Posts by jmch 3705 publicly visible posts • joined 6 Mar 2017
"Every time a customer buys a printer, it's an investment for us. We are investing in that customer, and if that customer doesn't print enough or doesn't use our supplies, it's a bad investment."
That's basically admitting that they're not making any money from selling a printer. We are investing in that customer literally means they are subsidising the printer sale and seeing that loss as an investment to be recouped on ink sales.
Saying an energy breakthrough is needed for AGI is getting it completely backwards.... it assumes that the current statistical models are the way to build AGI, and just throwing more data and processing (and therefore more power) at it will get there.
"OpenAI's old GPT-3 system reportedly consumed 936 megawatt hours (MWh) [to train]"
A human brain consumes around 20W, and even assuming the "training time" is every living moment from birth to a wise old 80 years, is a total consumption of just over 14MWh. To arrive at a reasonably competent 20-year old trained brain consumes 3.5 MWh. So real human intelligence is not only far better than current-best LLMs, it is also more than 2 orders of magnitude less energy-intensive. Also interesting to note that humans do not need giant volumes of text to train, and indeed being able to memorise and regurgitate vast amounts of information is an exercise in memory, not intelligence.
What is needed is a completely different approach, not 'more-of-the-same'... and (in the spirit of being constructive not just criticising), I would note that what humans ARE being trained on is vast amounts of what is not only 'video' but "3-D" / depth-perceptive video combined with all of the internal sensory input that informs the brain about the current state of the body. As far as I can work out, most AI researchers are trying to build a 'superbrain-in-a-jar' kind of intelligence, but I think that it won't be possible to build that completely divorced from physical interaction with the real world.
"straightforward translation to bricks-and-mortar ethics"
I pass by a warehouse where the key is stuck in the lock on the door outside. I open the door to go inside and shout "Hallo, anyone there? You left the key in the lock!!". What that would usually result in is a "Gosh, thanks, I forgot that" - not a police report for trespass, and certainly not even a prosecution let alone a conviction.
" When they can sell every car there's no need to reduce prices."
And yet, "they" can't keep selling every car they make at the current price because at some point all the richer people will have one, and the less-well-off people who can't afford that will still want one. And what will happen is that "they" will introduce a lower-spec, cheaper model to supply that market. This has already happened.... The original Tesla roadster cost $100k at the time. The first Model S (which by the way is over 10 years ago) sold for less than $60k. The first Model 3 sold for $35k.
And if "they" isn't Tesla, it will be one of the dozens of electric car manufacturers around the world or some new entry from India
"Baseless assumption based on literally nothing."
I think it's a pretty safe assumption, based on a couple of centuries of technological development, and a working knowledge of capitalism. Even if ICE cars are completely banned, electric car makers are still competing against each other for market share. The developing world is getting richer so demand for cars and other vehicles will anyway only grow. The high cost of battery cars is tied to (a) new technologies, which have to recoup R&D costs and/or buy patent licenses and (b) battery cost. The technology cost will continue to sink for older technology while the best and newest is more expensive.On the battery front, battery cost is less than a fifth of what it was 10 years ago*, even as capacities have quadrupled.
When there are customers willing/able to buy only a cheap electric car but not an expensive one, you can be sure that some car company will fill that niche.
* https://www.statista.com/statistics/883118/global-lithium-ion-battery-pack-costs/
"you seem to be forgetting that 80% of the *actual* world population does not live in developed countries."
No, didn't forget that at all. The vast majority of people, even in poorer countries, live in cities or suburbs. All the most populous countries are in warm or temperate regions with the majority of population in coastal plains, not mountains. It's also true that the vast majority of them could never afford a Tesla or even a e-VW. But many of them could afford a BYD or a small, locally produced electric tricycle-type vehicle.
"Current EVs cost too damn much"
Even considering cheaper running costs, the unsubsidised sticker price is still quite high... but it's getting closer and closer to a similair ICE car. The real question with regards to long-term costs is how much value it retains after 5 or 10 years. Currently 2nd-hand prices are low because of doubts about battery longevity (in fact AFAIK Tesla buys up used Teslas to recondition or change battery before resale, as otherwise they don't sell well), and that means a very high annual depreciation. If battery life of the newer cars proves itself stable after 10-15 years and the car keeps a strong aftermarket value, the overall annual cost would drop even with a high sticker price.
