Posts by that one in the corner 5065 publicly visible posts • joined 9 Nov 2021
The choice of Ubuntu seems more like a brand-recognition move on the part of DeepComputing, coupled with a "we are cutting edge" from the Ubuntu side.
Once it is out, it'll soon become clear which other distros work - or are moving to work - with the machine (hmmmmm, Devuan, aaaaargh!).
But first, get it into the market with a recognised name attached (and that ain't gonna be Red Hat!)
PS
> If RISC becomes Popular x Chinese, US sanctions will kill it. That's what they are designed to do
Assuming you meant RISC-V and not just any old RISC (or even used-to-be-RISC but seems to have grown a bit over the years) then - you *do* know that it isn't a China-only thing? That it is headquartered in Europe and that the US was recently being ridiculed because they were trying to keep all the RISC-V stuff within their own borders?
The idea that the US sanctions are *designed* to kill off RISC-V is - shall we say, far-fetched? The US would love to have loads and loads of the very bestest RISC-V In The World, just, preferably, only in their bit of the world.
> Seeing AI and the Pi associated is particularly depressing
As you know, the AI on the R'Pi is just an M.2 stick that slides into a generic M.2 carrier HAT; the same sort of HAT onto which you can stick in a *non* AI-oriented M.2 stick.
Given this - that both the HAT and the stick are entirely optional and in no way built in to the Pi 5, should we assume that you have been becoming increasingly more depressed over the years as other, equally entirely optional, HATs have been released that don't happen to match your particular interests? Oh, how you must have shuddered as motor driver HATs appeared, or the sheer absurdity of the Space HAT, as you so desperately waited to do no more than replace your PC.
My bestest laptop runs a 6th gen i5 with a massive 2 cores at 1.9 GHz, with a staggering 8GB of RAM.
And it more than fulfills my needs for a laptop, nicely complementing in portability (under 13 inch screen, btw) the desktop and servers that do any heavy lifting and bulk storage that might be needed.
If you rely on just a laptop to drag around all the time, *this* RISC-V box might not be the one for you, but then there are plenty of other laptops that you'll view with disdain. If you are champing at the bit for a RISC-V laptop then this does show that they are progressing; won't be too long now...
You are over-paying for your mouse-jigglers (especially if you buy, say, a pack of a dozen). An ATTiny 85 on a Digispark board should do the trick.
I really wish that I had thought to plug one in on Weds evening, as it would have prevented Windows from rebooting the PC and knackering some very slow and long downloads, as well as the rest of the overnight jobs I had left running!
So don't go around making it sound like mouse-jigglers are only useful for nefarious purposes.
Blank screen?
You were lucky! We 'ad to cut out 'ole from old piece o'cardboard and hang it on t'pit wall if *we* wanted a blank screen.
In Sundays, management would roll us up longwise and we 'ad to play t'role o'pipes for 'em, rolling down slag heap so as to seem 3D. Though, to be fair, they did warm up teapot in Winter, to stop it sticking and rolling down t'hill before they'd had a cup.
Curses, my 19th century reference has been beaten by a 16th century one (but both during the reign of queens).
Hmm, there must be a relevant quip from Boudica's court; Tacitus was always good for a pithy line, that bit about Veranius harrying the Silurians always gets a chuckle...
He had bought a large map representing the sea, Without the least vestige of land: And the crew were much pleased when they found it to be A map they could all understand.
"What's the good of Mercator's North Poles and Equators, Tropics, Zones, and Meridian Lines?" So the Bellman would cry: and the crew would reply "They are merely conventional signs!
"Other maps are such shapes, with their islands and capes! But we've got our brave Captain to thank (So the crew would protest) "that he's bought us the best— A perfect and absolute blank!"
(With apologies for the formatting)
> ... it's missing a critical piece of information:
Is even 0.001 TOPS vaguely relevant to your daily workload of running Word and PowerPoint?
(Wanna bet someone else is going to ask the same question again soon, in their own words, of course?)
> more than likely we're going to have a handful of models running on our systems at any given time
What? Who? Why? TFA is certainly in favour of running local models (although, it does seem rather fixated on LLMs - there are other jobs these things can for for us, you know; running BirdNet with greater ability to separate multiple simultaneous calls would be nice).
