Posts by that one in the corner 5494 publicly visible posts • joined 9 Nov 2021
FB are making money by using statistical models to evaluate statistical data and deriving actually useful results from that.
And as a side hustle, by selling time on the bullshit generators to other companies, which are using them to generate more ads that they then pay FB to display.
Sounds like FB have a handle on how to use this AI malarkey. For their own good, at least.
> No wonder there were so many bugs in the first place
Care to list out this massive set of bugs you know of? How long is it really? And compared to any other errata (let alone the non-published, shush, we'll get away with it, internal addendum)?
Anything in that list that prevented the device from being useful in the majority of cases?
The GPIO pullup issue is unfortunate, but then some of us were used to having to solder pullups/pulldowns on absolutely everything and view these built-in ones as a bit of a luxury[1] (oh, ok, it adds a couple of pennies to the BOM - IFF your design actually hit it). Actually, I know of a few small run builds - not with the RPi - where the guys were hardware- rather than software-oriented and added all the pullups anyway, as that was easier and more cost-effective from their p.o.v. than worrying if they'd set up the registers properly in their code.[2]
As for the glitching mentioned - you did understand, didn't you, that this is not talking about glitches in the silicon design but is about people introducing glitches in power lines etc from outside the device and feeding those in, with the intent to make it misfire? So operating way outside normal and specifically to see if they could break security features?
The public bounty for breaking security features is a Good Thing - it encourages the participation of some truly devious minds to have a go AND talk openly about what they are doing on the forums.
> company can't be arsed to hire proper talent
And just how do you expect a company to find these talents without setting this kind of challenge? You may find it hard to understand, but some of them actually consider etching off packaging and probing around under the microscope to be a fun hobby.
[1] cue "pullups? You were luck! We had to tie 60V to our toe and stick fingers on the i/o lines"
[2] and yes, their code was very pedestrian and could be "improved" - especially for readability! - but that is the joy of small-run: Done. Out the door. Next.
At least Midgely was solving some real problems and his solutions did genuinely work; it was only afterwards that we realised there was potential for problems with one of them (the lead was probably foreseeable). So Midgely wasn't ever facing a stream of vocal criticism each time his work was released.
But if Agent P. is blissfully unaware of criticism...
> it happens that a single person may have multiple roles, or jobs within the company, each role or job having a separate set of valid filesystem accesses.
Which is what groups gives you.
> there are not enough groups, and group access is insufficiently fine-grained
This means you haven't bothered to analyse your requirements and defined, then created and used, enough groups to control your own specific environment. Ok, there are limitations from some file systems (NFS, IIRC) but kludgey careful use of groups gets around those (i.e. define the set of groups important to whatever is being accessed via (NFS or ...) to be within the range you can use there).
Nor is it clear how having a different home folder helps - aside from selecting which login script gets run and which ~/.myapp-config.d gets invoked, which are separate from shared resource access control, aren't they?
> Further, there are things which happen in login scripts which may be required for Department A, yet should not happen for Department B.
Can you elaborate on that, as it is not something I've encountered in practice (simply because I've not worked in every environment on the planet, not because it is necessarily unusual): do you mean w.r.t. job function - e.g. logging into the correct time management system? - or physical constraints - e.g. Dept A all use printer PrintA, Dept B uses PrintB?
Either way, it really isn't clear to me why adding this into the OS ((making it more complex for everyone) really helps in such situations instead of, say, shoving into their login scripts a question box that just asks "What are doing today" then does the necessary.
> The AI bit is irrelevant. It merely makes this probably a bit quicker (and more
costly once the AI hypesters (and their supply chain) start charging realistic prices to their clients, instead of, oooh, running on debt financing of their GPU farms.
But don't worry, the corporations will just pass their sudden increase in costs onto the consumer (except those who con government into "bailing them out of this crisis that nobody could have foreseen").
And avoid the Big Corp Supermarkets with their ePaper price tickets.
At least you'll be able to spot when your local shopkeeper is running ahead of you, swapping paper tickets on the shelves, after he spots the platinum card peeking out of your wallet.
