Rick's b.log - entry 2025/11/02

mailto: blog -at- heyrick -dot- eu

This is not the future we were promised

The reality, however, is... In the future computers/machines/AI will pretty much suck at a lot of these mundane jobs, but they'll be just about good enough that companies can use them instead of employees, but things won't get cheaper because every person laid off is extra profit in the pockets of the few obscenely rich people that own these sorts of things. "The paperless office" is a joke, if anything the introduction of computers into the office made things worse because it was now that much easier to have so many more pieces of paper; and a general lack of access controls, boredom, and terrible user interfaces mean it is now much easier for somebody to screw up potentially exacerbating the problem (only, these eventual issues will be somebody else's problem - but good luck getting anything resolved "because the computer said...").
For customers, things are starting to turn paperless. No more printed bills, no more tax forms, it's all done electronically now. This is in part due to the cost of postage these days, but it's also because by pushing people to use "the official app", one can quietly collect a lot of telemetry in the background to sell (imagine, if you will, a random official from your energy supplier sitting in the corner of the room quietly watching what you are doing; then add a person from your bank, one from Amazon, somebody from your broadband provider, and another from.... you get the idea).

Why do I even have to answer this in order to pay my bill?

Meanwhile, in the greatest theft of original works imaginable, AI companies have been retrieving and storing copies of just about everything they could get their grubby mitts on in order to try to claim "fair use" rights to music, books, photos, websites, etc etc.
I think you will understand that - morally - "fair use" is best defined as "if you make some money from my work, share some with me". But this is not what is happening. Instead people's works are being effectively stolen to allow a machine to create lookalike copies or a bunch of mindless memes in seconds.
To put this madness into context, OpenAI, the company behind ChatGPT, DALL-E, and Sora manages to lose tens of billions every year...and they're in the process of trying to do an IPO (that means get listed on the stock exchange) with a valuation in the ballpark of one trillion dollars.

To put this into perspective, the are about 1.2 billion websites but only around 16% of them are actually active. So let's round that figure to an easy two hundred million active sites. Using very simple maths, one could assume that the valuation for information stolen from those sites to "train" their AI is $5,000 per site. Let's knock off a grand to account for the creation of the software and other nerdy stuff - that's two hundred billion, already stupid money); and that still leaves four grand per site.
In reality it will be radically different as sites do not all have the same relative value. That is to say that The Guardian will be worth more than HeyRick, which will be worth more than The Daily Mail, for example.

And through all of this, machines effectively suck really hard at jobs that are not rote identical tasks. A machine that bangs doors onto cars can do so quickly and effectively because that is all it does. Machines that drive on public roads? Despite all of the hype and the nonsensical "Full Self Driving" name, fully automatic cars have had rather limited success and their share of accidents (including fatalities). Granted, things like Waymo have to deal with us meatsacks and we're half the problem... but that's just how it is. Meatsacks have to watch out for other meatsacks doing dumb things, and this particular meatsack looks in the rear view mirror as much as ahead, because if anybody is going to do something idiotic, it's likely the car behind getting irrationally impatient. The more random reality is, the less machines are able to cope because they can only cope if the software provides a mechanism to understand the problem and mitigate it. Most of us meatsacks, but not all, are running rather superior firmware...although there are liquid substances that are known to dramatically interfere with the operating system...

As for that blessed human creativity? Well, it turns out that a machine can actually be rather good at that. There are tell-tale signs. I'm sure by now you've come across websites offering solutions to various problems (like such and such happens with my phone, how do I fix it) that seem to be oddly repetitive in their phrasing. These sites were very likely created by AI to maybe aid you (if they're at all accurate) but mostly to draw your eyes to all the embedded advertising. And, yeah: force close the app, clear the app cache, remove and reinstall the app, clear the Android cache via the secret recovery menu, reset Android... the latter two should never be performed by the inexperienced, and 99.9% of the time are completely unnecessary regarding an app that won't work correctly and have potential severe consequences (badly hacked versions of Android will fail if the cache is cleared and may then fail to reinstall itself - Klipad I'm looking at YOU); and resetting Android will wipe all of your data/photos/other apps (and badly hacked versions will fail completely at this point - Klipad I'm still looking at you).

But even so, AI generated photos are getting to be pretty good these days - especially the more abstract ones so one doesn't have to worry about mutilated fingers. The picture below was created in seconds by a machine from my description. All I had to do was sit through a minute long awful and intensely annoying Temu advert. Again. The app seems to show the same stupid advert over and over. 🤷 Amusingly, rather than "yay, Temu!", I now have a deep hatred of the company and wouldn't go near it if you offered to pay for whatever I wanted to order. 🤦 Advertising has power, and it's not necessarily good power.

Line up with the portal and off you go to the other side of the galaxy.