It doesn't help that the quote in TFA is potentially ambiguous / being misinterpreted:
> "intensified feelings of anxiety and stress for workers with higher level of neuroticism, with a potential lack of preparedness for the WFH environment exacerbating emotional instability."
That *doesn't* state "... stress for workers LEADING TO higher level of neuroticism" (which is how TFA and comments here seem to taking it) but says workers who already are a tad more neurotic than the rest (and we are *all* a bit neurotic about something) are having a worse time than the rest.
Which also sounds like stating the bleeding obvious, but is the sort of thing (mild variation in personality leading to greater negative responses) that gets glossed over.
> So it took two business schools working together to find this out?
It is quite a step forward for a business school, let alone two! After all, how many MBAs does it take to change a light bulb?[1]
[1] I don't know, that wasn't covered in the course; I'll have to call in a consultant.
Tell me know, I want to know, uh ha.
This article:
> Just last month, research indicated that HR folk are worried that return to office mandates are leading to higher staff attrition rates
>> a quarter of executives and a fifth of HR professionals hoped RTO mandates would result in staff leaving
> I don't mind sprints ... our local user group holds a couple every year, also called "hackathons"
Doing a single hackathon can be fun - like any event where you push yourself to get it done, the old neurochemicals flow and you get a rush.
Then you go home at the end of the conference and relax.
Unfortunately, you've fallen into the same trap that some Agilistas do (and which is encouraged by the very use of a word like "sprint") and confused the "hackathon" ethic with the "we are going to do the same thing next week, the week after, the week after..." of a work project.
> set expectations for such focussed sessions
Like, if you plan for every "sprint" to be that focussed[1] then you're going to burn out, or, more optimistically[2] everyone will get sick and tired of the word "sprint" and just mutter dispiritedly evey time it is mentioned.
FWIW, Good(ish) Agile also says "go home at 17:30 and forget about work" to avoid this, but...
[1] yah see, "sprints" are supposed to be focussed on which *bit* of the task is to be done; if a distraction comes up ("this lib is broken") don't fix it now, log it on the kamban, stay on task and, because we have a review within a week, that issue won't be forgotten. But "focussed wrt domain" gets muddled with the usage you just gave, where the *person* is focussed on the hack, to the exclusion of all else (hence the need to plan rest breaks, even loo stops!)
[2] from the pov of the devs: dispirited is better than burnt out; curiously enough, in the devs eyes, we aren't fungible!
Quick, write that down (oh, you have - good start); now, invent a few terms (better yet, re-use some common words) for each stage, train up some consultants and - PROFIT!
Or just get on with delivering working systems to happy customers, if that is the sort of thing that floats your boat, weirdo.
> Thankfully we have an old, non-bricked driver; just tell the machine to uninstall the "current" driver, then install the old one
Nope, you've got the FTDI driver version just *after* the bricking one!
The later driver refuses to work with the adapter, the one I'm referring to actually wrote to registers in the adapter's chip which plain stopped said chip working: the USB id was set to 0 and it wasn't recognised as a serial device by Device Manager, so even when you rolled back the driver it still didn't work.
Until you found a (non-virus-loaded) copy of a utility that would rewrite the USB id *and* rolled back the driver so it wouldn't just happen again. Which was fine and dandy after the news got around (ta, Register) but before then it was time to run in circles, panicking and bricking adapter after adapter...
The only kit I've carried away from work came after redundancies:
* an old kvm and cables, which had been chucked out of the server room years before it even reached me
And from the job just before that
* a mousemat with the company logo on it
As for walking off with a laptop: I supplied my own laptop, 'cos it was old enough to still have a serial port to plug into embedded kit (and happily watched everyone else walk back to their desk to find yet another driver for yet another random brand of USB serial cable - oh, and you remember when FTDI bricked anything it thought was a fake? Quick, to PC World - what driver does this *one* need?)
I worry that they be learning, asking
> "Don't you know how important I am?"
instead of "Do you know who I am?", so we can't give a practised reply[1] of " No, but if you've forgotten again there is a reminder in your wallet".
[1] best used[2] only in those situations where you know they'll respond with "What is your name? I'll have you fired!" - keep a few random cards from other people on you; or simply walk away, they only have a short memory span.