"something like a three wheel tricycle with a maximum speed of 30-40kph, but with enough power to lug 400kg of crops or the wife and three kids 20km to a market town ... and back"
Absolutely... an electric Piaggio Ape*! And in many other cases even in the developed world, an electric bicycle or tricycle can be more practical than an electric car.
(Not ape as in gorilla, but pronounced A-pay, in Italian it means bee)
"the current crop are best suited to users who:"
You're absolutely right on all counts. I would point out, however, that "Live in mild climates", "Lives in cities / suburbs" and "lives in a non-mountainous area" covers at least 70-80% of world population. In other words, EVs as they are already comfortably cover the needs of, I would say, a good 20% of world population, with a further 20-30% for whom it's 'good enough'.
"Future EVs will be better. Probably on all counts. But it'll be a long time (maybe never) before EVs are the right vehicle for every use case."
Absolutely they will get better. The only way future EVs will completely replace EVs is if they can have an 800km range with 10 minute recharge time (meaning >150 kWh battery capacity, >800kW charging power and 500 Wh/kg specific energy density). Theoretically possible, but practically???
Oslo temperatures last week were around -20C.
At that temperature, diesel engines can also start giving trouble, and I believe petrol freezes at around -40C. So -50C is an extremely hostile environment for all vehicles anyway, requiring special antifreezing additives to the fuel as well as to other required fluids (hydraulics and, ironically, cooling fluids). I guess the options in order are:
- if you live somewhere where it gets ridiculously cold, just don't get an electric car
- if you live somewhere where it gets ridiculously cold, just don't go outside when it's ridiculously cold
- if you live somewhere where it gets ridiculously cold, just don't!!!
"By centralizing ALL software the Apple and Google app stores have solved that problem for developers."
...and have also created another, that is, there are now tens or hundreds of thousands of apps on each app store. There are dozens of apps for even the most trivial functionality. For more common uses there are hundreds of apps. There is really no way a user can properly search through all of that, and the search functionality in-store is (possibly intentionally) crap. Many times I have searched for an app using a specific name, and the app either doesn't show at all, or is way down the list, with a bunch of competitor apps (or completely irrelevant apps) higher up in the list. The searching and ranking algorithms are, of course, highly secret but it would not be surprising if they are prioritised by whichever one gives more revenue to the store.
The 'Best Buy model you describe is quite different. Any computer store could offer a variety of software and games, and developers could sell their software at any one of those stores, as well as online or mail-order with complete freedom. The 50% retail cut is a 'normal' retail cut for a physical store that needs to pay rent and hire staff in a very high staff:sales ratio, and stores could compete against each other by taking a lower cut.
Apple's 30% isn't in itself monopolistic, what's monopolistic is the ban on competing App Stores. The 'payment provider' part is only a small part of it. Apple's insistence that all Apps have to be installed from their App Store is like if a car manufacturer disallowed servicing by any independent garage *that was competent to do so*. I completely agree that having strict App Store rules protect the Apple ecosystem, but in a true competitive market, anyone could run an iPhone-compatible App Store *as long as they could demonstrate equivalent security to Apple's*
"Apple: You signed a contract saying you would. We are suing you for that, and then we get to look at your files. We will prove it for you"
The 3rd-party payment providers can simply not store any referrer data (they are not obliged to keep it and usually this data is kept only for marketing). If they can save themselves 27% by dumping the data as soon as a transaction executes they surely would do it.
"Cupertino insists on collecting a cut of sales, up to 27 percent, even if a non-Apple system is used."
Surely that is going against the spirit, if not the letter, of the judgement? As I understood it, Apple demanding on-platform payments be made through Apple Pay is not monopolistic if links to outside payment providers are allowed, so surely insisting they get a cut of payments through externally-linked systems makes the whole system monopolistic?