Killer app?
Are there any even vaguely useful apps that require an AI PC?
Seriously.
Searching throws up a few toys. So far, not even anything that looks like it'll help anybody who is already doing anything locally with ML, as they (still) need the beefy GPU (they already have) to make any real dent in execution times.
For the improved model of the Tesla Toaster, the classic hot cheese meal. Pay up now to go on the waiting list and it will be delivered next year^^^^^^^^^in 2027^9^^^^sometime.
Asked if it was going to be as jaw-shattering as the current offering, Musk replied that "It's really not that hard".
You are describing IPMI and its ilk (I'll just use IPMI as a blanket term) or the use of a hardware KVM-over-IP, not netboot. If you don't like using the name of a specific implementation, just use "IP KVM" or similar, rather than usurping "netboot". After all, a decently run render server farm will be using netboot as well as a KVM, IP or otherwise.
Oh, and (actual) IPMI is at least 25 years old, so falls outside of the "some things have got 10x better (within the last 20 years)" claim. And that is just when it is integrated into the main chipset. Bolting an extra computer onto the side of the bigger one, in order to control the latter, is older still: if you needed it, it was available back in the 1980s (at least) as well. So even PiKVM etc are just the current versions of the Old Ways.
Network booting has been available since the 1980s, when it was useful as a way to save money - give your computers a netboot-capable NIC (which could mean buying the BIOS-extension ROM to plug into the card) and then let it boot from an OS image on the one machine on the LAN that hard a hard drive.
Changing the OS image just needs a settings tweak on the OS server and then reboot the remote - which could be (can be) done in any of a number of ways, depending upon what is running on the remote computer (e.g. just log in via telnet and issue a shutdown -r or have the remote poll for a reboot command or - just wait 'cos it'll crash every 20 minutes anyway and the watchdog will bring it back up).
> Some things have improved 10X.
If that were true for netboot then we would all have a really easy way to set it up for everyday use in every situation, whereas we seem to be more wed to every computer having its OS on its own drive than ever before. Yes, some IT departments can set it up, but by now the advantages of having every PC in the office loading a fresh, uncorrupted, totally up to date, copy of The Company OS Image should be seen everywhere. It should have long ago become so easy that every school used it by default.
Although, as we're now all meant to be "doing it in the cloud", and those are just VMs booting from a prepared image...
Glad you had some good recollections.
Although I have to admit that I'm more of a Tyrannosaurus Rex man than T.Rex; more "Stacey Grove" and "Salamanda Palaganda" than "Jeepster" or "Born to Boogie". Although, "Ride A White Swan" - let's just call that a transition piece.
> And there's yer [citation needed],
Well, in addition to the reference I gave in response to the first of your posts...
Let me just Google that for you...
How about this:
‘Auntie’ has been the jocular, and once derisive, nickname for the BBC (British Broadcasting Corporation) since the mid-1950s .The first known citation is believed to be a Leslie Illingworth cartoon, which appeared in The Daily Mail in 1953 welcoming the news of commercial television in competition with ‘Auntie’, namely the BBC.
That disagrees about the date of the first citation (please feel free to nip down to the British Library newspaper archive and check for yourself) but the date and content of the cartoon are in agreement with the other article's citation of Paul Holt, published in the Daily Herald (London, England) of Tuesday 18th November 1952: talking about the 1953 start of commercial television in contrast to the BBC "The film men have decided to bank all on the calculation that dear old Aunty BBC will continue to offer viewers those neat, nice little family dramas while they go out after bigger stuff."
The start of ITV was a Big Event in 1953 and a cartoon at the time would have a wider audience than the 1952 wordy article, so giving Leslie Illingworth credit for bringing the phrase into the wider usage is quite reasonable - even if he did not coin it, as I'm sure you are aware, Leslie Illingworth and his peers (as if!) were reknown for helping the public remember the occasional - appropriate - phrase.