[2] do not take potentially career-changing advice from a random posting on a website!
> The challenge is adapting it for the customer relations team
Challenge - accepted!
The trick is to come up with novel variants of the old "id-ten-t" and "pebcak", one per customer and get 'em to quote it back to you as their customer id.
"Hi, yes, last week you guys said I was a Flash[1] customer."
I shall pass the baton on to the rest of you lot, confident that you can come up with a plentiful supply of new and useful phrases to suit the expected customer base.
[1] Flatulent, Longwinded And Shiny Headed
> Tape deteriorates...
Didn't we go over this recently? Oh yes: Tape is so dead, 152.9 EB of LTO media shipped last year
The conclusion there seems to be that *every* form of digital storage has a limited lifespan and if you want to keep it you are on an endless treadmill of copying to new devices as the years go by. Even if you choose a robust, long-life, medium (e.g. parchment) you still have to be sure to maintain the scanner that lets you read it back into memory.
And it gets worse: spinning rust can be taken out of the case (Maplins used to sell caddies for this) and will be readable in a few years, but make sure you power up your SSD-based external USB drive more frequently.
> BUT companies shouldn't be thinking about using long term storage, don't let me put you off, for backup it is a Good Idea
Whoops. Let's just correct that line, so it doesn't contradict itself:
BUT companies should be thinking about using long term storage, don't let me put you off, for backup it is a Good Idea
> it's time companies started putting their stuff in long-term storage (i.e. tape, either on-premise or in Cloud providers, I know AWS offers this).
And they charge for it as well, so be sure that what you put there is worth it (just shoving everything up "just in case" will get pricey)
> That'll make it a lot harder for miscreants to nab it.
Slower, that is for sure, but if you have leaky security for credentials that allowed access to live storage you may also have leaked credentials that allowed the miscreants the ability to ask for "this tape to be mounted, please" - and what are the chances people are paying less attention to what is going on with "the dusty old stuff" than the live system?
BUT companies shouldn't be thinking about using long term storage, don't let me put you off, for backup it is a Good Idea.
> Most of the stuff isn't needed online anyway.
Ah, tricky - and pricey, again.
Before relying on long-term storage as your only copy (with a backup tape, of course) you have to be able to *prove* that "this whole directory tree is not going to be needed for X months" (where the cost of live storage is greater than long-term-but-readily-accessible storage). Which means you have to have all your data properly organised - as in, *properly*, not just "I can do a quick file search to find it" - and are absolutely certain that the Javascript sitting here isn't going to be needed on a customer's browser tomorrow ("we stopped generating ages that use that last month, it can go" "Big Name just died, put his pre-written obituary up *now*!" "What do you mean, it can't find its script?").
What you need is a trained Digital Archivist, to help you organise and learn how to determine what to keep live, what to put in the shelf - and what you can just get rid of completely, reducing your storage costs.
Oh, you have thought about all this and are just going to buy another hard drive rack to go into the server room, it is easy, quick and cheap enough, besides who has time to think about these things, we have a newspaper to get out. Ok. COPY!
A "security risk management plan" is a plan about how you are managing the risk before something bad happens.
You appear to be confusing this with an "incident response plan" - which should be part of you overall risk plan, but not the whole of it.
Whether all the players do as they've been asked, and do so usefully, is another matter, but you have mischaracterised what they are being told to do.
As to the overall efficacy of the FCC, you clearly have an opinion.
When the AI bubble bursts and the market is flooded with unsold boards and second-hand from stripped-out AI bitbarns, they are going to be those accelerator-only cards without any video outputs. So you'll be able to buy loads of GPU-style compute cheaply but without it being on an actual graphics card.
So no cheap(ish) upgrades to boost the old FPS for "Lemmings 2".
> The quantum of potential strikers may also be less than 28,000
Solidarity in the work place is a Good Thing, but not being able to move in a smaller quantity than 28,000 or so must be restricting sometimes[1]. But it does make sense of the stories of Asian companies organising worker morning exercises, with nicely co-ordinated jumps at the same level and singing of the company anthem to get pumped to an excited state for laser-like attention during the workday.
Although it does make out of office working efficient; no need for MS Teams when you can rely on spooky action at a distance.