A little thought experiment here.....
How about the AI take over the jobs of world leaders and all the circle-jerk wankers who flock to Davos???
We would probably get considerably less grandstanding and willy-waving (and hence less conflict and wars), and surly however badly the AI screws up (as surely it will), it would still not be on as grand a scale!
True story - a couple of years ago my uncle was in his garden and came across a cheetah. Carefully backed away, got inside and called the police. A few urgent enquiries later, it turns out the cheetah was being kept as a pet by someone in the neighborhood and it had decided to go walkies!
"The problem of EU is corruption. Too many people want as big slice of the pie as they can get their paws on."
And not in China or the US? pull the other one!!
The real problem with tech development (as well as many other things) in the EU, is that the priority for most people and countries is always country first, EU second (or further back), while in the US and China it's country first, second and third.
"The lawyers of thermodynamics would like a look at the contract"
The laws of thermodynamics don't need lawyers, they will take their tax anyway. But it seems trivially true that with an operation requiring cooling and an operation requiring heating transferring heat from one to the other, that whatever losses there are in the process, it is still more efficient than each of them doing their heating / cooling independently.
"What I haven't been able to find is how the deal is structured, ie who pays for Deep Green's electricity? Them, or the leisure centre?"
Whatever the contract details, since partnering together means lower energy requirements for both operations combined, there is scope for both to benefit. In any case I assume that "save the site more than £20,000 ($25,430) a year on heating bills" is a net figure (although it would be certainly in keeping with UK government projects in general if the private company were getting all the benefits of the arrangement to themselves).
Doesn't really matter. Much as I hate the idea of Trump 2, it's going to happen. Biden got elected as a centrist, but has allowed himself to be sucked way to the left by his party extremists, hence his high disapproval ratings. His age is also definitely a factor. He should have already stepped aside and allowed someone with a better chance to run, but his pride is getting in the way.
On the other hand Trump has used his vast experience as a fraud, shyster and conman to convince Republican voters that he will deliver what they want, even though in his previous term he delivered nothing of what he had promised them. Rather than draining the swamp, he appointed more insiders. He didn't seek funding for his wall, and in the end built about 50 miles of it. More immigrants were deported under Obama than on his watch. He didn't pursue any legal cases against Hillary Clinton, and then blamed his AG for that. Etc etc etc. This time round, he has been non-committal on abortion rights even as the other Republican candidates loudly beat their chests on the issue (one of his talents as a conman is that he knows how to read his mark i.e. the American Public, and he can see that coming out fully pro-abortion will hurt his chances). He also wasn't required to attend in person either of the 2 hearings against him last week, but he went anyway and loudly complained about being forced to go, turning his own frauds and illegal actions into a vote-winner for the dupes that are his marks.
The left, on the other hand, is completely failing to read the room, and keeps pushing it's critical race theory claptrap on everything, while de facto advocating completely open borders, defunding the police and third-trimester abortions. The vast majority of sane Americans not living in a woke opinion-reflection bubble reject all of that. A more balanced (and younger!!) Democratic candidate supporting reasonable European-style abortion rights, stricter (but still fair) immigration rules, demilitarising (rather than defunding) the police, and meritocracy rather than reverse-race-ocracy would handily beat Trump, but such candidates have been weeded out from important roles in the Democratic Party because they don't conform to the left groupthink.
Sorry, off topic, rant over!!!
"The US system of chosing a president isn't democratic in the first place."
Not at all. The 'some votes are worth more than others' is a known issue in every single representative democracy. Winning the presidency / government while losing the popular vote is a possibility in any system, whether it's first-past-the-post, proportional, single transferable vote, whatever. The problem is just made bigger when you have only 2 parties.
The US electoral rules might be weird, but they are also known to all participants, all of whom can devise an electoral strategy that will optimise their chances of victory based on a known set of rules. If anything, the bigger issue is local district gerrymandering (redistricting should be done by statisticians and land surveyors not politicians )
Also, 'dinner' at 10.00 am and 1977 calories a day seems to imply squeezing all of those 1977 calories into a single breakfast-lunch-dinner single meal. Surely not that healthy. Also AFAIK intermittent fasting works best with a mixed schedule rather than same schedule every day, and to really gain maximum benefit it's better to eat normally some days and completely fast on others (24-36 hours or more).