The use of "Auntie Beeb" was very familiar in the 1960s as well, when BBC Radio became the "target", as its output, even on "The Light Programme", was being unfavourably compared to the Happening and Now overseas stations like Radio Luxembourg and the not-quite-UK-based pirate radio stations like Radio Caroline.
Of course, after that, the phrase was kept alive by "Cuddly Ken", Kenny Everett, and his wireless show - even whilst he was on Capital Radio, telling us how he got fired by the BBC for being flippant and pushing Auntie just a bit too far (although was what he was saying about politician's wives or the Musician's Union that made Auntie's lorgnettes fall?).
For the majority of the time, the BBC itself most certainly did NOT refer to itself as "Auntie"; your recollection of the show "Auntie's Bloomers" was from the 1990s, when the BBC was poking fun at itself and was ready to loosen up a bit. It didn't even refer to itself as "The Beeb" - this not even being acceptable on "Just A Minute" (although that may be because of the number of players who fell foul of the buzzer for repetition when reverting back to "B - B - C" after being caught out on the deviation of "Beeb"). Irreverant broadcasters such as Terry Wogan might use the abbreviation, but it wasn't used by any formal representative of the management and it was a genuine surprise to some when an official channel name formally recognised "The Beeb", in the form of CBeebies.
> Ok then, give some non-BBC usages of the term
Talking about the Pomona Times in the US:
>> An early use occurs in the following from The Daily Progress (Pomona, California, USA) of Monday 1st April 1901: “Felix Grundy’s” reputation for veracity and honor is not surpassed by that of any man in Pomona, not excepting the proprietors of Auntie Times.
The same article also refers to a mention of "Auntie Times" in the 'Bystander' from1935, although that one is referring to the London Times newspaper, rather than the Ponoma Times. It then goes on to give a few more before giving its first reference to the usage directed at the BBC, which he dates to a piece in 1948.
Because the phrase is "Auntie Beeb"?
Please don't fall into the trap of assuming that, because nobody in your circle does something, that means nobody anywhere does it.
And, no, trying to dig up references - or a lack or references - as convenient URLs is not a proof, either.
I agree, the phrase is not one even I utter every day, but then again I don't use the words "diplodocus" or even "T-Rex"[1] every day.
[1] yes, the dinosaur reference is deliberate, saves anyone else the need to belabour it
> which, if you're not up to speed, the BBC has been held to account in the recent past for not being truthful, or good for that matter.
The Beeb has always been under the eye of the watchdog - and broadcasts whenever it has been caught not doing its best.
> Look up Jimmy Saville, Rolf Harris etc
Clearly, not going to justify anything that happened with those - except to point out that, unfortunately, there are bad apples - and outright scum - at about the same (low, very low - that is why it is newsworthy) rate everywhere, in every organisation. With the same levels of cover-up pro-rated by the power of the individuals. The BBC is large, so, with deep regret, statistics imply...
But where were we able to follow the court cases? The BBC news was giving them coverage. All that kerfuffle with Cliff Richard getting rightfully annoyed with the BBC? Heard about on Radio 4 on the drive to work.
> I for one certainly do not look at the BBC with affection and call it Auntie
Tarring everyone and every output from an organisation by highlighting the worst is - well, I don't accept that as rational.
We have had Andrew Wakefield, Harold Shipman, a number of "Nurse Death" cases and there are plenty of issues around funding and NICE, but I'm not going to turn my nose up at or my back away from the NHS.
The BBC is not what it once was, from my p.o.v. But it can not be - the environment around it has changed. As have I (with the best will in the World, Pogle's Wood no longer holds my attention as it once could, even just as a lover of animation).
The London Times was another institution that was once referred to as an Auntie to The Nation - the BBC has held its place far better than some.
> Auntie is a self description of the BBC, by the BBC.
Twaddle.
The Beeb was given the appellation "Auntie" by the public - well, by the print media, such as cartoonists - as a way to describe its output as being a bit staid and polite when compared to the bright and brassy (not sassy, far too early for anything in Britain to be sassy) offerings from those rude upstart channels that were obviously just going to be pandering to the desires of the advertisers that will pay their wages.