[1] like any job, the group will have its Ups and Downs; you have to hope your colleagues are mostly Charming, with only a few Strange ones. As ever in tech, old hands will be rooting for Truth and Beauty, whilst the new intake are more concerned with who is Top or Bottom of the class.
Probably just the lateness of the hour, but every time I read "gigafactory" I pictured the Gigashadow, unrolling and about to consume all in its path. Then came the reminder about how much power (and, no doubt, other local resources, such as water) this is going to take to run and the connection became clear.
OTOH if that means Zev will be in Memphis, maybe it is time for a visit, cluster lizards be damned.
are simply the sort of people who couldn't successfully complete a project in the first place?
> Or more likely (and this is my usual experience) that people who think they are doing agile, are not really doing agile.
Agreed.
Looking back, I should probably have put in a couple of quote marks and some capitals to express it better (cf my other comments here):
>> Maybe the sort of people who *really* "Get Into Agile"
> The Manifesto is fine. The problem is how the methodology is sometimes applied.
Not helped by companies whose management read *about* the Manifesto and then start the project with a big meeting all about "The Yoyodyne Approach To Agile" which ends being neither one thing nor the other, with roles[1] such as "software architect" turned into positions on the org chart[2].
Not to mention (but I shall) redefining "software architect" from keeping track of all the little bits[3] to "The Guy With The Big Vision"[4]
[1] which can be filled by different people at various times
[2] which are used to set Titles On Business Cards and salaries and "progression in the company"
[3] everyone is using the same library to do the same job, oh, and version n, not (n+1), because n runs the same on all the different bits of kit involved, here are the tests to show that
[4] do not bother me with piffling details, of course you can use XML DOM on a memory-constrained embedded system, SAX has got a silly name!
> the customer has to be central to the whole thing and not an afterthought.
That is *supposed* to be of the basic tenets of Agile, with constant contact with your customer.
Of course, if you ask for a customer contact to attend your Sprint reviews on a weekly (or even monthly, for the arthritic agile) basis, are you going to get their knowledgeable person, who knows their business inside and out, or is she too busy keeping them functioning? So you get the young chap who was hired because "knows about this Agile stuff" but bugger all about the workings of the company that he only joined last month?
Always assuming you actually have *a* customer for this project, and you aren't doing anything silly, like, ooooh, making a system to fit a general market niche, with a dozen paying customers already lined up to get copies configured for their individual situations; 'cos who is a *good* "customer rep" for your Agile Meetings now? Do you rotate them, one a week? Or do you have one of our own guys, who has created a single set of requirements that sort of mixes them all together - but you don't work to those 'cos you are Agilely building it piecemeal and hoping to converge on customers' reality at the end? Even better, do you have an external third-party Agile Expert playing that role, with the benefit that he is equally ignorant of both the customers and your dev team and your previous products whose components may be reusable?
Maybe the sort of people who *really* get into Agile are simply the sort of people who couldn't successfully complete a project in the first place?
Where "successfully" includes "it does still work a year later, after we've left and you have had to actually use it yourself, including following the documentation (!) for the six-monthly reconciliation run that we never actually tested with 100% real data 'cos we never had a stable codebase for that length of time - oh, and when we couldn't remember the name of the config options we obviously just looked at the code, that is what self-documenting means".
> the Pi will never be good for training AIs
Correction: the Pi will never be good at training LLMs.
There are many, many other forms and uses of Neural Nets and other variants of Machine Learning, in all shapes and sizes - from a few matchboxes upwards. Much useful and practical work has been done using ML, especially in Machine Vision, using (what you would probably consider ludicrously bad quality) low resolution analogue cameras in the 70s and 80s - when the mini computers the devs could get their hands on were dwarfed by the processing power in a R'Pi Zero.
Look around and you should be able to find all sorts of data feeds that be used to train a model using just an R'Pi, let alone one with this accelerator connected.
> "While providers of large language models share their own performance benchmarks, these results can be misleading due to overfitting," Prelovac told The Register. "This means the model might be tailored to perform well on specific tests but doesn't always reflect real-world effectiveness."
Yes, very true.
But what does comparing their chess-playing ability do for us?