Our hunter-gatherer and cavemen ancestors evolved while occasionally going a few days without food, and certain body-repair mechanisms (autophagy, gut repair) kick into high gear after many hours without food.
Surely mixing that with a lifestyle more beneficial to mental health would work better overall?
"I have nothing against vegetarian diet, as long as it doesn't try to hide in meat look alike"
Absolutely this. Nothing wrong with vegetarian or vegan dishes, many of them are extremely tasty besides very nutritious. No reason to screw it all up to pretend it's not meat, and fail on both counts!
"For some of us those "exceptional cases" are the norm. "
Yes, but also, very few people have such driving patterns, so it is an exceptional case. In your case and that of those with similair driving needs, a hybrid would indeed be the way to go.
(And, as an aside, while a modern diesel or eventual hybrid could get 700km off a 5 min refuel, the human driving it will certainly need more than 5 minutes rest every 700km!! )
"The answer to slow charging is fast charging"
Absolutely this. To be honest most other current battery specs (energy density, range, charge/discharge cycles), while improvable, are already good enough or more than good enough. Being able to charge at 200kW (which is around the current record for consistent, rather than peak, charging rate) gives, for most electric cars, an extra 100km range per 5 minutes charge*. If you could manage to double that in the typical (rather than record-holding) electric car, adding 400km range in 10 minutes and you're instantly almost-on-par with ICEs even for those exceptional cases where you're travelling more than the car's full range in a day.
*if their efficiency is 167 Wh/km, which is current real-life efficiency of a mid-sized electric saloon
"The current battery technology in EVs is very poor, and in a decade or so we'll look back and laugh at it."
Well, yes, but also no.
If you look at the old NiCd dumbphone batteries , or the Li-ion batteries from the first smartphones and electric cars, the current generation of Lithium batteries are very advanced indeed. Lithium battery fires used to be a much higher risk than now (cough, Samsung, cough), and it's possible to alter the chemistry to have a safer (albeit less performant) Li battery. Energy densities in the mainstream top-end eg Tesla are past 250Wh/kg. So I wouldn't really say current battery technology is "very poor".
Equally, though, there is quite significant room for improvement, with theoretically proposed battery chemistries delivering up to 1.5kWh/kg and many engineers confident that we can eventually crack 1kWh/kg in practice. And new technologies (Lithium Sulphur and Aluminium-anode based among others ) promise higher energy densities, faster charging, longer life (charge-discharge cycles) and better safety.
Honestly I think a decade on is still too close for a major breakthrough on all fronts, but equally honestly, I don't think we need a major breakthrough. If in the next 10 years we make the same steady progress as the last 10 (7% year-on-year increase is a doubling in 10 years).
- Doubling best energy density to 500Wh/kg means a 100kWh battery that weighs 1/2 tonne (light enough to bring electric cars in the same weight range as ICE)
- Doubling best charging rates to 350kW + means a 10-80% charge on such battery in 12 min (which would give 350+ km real-world range)
- Doubling charge-discharge cycles to 2000 (even assuming just 10-80% every time) means the battery will still work pretty well after 700,000 km. In real life it probably means you keep close to 100% capacity for at least the first 200,000 km
So, yeah, looking forward to those amazing batteries we might have in 10 years (if we manage) but the ones we have now are certainly no laughing stocks!!
If you mean standardised across the whole industry, that doesn't make much sense to me, it's like saying that ICE vehicles would have had a big boost if everyone standardised on the same engine and gearbox. ICE engines were highly proprietary, and the bragging rights involved in having the most powerful (and later, most economical) engine drove development forward. What was standardised on (or at least settled on a 3-5 possible standard options) was all the gubbins around it - pipework, electrics, wheels and brakes etc.
