Posts by mpi 511 publicly visible posts • joined 15 Nov 2021
Additionally, the companies peddling this consumer AI revolution are kinda meeting their worst opposition: Themselves.
If I need an LLMs capabilities, be it for my own projects or as part of a product, I can just shell out a few bucks and have full access to GPT-4o or whathaveyou, all of which are more accurate and powerful than the models I can run on even the most sophisticated "AI laptop" for much of the foreseeable future.
And whos positioned to be the prime player in this market of "AI-by-API"? Well, Microsoft of course.
So, if I'm a company, why would I shell out an ungodly amount of money to replace my laptop fleet with "AI laptops", when instead I can keep the hardware I already have for soome time and then replace it with the next generation of cheap-but-adequate, and just get AI capabilities by subscription?
So the next generations do NOT in fact automagically turn into "digital natives" with digital superpowers as was foretold? What an amazing surprise! What's that? Instead many of them struggle to understand even the most basic concepts, like file systems or programs, because they only ever got to experience the "digital world" through the confines of smartsomething devices that were designed to maximise the output of an attention economy?
Well color me shocked. Flabbergasted even.
And what's that? Curricula designed after computer SCIENCE (which is, in fact, a highly specialized scientific discipline) are NOT a good way to teach the general populace about how computers should be used? Amazing! Who would've thought that? Oh, btw. did you know that making people study Oceanography doesn't, in fact, get you more people who can swim? I know, right, it's almost as if there is a huge difference between these topics or something.
> https://pm.besharp.at/jira/browse/CC-6550
A political donation from years past and a crypto integration in Brave that takes all of a a couple of mouse clicks to completely de-activate?
That's the sum total of criticism against Brave Browser, which is, hands down, best in class when it comes to ad blocking?
> and its not inefficient.
Yes it is, for a very simple reason: Physics.
It doesn't matter how you wirelessly transmit energy, the losses will ALWAYS be a lot higher than letting electrons bumb together in a piece of metal. Again, that isn't me saying that, it's physics saying that.
So you either use A LOT more energy to reach the same charging speeds, or you charge A LOT slower. Both options suck for widespread adoption. The former will make already struggling electrical grids even less capable of coping, the latter is simply unacceptable for consumers. Keep in mind: Any charging technology is not just up against other charging technologies...it also is up against gas pumps, which put a car from zero to full energy within less than a minute.
> The red line is where the IP owner is suppressing progress
That existence of patent trolls already suppresses progress.
Because their existence creates a chilling effect on inventors who aren't multinational companies with deep pockets and highly paid legal teams.
...'opt in is only meaningful if users can make an informed decision,"'
If I cannot make an informed decision about something, most of the time, the correct assumption is that I don't want or should buy/activate/use the "something" in question.
For example, if I cannot make informed decisions about how the stock market works, I don't buy stock. If I cannot make an informed decision about how to handle a powerful Motorcycle, I probably shouldn't try to handle a beast that can go 250 km/h.
So yeah. No. Luckily, there is no shortage of privacy enabled chromium based browsers these days.
> t's not the right thing within a single project which progress with social norms.
Yes it is the right language within a project run by real humans, who have real time limits.
> In any project there is a degree of obligation between upstream and downstream code
If up/downstream code were agreed upon by BOTH parties; yes. This is not the case here. Here you have a bunch of people showing up one day, putting their new stuff into the project, and expecting everyone else to just accommodate it.
> The Rust people are building safe abstractions to large parts of the kernel.
It remains to be seen if any of the alleged benefits of Rust will materialize for the Linux kernel.
> But what I don't understand very well is why folks whose needs are simpler far insist on using git.
Because it is neither "complicated" nor "obtuse" when used for simple use cases.
`git` is the perfect example of a well executed tool: Its handling is simple for the simple usecases, and only gets more complex as the the usecase itself grows in complexity.
Example: Single developer: All you realistically need are git add/commit/push. If you want to be a bit fore fancy, git branch/checkout/merge. That's a handful of commands, no need for any of their special bells and whistles, that cover the vast majority of source control requirements already.
> the Rust people have stated many times that if anyone needs to make changes to interfaces impacting Rust bits they're happy to be involved
Yeah, and that's exactly where the friction is coming from. Because, no, it's not enough that they would be "happy to be involved."
THEY want THEIR language in there. They are the ones making the demand. So they are the ones WHO ARE REQUIRED to make it work, and they are the ones WHO ARE REQUIRED TO MAINTAIN IT. No one else has to deal with their problems, and no one else has to spend time and resources to solve their problems for them.