For music? It is now capable of pretty impressive results, including capturing various human emotions. The machine doesn't understand any of these emotions, but love is love whether it's a person, a nice cup of tea, or chocolate. Once it gets to recognise patterns in the music we create, it can reproduce those remarkably well.
Let's put it like this - I'm not a musician. I would like to be capable of playing the piano but a lack of time/concentration and dyspraxia mean I probably never will. I wrote SimpleSeq because I saw that there was no current piano-roll style music editor for RISC OS; and the music software that is available (Rhapsody4, we won't count Maestro!) lacks a lot of the things that are possible with MIDI (and, even then, General MIDI (version 1) lacks a fair bit of expression that is possible with a real instrument). Anyway, I wrote it to try to allow me a way to turn my ideas into something "real" because I just wasn't getting on with traditional notation.
Had I known that software such as Suno would become available, well, honestly I wouldn't have bothered. Suno can do in seconds what I probably couldn't do in the rest of my life. And the remainder of today's blog entry is basically a big demonstration of that.

Coming Home to Tea

So I fired up the Suno app to ask for a song about coming home to a nice cup of tea. And, because I could, I created several different versions of the song. The lyrics are the same in each case. Yes, there are thirteen versions of the same song here.

Before I begin, can I say "Holy bloody shitballs!". The free version of Suno used to offer the 3.5 model which, as you will have experienced in the past tended to give rather autotuney-style vocals and fairly simple arrangements. It was a bit of a challenge getting it to do dual vocals, and it only managed to create one song in 6:4 time, the rest were pretty much straight 4:4.
Well, Suno have released the version 5 model to paying subscribers, so us freebie users now have access to the generic v4.5 model. The difference? It's like comparing a Ferrari to a Morris Minor.
You'll hear what I mean.

With Suno, every generation results in two songs being created. In each case in the past I would pick which of the two I liked, and if neither fit what I was looking for then I would just do it all again.
Free users get "credits" to make five generations per day. With every generation providing two songs, that's a potential ten songs a day. I would have said a more realistic figure would be two or three songs a day as you'd want to reject those that didn't quite measure up. But with the new 4.5 model.... well....

For each of these songs, I requested an unusual time signature and syncopation. The reason for this is to give a completely different ambience to the usual common time songs. It's no longer straight one, two, three, four, one, two, three, four. I don't always get it, but when I do it makes the song sound so much richer because it's just a little out of the ordinary.

Note, also, that the song numbers are not necessarily in order. The order they are presented here is the order in the Suno app list (going backwards from earliest), which is not necessarily the order that I saved them (the number suffixes). That's why it goes 2, 1, 4, 3...

Generation 1

The second version of the song was a lot like the first but with a slightly different arrangement, and note how much richer the instrumentation is, not to mention the vibrato vocals that now sound like a person more than a machine. And, wait, is that a subtle but of choir vocalisation in the second chorus? And that note is held for a good ten seconds. I think we can say that this one captures the feeling of coming home to a nice cup of tea.

Generation 2

Here's the second version. This is going for a much gentler melody, more restrained then the one above. I don't feel as if this one is running at 4:4, it isn't so easy to work out what time these songs are actually in given the erratic percussion (which in itself is not a bad thing - I don't like percussion that is just a regular tick, tick, tick in the background) but this one seems more like it would be 8:4. Or maybe it's just a really complex 4:4 taking the bars in pairs? Either way, this arrangement is superb, isn't it?

Generation 3

That's quite a captivating opening, isn't it? The older model couldn't really do this, but the 4.5 model absolutely delivers, and how. This is one of the shorter songs, but it uses its time well.

Oh, a moody sounding start tying into what the lyrics are about. By the prechorus we're already in the realms of epic, and going full symphonic for the chorus was a joy to listen to. This has taken what started off as a simple ballad and turned it into something much more powerful and dramatic - I mean we've gone from a relatively calm piano piece five songs back to Xandria.

Generation 5

After hearing some choral parts in the fourth generation, I was like "okay, let's toss a choir into the mix and see how we get on.
There isn't much choir in here, but it's a impressive progression from a calm beginning to a powerful ending that's almost symphonic metal. It's also one of the few to fill the entire four minutes.

The second version starts with a richer sound than the one above, it sounds less like a piano and more like a harp. As for choir, they turn up in impressive form and if you thought it would be hard to top the sixth version, well...
Note also the delightfully complex percussion. Suno clearly gets that I'm not looking for a boring-as-hell four on the floor.

Generation 4

The first version sounded really quite quirky and interesting. Unfortunately we didn't get much beyond the first half of the first chorus before the wheels fell off, it was speeding up and slowing down and sounding like it was pasting together random bits of music and not necessarily in any logical form. That's not to say it wouldn't have a place in a dystopic cyberpunk animé, but for a song to listen to it was weirder than the (red) waiting room in Twin Peaks. Count this one as a loss.