With EVs there is clearly value for a manufacturer to have their own internal standards, or shared across partners (Hence VW-group electric cars sharing platforms, same with Kia-Hyundai partnership etc). But this has also been standard industry practice for ICE vehicles as well for a couple of decades (VW group again, and Renault-Nissan come to mind). But when it comes to battery packs, charging, and motors, these are the technology 'crown jewels' of a car company, they might be shared across group partners at most but are otherwise closely guarded.
"even if we quit using plastic tomorrow, this stuff is likely to be with us at least through the lifetime of anyone alive now including those born five minutes ago "
Far longer than that actually. Plastic takes hundreds of years to break down, and it actually simply breaks down into smaller and smaller particles, hence the proliferation of micro and nano particles. There are some interesting prospects of developing certain types of bacteria that feed on plastic, but the only way to be free of plastic waste is if such bacteria were ubiquitous (and that in turn means the lifetime even of in-use plastics would be severely limited)
I also avoid water in plastic bottles, I drink filtered tap water.
The issue is though, that as per the article, a large part of the microplastics found isn't coming from the bottle itself. Bottled water in a glass container (and tap water too for that matter) is sourced from a spring / underground aquefier and filtered, or derived from reverse osmosis. Either way there are plastics involved in the processing machinery. I wouldn't be surprised if plastics are being picked up from the pipes that deliver tap water.
However since I don't have a private pristine brook of snow-melt water in an unpolluted corner on the slopes of Mount Inaccessible, I'll just have to make do with avoiding plastic bottles.
"In order to understand that it was merely identifying the angles and starting points of the ridges, they had to study the AI system's decision process. Thus, the team concluded that the AI was using an unexpected forensic marker."
What is interesting to me is that the design of their pattern-matching system (not AI, but I get that real researchers sadly need to use this term to get any funding) allowed them to identify WHY the results being given were being given, which I believe is (by design or because of huge complexity) not possible in the more complex LLMs.
It IS very surprising to find that part of a single individuals' prints could be identical across fingers.... not sure exactly what the forensic utility is though, since normally you want to use fingerprints to distinguish between different *people* not different *fingers*. I suspect that rather like DNA, forensic comparison uses a simplified mathematical model of the fingerprint or DNA to be able to more easily compare them, and model matches might not necessarily be aligned with real-life matches. So in any case it's good that some "thought-to-be-correct-but-not-really-proven" assumptions such as unique fingerprints are being challenged.
In the case of ridesharing companies, is a driver allowed eg by Uber to also drive for Lyft or any other company, and vice-versa?
Can a driver individually negotiate a rate with the ridesharing company?
Indeed, can a driver even communicate with a ridesharing company to even try to negotiate anything at all, or can they only interact with a web of circular, self-referential web pages and 'help' lines that lead nowhere?
Can a driver independently record what they are owed (including knowing what, if any, tips has the customer added to the fare), or transparently access the data ,and can they bill the rideshare company and/or successfully contest a disputed payment amount? Or is it just 'take it or leave it'?
I don't know the answers to all these, nor do any one of them individually determine contractor vs employee. But that's exactly why the decades-old standard looks at those 6 separate questions, the combined answers to which determine the classification.
"OpenAI has said it would be "impossible" to build top-tier neural networks that meet today's needs without using people's copyrighted work."
OK, so either *don't* build a "top-tier neural networks that meet today's needs" whatever the eff that is
OR
Pay people for their copyrighted work to be incorporated in the training set
Simply hoovering up all the data and leave the lawyers to argue about it for the next 2 decades is just plain wrong
"Fauci uses the word immunity."
Not only that, he says "I wouldn't be surprised if it gives lifelong immunity".
So, OK maybe there were other sources saying it might be yearly like the flu one, but this was supposedly the single most prominent health authority in the world at the peak on the pandemic, and he's saying *lifelong immunity* to answer a direct question asking if it would need to be taken yearly. It's not something misunderstood or taken out of context.