Aka., if an interface changes in a way that impacts rust code, the rust developers are the ones who need to make sure THEIR code works with that, not the other way around.
This isn't complicated, or an outrageous point of view: If someone develops a plugin for my application, and later I change some part of the API, it is THEIR responsibility to update their plugin to deal with that. I am not required to maintain their plugin for them.
I will give them deprecation warnings. I will use semver-semantics correctly. I will even manage a mailing list for plugin developers, telling them in advance "hey guys, fair warning, you use X and X is going to change in update Y, so you might wanna update, or your shit is gonna break."
But if then demand that I put in extra hours and free work to maintain THEIR project for them, then I'm afraid all the're gonna get from me is this.
> Yes, the physical products that have to constantly change include: CPUs. And GPUs.
Not in the same way software changes.
The equivalent would be to redesign half the ALU core while the assembly line for the new CPU is 75% built, and we decided that our word size is 70 bits instead of 64, because big number sells easier. Oh, and the whole thing now has to be triangle-shaped, because marketing thinks that looks more "zazzy!". And before I forget, Janet from sales read this thing about "ternary logic" somewhere on facebook, and we had a C-level meeting, and we really want that in the product as well, so make it happen, alright?
> They are announced on as little as an 18 month cadence.
Cute. Meanwhile, software projects sometimes change core requirements single-digit DAYS before launch. And in some cases, AFTER the launch.
Oh, and btw. those 18 month cadence are the result of an incredibly complex planning and logistics process that starts YEARS before the announcements go out, including multiple R&D teams working in the next few generations of chips while they still anounce this years.
> and has a pretty good track record of backward compatibility
Python has a pretty good track record, true.
Python packages are a very, veeery different matter.
The amount of time I wasted fixing old, or even abandoned, python packages, where the maintainer thought it would be a swell idea to pin one specific version of a dependency as a requirement, even though nothing in the package actually relies on that specific version of that dependency.
And those are still the easiest to fix, usually I just need to update an old pyproject.toml or setup.py to make pip stop complaining.
It gets alot worse, when suddenly a package relies on some C-libraries, and there actually is a reason that dependency was pinned, so I have to install an old version of the package, which requires an old version of the lib to work, but the system no longer has that lib, because its 10000 years old...
People should absolutely be able to eat hamburgers.
People should absolutely not be able to buy a 10-pack of hamburger patties, then forget about it, re-discover it 3 weeks later as a pile of rotten meat at the back of their fridge, and throw it away without a care in the world, because this incredibly energy and resource intensive produce is so dirt cheap, and there is such a massive overproduction completely beyond any demand, that they don't even really have to care about it. Not to mention that part of the reason it is that cheap, is because we all subsidize the industries behind it with billions of tax payer money, that could otherwise go into education, healthcare, etc., and/or be used to improve public utilities such as better fibreoptics infrastructure.
And as long as that is the case, DataCenters are pretty far down the food-chain *ba-dum-tsss* when it comes to the question: "Where could we spend the worlds resources more efficiently...?"
If you want to "prioritize the world's diminishing resources for something more useful", how about
- banning excessive meat production and agriculture
- getting rid of individual traffic where it isn't absolutely necessary
- completely banning private jets and all commercial flights that can't meet certain passenger thresholds
- renaturation of sprawling suburbs and other inefficient housing solutions, replacing them with densification and urbanization
- re-prioritizing resources meant for more road-infrastructure into public transport solutions and sustainable energy technology research (yes, that includes fusion)
Because all these things damage our habitat ALOT more than datacenters ever could.
And if any of those are not an option, then I guess datacenters are non-negotiable as well. Because I really don't see why it should be considered okay to burn 25kg of feed (most of which is legumes fit for human consumption btw.) and waste 50,000 liters of water, to produce 1kg of beef so people can eat hamburgers 7 days a week if they want, but datacenters, which use ~1.3% of global electricity output (2022 numbers), are somehow a huge problem that needs to be stopped.
What? Of course it was interesting!
It was wild, wacky, full of new ideas, and a blast to use. People were making things up on the fly. There is a reason we called it "Web-Surfing" back then, because it was that exciting to interact with.