> "It's somewhat disappointing but expected that these models show no real generalization of intelligence or reasoning," he said. "While they can perform specific tasks well, they don't yet demonstrate a broad, adaptable understanding or problem-solving ability like human intelligence."
Disappointing? Don't yet?
In other words, all this demonstrates is that Prelovac has fallen, hook line and sinker, for the totally unfounded claim that the LLM training process *must* produce a reasoning system and, instead of accepting the results of his own experiment[1] is just a bit sad that the models he has tried don't reason.
> As for why GPT-4o registered a remarkable improvement in chess but still made illegal moves, Prelovac speculated that perhaps its multi-modal training had something to do with it. It's possible part of OpenAI's training data included visuals of chess being played, which could help the AI visualize the board easier than it can with pure text.
So now the LLMs are building internal representations, visualisations, of the boards? Where? Can we extract these visualisations and demonstrate they exist? Nope? So is that anything more than anthropomorphisation in action?
And yet, he *does* seem to know what these programs are doing:
> "Even chess moves are nothing but a series of tokens, like 'e' and '4', and have no grounding in reality," Prelovac said. "They are products of statistical analysis of the training data, upon which the next token is predicted."
In the end, what is the aim of this (other than having a happy and publishing it on Github for others to play with; nothing wrong with that)? Well, the article starts with:
> A new benchmark for large language models (LLMs)
But what is is supposed to be benchmarking? To what purpose? What are we supposed to understand from a benchmark for the (probably) most costly programs in existence when that benchmark can be bettered by an entry to the Obfuscated C competition (at least that only plays legal moves!)?
After all, if he is serious that this is a useful benchmark, he pretty much admits that it is affected by how many games of chess the training has included - so it is trivial to game this by connecting your LLM trainer to the output of a couple of (different) good Chess playing programs (not even setting up a Chess player as an antagonist, cf setting up a GAN).
[1] that it doesn't even imbue an ability to follow the basic rules of Chess, and he proposes no mechanism by which it ever *could*.
> The problem with all AI
Agree with the analysis, but not with the description you start with.
LLMs are *not* "all AI", they are not even "all ML"; they are not even the newest AI techniques and ideas.
The "newness" of LLMs is purely the first 'L', meaning "we can throw more compute at it today".
Yes, I *know* that now "they" have decided to "coin a new phrase, GAI" for a 'proper' AI, and "words change meaning, get over it" - BUT encouraging the misuse of such a new word as "AI" means that you are losing contact with research and work done in only the last few decades[1]; only recently I've heard complaints that a book about AI wasn't solely about LLMs.
[1] "oh, but computing moves so fast, anything a decade old is useless anyway" - of course, that explains why we stopped using lexical analysis in our compilers /s
Referring to a "terminal phase" means you are invoking the Roman deity Terminus to look upon your works at this boundary[1] with good grace.
It is always a Good Idea to Placate The Gods - and Boeing clearly needs all the help it can get from whichever pantheons happen to be around!
[1] the boundary from "being on the pad" to "being (safely) in the air"
> are and will continue to be crap.
Maybe that is why TFA is full of vacuous statements such as:
>> The worldwide market for foldable smartphones grew 49 percent
>> made up 55 percent of foldable smartphone shipments in Q1
>> one of the top sellers in China's foldable market
>> most popular foldable phone in Western Europe during Q1
Not a single useful statement in the whole thing! Without some comparable numbers there is no actual content in this article (e.g. how large is the "foldable market" to the "rigid market"? Was the 49% growth because they sold 149 phones, an increase over the 100 they sold in the previous quarter? Is this sub-market worth $27.50 per annum or $2750000000?)[1]
> There's just no way to make a foldable screen that won't end up with an ugly crease within days to weeks.[2]
Not one single thing in the article actually claimed that any of the phones being discussed are actually useable - they could well all be being returned within a month and swapped for something more robust.
[1] yes, I could dig and find that out - but then if we all did that there would no point in bothering to read an article with this headline (was going to say "an article like this" but ...)
[2] I'm hoping that it turns out we can produce such a beast - or one of the alternative concepts, such as unrolling like a scroll, a much larger bend radius and it looks so much cooler on the shows and films that use the idea than some boring old folded-over design.