What constraints? Be specific!
A screwdriver has specific constraints - they are usually clearly written on the packaging and sometimes even engraved on the shaft: this is a Philips number 2, this is an 3 mm flat. Neither will perform well on that Torx head bolt.
And if the manufacturer hasn't told you the maximum force it can apply to the lid of a paint tin, you can perform the experiment on a sacrificial unit and be in with a very good chance that the same spec qua screwdriving will perform similarly qua tin opening when you go and buy a replacement.
This compiler has specific constraints: it only does C/C++, so don't feed it COBOL. It is also old, so predates these language features. Because it is not the latest, we have even had time to characterise its bugs, so avoid doing anything in this addendum.
This editor has specific constraints: it only works on text files up to 2 GB less 4 KB.
And so on and so forth[1]. Those are all tools (they clearly say what jobs they can do) - with specific constraints (they - and sometimes also we, as we learn their extreme limits - clearly say what they can *not* do).
Now, where are these *specific* constraints defined (and published!) for the Replit (and other "vibe" coding - things)? We can all find the fluff and hype, but specifics? It is to laugh. HA!
So even after he has shared with us his masterly insights (see above), we know that he is *still* under the hypnotic thrall of the vibebeast.
Hang on. "Hello, is this SaaStr? I have an idea - an AI app for de-programming! Hello?"
[1] that is Forth-83, so yes, you can use multitasking. See, it is in the specs, along with the constraint (co-op tasks, easy to use but you have to yield or they'll block)
> So I am curious why there would be any interest in the 80286 given it's shortcomings.
Because there were a *lot* of them about, a lot of people did use them, both as "better DOS boxes" in Real Mode and a variety of OSes in Protected Mode.
Having accurate emulators for - well, everything - is an important goal to keep information alive. You never know what old project may suddenly be revivable and be acclaimed as an important historic artifact.
PS you haven't give your opinion on the TMS9900, the IMP-16 or even the NEC mu-COM-16. Did *anyone* use the mu-COM-16, outside of (presumably) some NEC equipment?
Clearly I have been doing it wrong all these years.
The step from running on a data copy to dog fooding my own work, let alone the push to QA (you are a dev and you get to push to public access yourself?!) is usually filled with dread that I missed the blindingly obvious and I'm about to be told that "the horrid bug has gone, good job on that, but is there a reason the video is now upside down?".
The "rush" involved is towards the car park and a calm getaway before the roads fill up.
To point out that the software was working precisely as spec'ed, but the End User wasn't using it properly.
No, that wasn't in the spec, yes, it would be useful[1], here is an estimate.
[1] note: always "useful" or "an interesting addition" whilst on site, never "so bloody obvious an omission that a blind man could see it, I kept on telling the managers we needed this, but, no, no, it was always far more important that we kept changing the shade of green in the prompts until Sales were happy it matched the letterheading".
Perhaps the first glimmerings of AWS being realistic about overhyping current "AI" trends and giving us the tools to defeat them?
Or just named by the same team who didn't think they needed to put a hyphen into the URL for their Expert Sex Change forum and could never figure out what the conversations were all about?
Glad to know that your experience not working on US freight trains has been so useful in teaching you all about the protections built into things that are not US freight trains.
Meanwhile, for everyone who does work on, or just lives near to, US freight trains, TFA has something relevant to say.
> Bill Nye, Last Week Tonight with John Oliver ... Does that sound.. very scientific
Ah, good old JE, having to dig up a piece from a 2019 John Oliver show to have a dig at Bill Nye and his ilk.
If Last Week Tonight is JE's standard for a scientific forum to criticise, we will be in for a treat when he catches up with Professor Hannah Fry's turn at chairing Have I Got News For You.
Microsoft are investors in ChatGPT, the generic product that was tested here - but you sre implying that have given OpenAI access to a lot of material from systems that are outright owned by Microsoft.
Do we have any citations for this?