No misunderstanding at all, the way it was initially marketed was:
"1 or 2 jabs (depending on variant) and you're done"
"even if you're not a high-risk group you have to take it because 'herd immunity' "
"This has been very highly tested and we know it's safe and effective"
The reality turned out to be:
"this only lasts 6 months tops and you'll have to keep taking it again"
"actually, vaccinated people can spread Covid just as well as unvaccinated ones, they just won't have symptoms, so herd immunity is off the cards anyway"
"We tested this over 3 months on a group of a few thousand, we sort of think it's just about safe enough and effective enough to make a difference".
To be fair, in the middle of what was a giant emergency, the vaccines worked 'well enough' for enough people to bring the pandemic under control, but let's not pretend that they were even close to what they were marketed as at launch
If it's to supply on-street parking for residents to charge overnight, 7kW is quite sufficient for an overnight full charge. Also for residential areas, it's far more likely to be used by people travelling shorter distances. If someone has some errands and is parked for an hour, a 7kWh is enough for around 40km, which for many people will be as much as they are using in a day anyway.
Of course it would be ideal if there is an infrastructure capable of delivering 50kW+ to every parking spot on a street, and we might eventually get there in 2074, but you have to start somewhere!! As long as it is actually capable of delivering 7kW (noting caveats that have been raised over how much power the cabinet has available anyway!!)
Nice idea BT but doomed to fail unless you can solve the cabling/parking
Cabling should be quite easy...unless there are other cables running underneath and along the pavement, it would surely be trivial to dig a trench across the pavement, and put in a metal pipe with a flat liftable cover on top. Anyone needing to charge can just run their cable through the trench.
If a cabinet is at a corner, they could use a similair trenching system to extend a cable from the cabinet along the road until it reaches a legal parking spot.
I think a bigger issue is, if the cabinet power supply is built for supplying broadband switches, that's surely far less than 7kW, so even if they provide a Type 2 7kW plug, how much power are they actually going to be able to supply??
I remember a university assignment [redacted] years ago to implement the same sorting algorithm in C and in assembler, to compare the relative speeds. The C code ran in a few seconds. The assembler seemed to me to be already done as soon as I had pressed 'Enter' on the command line invoking it.
(of course it's not to be taken lightly, and there are indeed many things that can go awfully wrong coding in assembly.)
"People by themselves tend to be selfish, greedy and have too many children."
Not strictly true for the children part. All demographic information we have points to people having less children as the country become more prosperous, because of a variety of factors:
- in very agricultural / manual labour societies, a baby is a resource who can start doing chores from quite young and productive work for the family from their teens. In countries where better opportunities are available through education (and where child labour is illegal), having children is a financial cost.
- in countries with little social security, more children = more people who can take care of you in old age. With good social security, this is less of a factor
- in countries with crap healthcare, having more babies is a hedge against some of them dying
Birth rates have plummeted in richer nations to less than replacement (hence labour shortage and pension crises), while even in poorer nations in Asia and Africa it's shrunk dramatically the last 20 years. In fact the only groups in western nations with high reproductive rates are those who are deliberately promoting it as a way to 'win' a demographic battle eg ultra-orthodox Jews and settlers in Israel
In any case I absolutely agree that unnecessary consumption is what needs to be scaled back, and those people making a living through that have to recalibrate to a different job
"how much the Roman Empire's need for food contributed to desertificaton in N.Africa."
And how the requirement of wood for boats starting in the ancient Greek era, combined with fields for agriculture resulted in much of Europe being turned from forest to farmland
You have to combine both things
1 - return farmlands that were former forests back into forests. That is a one-off CO2 capture since as you say, once the forests is grown the CO2 in/out stabilises
2 - use marginal farmland to grow hardy quick-growing plants which are harvested and dumped down a mine (maybe in a few million years they would have turned back into coal!!!). Hemp for example is very hardy and extremely fast-growing. The only issue there is how long you can do that on the same patch of ground without depleting the soil (or having to use fertilizers which take a lot of energy and CO2 emissions to make)
3 - Return to using wood as a primary material for building and many other things, and have plantations of suitable trees, in whose structure the carbon would be captured. This is a medium-term solution though as eventually the structures need to be renewed and there's only so much you can build with wood
But basically, if you want to capture large amounts of CO2, using plants to do it is a no-brainer as that's what they do naturally
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