You know what's decidedly NOT interesting? The sterile landscape of run-of-the-mill, SEO-infused, ipad-interface-wannabe, corporate-feelgood-speech-bulls__t, empty, senseless and soulless garbage that is most of todays internet. The average webpage today is about as surprising as drywall. They look the same, they make the same mistakes, they use the same user tracking mechanisms. And most of it exists purely as a canvas to serve ads. What's really heartbreaking is that, despite the sameness and pre-fabricated looks (which in any other industry would indicate higher standards), they actually got WORSE in terms of usability, information density, and general usefulness. And they, of course, agree that it's perfectly sensible to load 30-40 MiB of data, to serve 10 lines of text under a giant "hero picture" that's usually about as heroic as a Fainting Goat.
The dotcom bubble didn't burst because the web was "not interesting". The dotcom bubble burst because stock markets doing stock market things. That had nothing to do with the capabilities of the technology.
> Currently, commercially speaking, worse than an FAQ section
Commercially speaking, people don't read FAQs, and customers expect a call-center option, like a chatbot.
Call Centers however, when done well (well educated and trained personnel) cost a lot of money, or when done badly (poor people, exploited to (sometimes forced) labor in some local strongmans warehouse somewhere, barely fluent in the customer language, reading stuff from a list) may provide worse service than even a purely algorithmic non-AI chatbot.
Plus, when people argue against chatbots, I always get the impression that they are using the worst examples in the industry. I have seen, and used, some truly amazing pieces of technology out there, which were a blast to interact with. I rather have one of those things, that solves my problem in 20 seconds, than wait on hold for some guy I had to read my order-number to 4 times, because he's wearing a headset from the early 2000s, that 42 people wore before him, while constantly wondering if the guy I am speaking to is even allowed a toilet break during his shift, and whether he has to sift through pictures of hate-crimes on his second screen to do "moderation" for some "social" media company while talking to me.
> worse then humans in every way
Except one: Speed. And when the documents in question are not mission critical, sometimes that's the most important thing. Because our society produces a lot of documents, more than humans can even glance at. Sometimes "good enough" is actually "good enough". And the number of use cases where this is true, is pretty astounding.
> jobs that LLMs are -or ever will be- capable of doing
The automation aspect is oversold. Lets be clear, yes I agree we ARE in a bubble, and some VCs will be very disappointed when it bursts. Boohoo.
What LLMs excel at, is not completely automating jobs, but act as a force multiplier for the people doing it. Best example: Test writing. I give it the class and an example test, and describe the edge cases I want tested in a few words, click a button, and off it goes. Do I have to check if what it wrote makes sense? Absolutely. Does it get these simple, but onerous and time consuming tasks right most of the time? Yup. Is that an advantage to me, because I get to spend that time more producitively? Yes.
Strongly disagree.
A 64 utf-8 character random stream is an amazing authenticator. Lightweight, easy to implement, and completely safe against against anything short of bruteforcing the entire keyspace or using the 5$-wrench-attack.
And all it requires is for people to use a password manager.
Because none of this, not space stations, not moon bases, will make any meaningful contribution to science and technology. It's purely a flex, nothing more, and sorry no sorry, but that's a bit a a steep price-tag for flexing, when there are so so so very many better things to spend money on.
As for my first sentence, I realize that this is somewhat of a bummer for space exploration fans, but it's true. Space Stations and Moon Bases sound cool, but aren't. Why? Because they all share the same fundamental design-flaw: They waste the majority of resources on keeping a bunch of, comparative to the alternative, very fragile biological components alive (aka. Humans), the job of which could be done at a fraction of the cost using robots.
Robotic drones don't require water and food or breathable air. I don't have to worry about getting rid of their waste products. They don't lose bone density or suffer circulation problems in microgravity. They don't suffer nasty unexpected health surprises (google "herpes virus reactivation in astronauts" sometimes). They have no need for bulky spacesuits eating into precious launch-capacity. They don't need a relieve crew every few months. And when the mission is over, I can just leave them there and sell T-Shirts with their silhouette instead of worrying about how to get them back home, preferably in one piece and alive (which in itself multiplies all the engineering hassles and resource requirements). And if disaster strikes, it is embarrassing, and blows up a ton of cash, but no one is going to have to set any flags to half mast, and see entire space programs shut down in the fallout.
"But but but ... Apollo?!? They had people in those rockets?!"
They sure did. Because there was no alternative. Computers and robotics in the 60/70s were nowhere near as advanced as they are today. But today we manufacture electronic components so tiny, chip designers have to worry about qantum-mechanics, and the thing I carry in my pocket to watch cat pictures, has likely more memory and computing power than all the computers in the world back then combined.
And technology gets better every year. The stuff we can build today would seem like science fiction less than a decade past. Just think what kind of robots we will have 10-20 years from now.