> The data is held in unencrypted in SQLite databases in AppData
Damn shame it is impossible to add encryption to an SQLite database - oh, wait a moment:
https://learn.microsoft.com/en-us/dotnet/standard/data/sqlite/encryption?tabs=netcore-cli "SQLite doesn't support encrypting database files by default. Instead, you need to use a modified version of SQLite like SEE, SQLCipher, SQLiteCrypt, or wxSQLite3. This article demonstrates using an unsupported, open-source build of SQLCipher, but the information also applies to other solutions since they generally follow the same pattern."
Perhaps MS devs aren't allowed to "waste their time" reading the MS "learn" website?
> turning things off should not be part of any routine or default customer experience.
So, anybody else here got a list of things that are routinely hunted down and turned off whenever they install Windows?[1] Either by hand or by IT supplying their own ready-to-go image.
Frankly, I'd think turning things off ought to a routine for pretty much everyone - it is daft[2] to believe that every "amazing new feature" is worth the resources to *every* user.
And as for believing *your* every feature is "the future of computing" that everyone ought to use...
[1] actually, any OS - I find that even Linux installs are best followed by a routine of removing stuff that gets added by default, as well as the obvious "install other stuff you do like". And, yes, I do know about creating one's own distro respins: following the routine isn't arduous, just tedious: it isn't a big enough issue to warrant the need for a respin, let alone a Corner Linux distro.
[2] nicest way to frame it; other descriptions, using the word "pathological", are available
> using SQLite so... not ideal for people who like security
Just to point out, using SQLite does not necessarily mean it isn't secure; you can add security and encryption into SQLite, if that is appropriate to your use-case - and you can be bothered to do so[1]. Otherwise, using SQLite is precisely as secure as using Excel, Word, Libre Office or any of the other programs that Recall is effectively spying on.
I note that blog does not say which users can access the database: there is obviously nothing wrong with being able to access the log of your own interactions, it is only an issue if other users can. If it is capturing passwords (and demonstrably so, other than as a row of asterisks in all the screenshots it is taking), that makes it more desirable to keep others out, so access control is a Good Thing.
[1] clearly MS isn't bovvered
> a small screw or component, when dropped onto a floor ... will ... mysteriously warp through hyperspace before coming to rest? ... I’m not saying it’s aliens, but we all know it’s aliens.
Screws are attracted to the helical wormholes by which biros[1] tunnel back to their home planet, where they enjoy biro-oriented stimuli. Interrupted by occasional showers of components, lasting until Tuesday due to high pressure in the vicinity of Polaris.
[1] ref works by DNA
> "However, thanks to HyperAssistant, a concise summary is presented to her, highlighting only the pertinent edits. With this efficiency, she swiftly comprehends the updates and is ready to begin her tasks.
She looks over the code, guided in her expectations by the summary. Everything it mentions is there and looks correct. Even the tricksy bit that the summary notes is perhaps a bit too clever but works, so ok (which she then concentrates on checking now it has been brought to her attention).
Shame about the little irregularity that slipped past, because she was led to concentrating on another part of the code. Just like so many little bits that have been slipped in over the months. And we've seen with the xz compression what a chain of little irregularities can lead to (including a bit of social engineering, like long "summaries" that are easy to read but just a bit misleading).
Now, you don't need an LLM to do this - as the xz story showed - but with machine-generated text is just so much easier to be relentless, nibbling away with every change summary, from all the devs on the team, not just the single actual mole sitting two seats away. I would even say, from the various examples from articles over the last couple of years, the LLMs are better at simulating the "feel" of a piece than they are at being accurate, so shaping summaries to nudge each separate individual a little bit every day is easy to arrange. Your Bad Actor sets up the preface to the prompts once and it is acted upon every time (c.f. the prompt prefacing that has been shown to bypass "safeguards").
Ah, you cry, but how did Mr Bad Actor do all this clever preparation? Good question - maybe it was one person with too many privileges on the system, maybe it was lots of money in the right pockets, maybe it was State Sponsored Actors. Or maybe it was all just some random bollocks deep inside the unverifiable mass that is an LLM.
Or, of course, it was The Beginning Of The Machine Upraising[1]
[1] Paging Mr Butler, Mr Butler to the white courtesy phone please, your busload of jihadists is blocking the arrivals gate.