I.e. that MS are effectively allowing any competitor to buy time on ChatGPT and get the benefits of the extra, "private", materials-worth of training?
Now, an MS-specific instance for Copilot, trained in the extra stuff, that is another thing.
PS even a "private" repo, you should not be storing your Windows keys etc!
All the fury about the LLMs containing this kind of information and how terrible that it wasn't stripped from the training data[1] or how weak the "guard rails"[2] are to allow it to spit the material out.
All that the article is *really* saying is that lots of things you wish you hadn't made available are trivially found by enough web crawling. The only reason you can't get the same stuff out of a simple Google search is because Google is now crap and won't do old fashioned "find me a match" searches. In this case, the LLM happens to be working better as a search engine (but don't expect all of its "results" to be *real* keys - some are, others were undoubtedly made up to fit the pattern).
Be glad of that.
Because of the naivety of the LLM, you have, once again, been warned that you published all those keys and they *have* been found.
And LLM scrapers are not the only things that will have read them. The AI slingers got them by accident, along with all the names of Barbie's pets. More directed scrapers are going through your Git repos, deliberately looking: your unsanitary habits are not only being used to create amusing articles about how awful LLMs are, tut, tut.
[1] how? Then imagine your reaction when you are told the training set is first filtered to find all the secrets that were made available to a web crawler - and you realise this means that the filter's log contains all the good stuff in concentrated form. See above.
[2] those "guard rails" really appear to be a sop, attempts at post-processing whose only practical results are that there is now a fun game for researchers to dream up more convoluted ways to hide their data requests (let's try some JSON today - boo, the guard rails are grepping for those now; how about HTML tags? Yay!).
> The usual problems you find with aluminium cable are caused by incorrect termination
That is the nub of the problem I was referring to: Al is no problem for the big boys who know what theh are doing with the odd kV here and there, but for the folk who are just getting used to 40V down an Ether cable, not realising those thin wires are not nice, well behaved and really rather forgiving copper can lead to unexpected, and definitely unwanted, problems.
> talent gap cited is less about absent AI skills
As there is more than enough information available about these systems: from the opportunity for all staff to try them out for free online, up to books, courses and coding examples to let the more technically inclined really dig in and see what is going on (Hugging Face is calling out to you and there are GPUs aplenty to rent by the minute).
> than human skills – the need for creativity, innovation, leadership, and critical thinking among staff to find beneficial ways to apply AI.
The critical thinking is probably what is getting in the way - you've foolishly let your staff gain all the AI skills and now they are critically comparing benefits to costs.
But never fear, as TFA points out, there are still consultancies who will bring in unbiased - i.e, totally ignorant about your company and products - people who are self-proclaimed "(thought) leaders" and "Creatives (with no other qualifiers, just randomly creative?)" to help find innovative ways of shoe-horning the machines into cracks in your processes[1]. Consultancies that are desperate to paint the growing loss of faith as a failure in your staff rather than common sense.
[1] looking for a spiffy short phrase to describe the results of doing this - "Process Spalling"? It needs more alliteration. You know the effect - someone shoves their pet hobby horse into a small gap in the flowchart and for a short time it expands, using up resources, then they get bored, it shrinks down, the gap reopens, then either the budget has to be used or it'll be cut at year end, so the hobby gets fed again - or someone else shoves The Latest Thing into the gap. Rinse and repeat, until one day the whole side of the process graph is undermined and falls off. To the shock and surprise of everyone whose work is usually totally unconnected with horses, hobby or otherwise, who now find they have been relying on this mess without realising it and are suddenly aware of all the equine byproducts that need to be cleared away.
On the lighter side, the next soap opera plot:
As Michelle and Jason stand together next to the celebrant, a voice calls out from the door: it is Miguel, Michelle's husband, who had been declared dead at the end of the last season. Miguel holds up his smartwatch and tells the stunned congregation that Jason had unsoldered the heart sensor, causing Xealth to report his demise; in reality, he had been sitting in a deck chair in Spain since the Christmas Special episode.
Doof-doof-doof