Humans in 20 years, on the other hand, will be just as frail, vulnerable and needing bathroom breaks and oxygen as they do today.
So, for the forseeable future, if you are excited about space exploration, you should be excited about robotics, not people in spacesuits.
> You fool around with the data and they can unequivocally trace the source of the modifications.
No, they cannot.
Because, without a consensus mechanism, any such trace is worth squat.
Let's say I have a blockchain, and share that with 100000 people. They all have the same chain. Okay. So now someone (who could be me), changes his copy of the chain, and tells everyone: "Hey, I found a discrepancy! The chain y'all have is invalid!"
Now, how is everyone going to check that I am wrong (or right)? Maybe I am trying to trick them now. Or maybe the copy they got in the first place was bad to begin with. They have ZERO way of deciding which of the two applies. And that doesn't just apply when I "revise" the chain, it also applies every time I add another block to it. How can they verify that the new block is valid?
That's why blockchains have this thing called "consensus mechanism"...a way that makes it really, really, really, REALLY hard to trick everyone into accepting changes to the chain, because in order to facilitate any changes, you have to solve a puzzle, that is really hard to do, but the result of which is really easy to verify for correctness.
Now, the thing about consensus mechanisms: In order for them to work, there HAS TO BE an incentive for all participants in the chain. With shitcoins, this is easy, beacause investing the work/stake required for the mechanism, gives participants a payday. But for a database, there is no incentive.
> this way the system can't be gamed
That only works if the chain is decentralized AND edits require a consensus-mechanism, such as proof-of-work or proof-of-stake.
Blockchains are not intrinsically tamper-proof. If I am the centralized agency that holds the blockchain, nothing prevents me from simply editing an entry at will, and then re-hashing the entire chain to "validate" the new reality.
And at that point, the question stands why it couldn't be just an ordinary database.
, or be able to make a privileged account complete some steps"
Ah, so I need to have full control of the system, in order to undo security updates to get full control of the system that I already have full control over so I can have full control over my full control.
Got it.
> The north pole has plenty of water ice
It sure has. Unfortunately, the north pole is about the furthest point away from any point where a settlement (which presumably wants to use solar power) would be established. And what with the astounding lack of overnight delivery services on the red planet, there is pretty much zero chance in hell to transport that water down south for processing.
Oh, sure, one could, in theory, establish a base in the north pole. Only, now all those solar panels (which are transported there how exactly btw. considering we will need housands of metric tons in panels alone, not to mention scaffolding, cabling, inverters, storage batteries, ...) which were running at half-capacity to begin with (inverse square law and all that), are even more useless than before, because now we don't even get sunlight at a good angle.
> Scientists and engineers have long studied In-Situ Resource Utilization (ISRU) as a way of building a settlement on the Moon without having to transport all the material from Earth.
Oh sure, you wouldn't have to transport the materials there, true enough.
You just have to transport the machinery, energy production facilities, tools, and additive materials required to use the local resources. And the labour to use them, whether automated or biological. And of course also the machinery, labour, tools, etc. to then use whatever the aforementioned machines produce, provided they don't break down. Oh, that reminds me, you also need replacement parts for all these thingamabobs, and a way to keep the crew alive long enough until whatever survival-critical things are to be provided by ISRU actually can be provided by it.
Kinda reminds me of the old "How to draw an owl" tutorial:
Because that hunk of junk would be one crash away from turning into a cloud of shrapnel, circulating the Earth at ~8 km/sec, possibly for centuries, ready to cut everything it encounters, including sattelites, future rockets and their crew, into space-confetti.
As long as its in one piece, its relatively easy to get rid off.
The order is a couple of items from a finite list, and the customer is physically present.
So, what is easier: Having an ASR system try and understand random peoples input, against a background noise of running motors and honking horns, and THEN have a LLM or similar system figure out their intents (including corrections, repetition, the occasional sarcastic remark, etc.) and translate that into an order...
...ooooor...
...simply have an outdoor-rated tough touch-screen (like the ones used on ATMs) positioned to that it can be reached from the drivers seat through the side window?
If you want it to be extra fancy, have a Convolutional Neural Net powered visual system (see, now its also an AI system!) detect the height of the car and move the screen up and down dynamically.
Why? The concept has already been successfully tested by NASA, see linked articles elsewhere in the comments. Note, that's "tested" as in, actually hauled into orbit and experienced an actual, real-life reentry, not just "tested" as in "worked in the lab".
Just because something is inflateable, doesn't mean it cannot be made of materials that can withstand extreme temperatures.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
