to walk your leadership team through it line by line, with Q&A.

Don’t try to do it on the cheap, anyone who knows what they are talking about is going to be charging £1500 per hour, and it’s a 100 hour job for a partner plus couple juniors. You aren’t getting out of there for less than £200k, and that’s if your business is small (<50 people) and simple. More complex businesses will be spending between tens or even hundreds of millions in legal fees just *analysing the legislation of how it applies to them*. Not implementing anything, just the legal fees.

You’ve *seriously* underestimated this if you think it’s like GDPR.

Is AI going to pay its way? Wall Street wants tech world to show it the money

This is nonsense. All of them got *extraordinary* good results by normal standards. A minimum of 15% annual increase in revenue and profit, for companies of that size, is unheard of. Let alone NVidia, up 260% revenue, 630% profit.

All that happened is that the increase isn’t as fast as analysts had budgeted in. But share prices are still higher than they were in May/June.

If AI can bring 15% growth rates, against historic 2-4% GDP growth, that’s unprecedented. But what *is* a problem, is that US employment data has been poor. That’s why share prices dropped.

Rising corporate income, is needing fewer people to do the work. Henry Ford knew that the population needs to be able to afford the cars they build. It’s just starting to dawn on people that we cannot avoid this reality. UBI is no longer a fantasy, it’s something we have to plan for. Maybe not today, but 10-20 years out. This bit of human history is going to be really rough. Groups that have historically been middle class and entitled will find that their skills are worth less than hairdressers and carers, and they aren’t going to like it.

Europe launches 'AI Factories' initiative in hopes of competing globally

Re: +++ EU TRIGGER WARNING +++

“Name some advantages”

Not having the EU AI Act. Specifically, not having to hand over your proprietary code and weights to the EU Commission for “checking” and “consultation”.

The company that I own, designs and uses AI for internal purposes. Not chatbot, or code completion, or HR, or security, or “fintech”, or “language” at all. Not do we sell it, or export it in any way. It’s NOYDB how we use it, and that’s my point. We are rather profitable, and thats all you need to know.

As far as I know, we are probably compliant with all the “low-risk” gubbins. But I have no interest in finding out how that mission will creep. If there were even a sniff of us signing up to that BS, I would simply shut that part of the company down, 9am Monday, and set up a satellite office in USA. End of story.

I’d absolutely wipe our UK servers too, to make sure the Commission didn’t get their hands on unearned wealth. Again, I have no interest in handing out our IP to potential competitors. We dont trade with EU, although we otherwise sell globally, over half is Asia Pacific, rest is mostly Americas, so we have no value at risk.

Re: +++ EU TRIGGER WARNING +++

This sort of comment is *exactly* why we left. It’s an article about the EU taxpayer stumping up money on some project. One can have a rational discussion about whether it is a valid goal; whether the execution is likely to be sound; and whether it would be value for money. I have worked on, and managed, a large number of these EU programs, so I have relevant opinions and observations about that.

But in fact you are *entirely uninterested* in the project output, as is the EU in general. All you want is to hold a toy aloft, and poke the child next to you “nobody’s having my toy”. If we were inside the EU, that’s all we would get - a chance to hold the toy in the air, and yah boo, China/USA sucks, eat deez nuts. If you’re rational, that’s a very bad deal.

Meta's AI safety system defeated by the space bar

You’re acknowledging that the so-called alignment techniques (ie play with the weights) work in R&D, but don’t work well in practice.

So I’m curious what you’ve got against the standard techniques that pretty much *do* work in production, and have done so reliably for over a year now? Viz, sanitise the inputs and outputs using GOFAI pattern-matching. Perhaps not in the chatbot “Can I persuade ChatGPT to say that Gandhi was an alien” nonsense. But for 90%+ of the LLM use-cases that *aren’t* chatbots, and many of which aren’t even “language”, works just fine.

Re: "the social media biz is working on a fix"

The *guardrails* are usually hand-coded parsers, on both input and output sides. See below for some concept code

https://github.com/guardrails-ai/guardrails

In principle, doing smthg like “work out the weight-vector of naughtiness, and turn it down a bit”, works reasonably well. In practice, that doesn’t work against adversarial jailbreaks.

The secret to better weather forecasts may be a dash of AI

Re: 2.8°C

Kelvin vs eV……not since 2019! Kelvin redefined to be a proper unit….

https://en.m.wikipedia.org/w/index.php?title=2019_redefinition_of_the_SI_base_units