Given what a mess the first version turned out as, I had quite a bit of trepidation as to what would happen with the second. It turns out that it screwed up one to make a brilliant second. I'm okay with that, as this is my favourite version. The choral vocalisations, switching to the backbeat for the chorus, the slightly zany arrangement, this is the song I was looking for. The only problem with this version were two clicking noises in between the first two lines of the final chorus, I think the instruments came in a little too hard, but it was easy enough to isolate and attenuate them in Audacity to remove the clicks. It's worth doing for this, the one I'd keep if I had to get rid of the rest. But hey, I have storage space, I can keep them all. ☺

Bonus: Four more variations

And just for a change, how about some Celtic folk? This is oddly restful, just like a good cup of tea should make you feel.

Now for something a little more like mainstream pop with a touch of emo just to keep it grounded.

And fittingly, a more goth version of the previous song is the thirteenth one saved, running for three minutes and thirteen seconds.

I think after all of this, the only thing left to do now.....is go put the kettle on.

Bonus music

My €0,02 on RISC OS

Specifically this topic that gets close to making some unkind and unwarranted accusations.

First of all, something that doesn't appear to have been mentioned yet, is that while the RiscPC is an ancient clunker, it is our current only emulation solution. If you want to run proper RISC OS on a device that isn't a supported ARM machine like, say, an x86 Linux box or under Windows, you're going to be using an emulator and that emulator is going to be a RiscPC of some description. It is what it is, and in the absence of anything more recent, the thirty year old thing isn't going away any time yet. You don't need to support the RiscPC if it's too old/incapable for the things you want to do, but understand that by making that choice you're not only blocking out the refuseniks with obsolete machine, but also everybody using an emulator.

With respect to the allocations database, it is grossly unfair to call ROOL "gatekeepers". The very few times my requests went unanswered it's because it just wasn't seen. This stuff is handled by a volunteer. A resend later and the request is granted. The one time it wasn't is because it wasn't actually necessary as it had been included as part of an earlier request (I think registering a SWI chunk gets the associated error block... something like that, it was over a decade ago). ROOL might refuse if it's a weird or bogus request or clashes with something that already exists, or is an unjustified use of a limited resource (like Event numbers or somesuch). If this is the case, the reason will be given and since the person that handles the requests is helpful, an alternative may be proposed.

Due to the history of the allocations database, there is private stuff in there, possibly subject to NDA, and quite likely all these years later with the original documentation missing. The only safe way to decentralise the database would be to replace everything with a question mark and hope that people send in their current allocations so the database can be brought up to date. I would not bother as it's a dumb idea that solves no current problem. Also understand that by opening up the allocation database, you will be actively dissuading companies working on stuff "in the background" from making a timely allocation. They may only be able to make a request immediately prior to, perhaps on the day of, the release of their new "whatever" - and what do you suppose might happen if there was a problem or clash?
That's right. It'll become a survival of the fittest, like file extensions on other systems. How many different pieces of software do you think tries to "own" .bas, for instance? And before Microsoft Word was a thing, .doc was commonly used for text files. Now imagine a world like that where one module can trample on another, multiple applications try to set the same App$Dir variables... it would be a mess. That's why things need to be allocated, so somebody can hold their hand up and say DumbArse$Dir is registered to ME!

As for maintaining the database, well, somebody has to. Why not ROOL? It's a logical choice given that they're the home of the source code and the various releases. You don't have to agree on ROOL's leadership or direction, and I very much don't given that I feel that current 32 bit devices are being pushed aside to chase the delusion of a mythical 64 bit system - but whatever, none of this is at all relevant to whether or not you can fire off a request and say "I'd like to register AppName DumbArse please". And, you know what, I'd probably get an email back in a few days saying "Done, but be aware that some people might find that to be offensive", which would be an entirely valid point to make. I have not, ever, found the allocations process to be withholding, gatekeeping, or anything like that. Only, once in a while, snowed under. Did I mention, this is being done by a volunteer? Be patient, be polite, it's a human not a machine.

How much ROMs cost is hard to answer. There's the physical cost of the ROM, then there's the time it takes to make it. And maybe even test that it works correctly (or, at least, verifies as being consistent with the source data and boots in a real machine). The price, in this golden age of capitalism, is often "as much as the buyer is prepared to pay". If somebody else can come along and do it cheaper, well, the more expensive guy will have to either lower his price or risk going bust due to not selling anything. He might try to claim some sort of exclusiveness but there's only so many people you can hoodwink by putting it into a pretty box. That, also, is capitalism.

Whether or not the Titanium is, or is not, "finished" also is not especially relevant. The Pi is possibly the most commonly used system for RISC OS. It is also woefully under-supported. Every single model has VFP, which is barely used (indeed the DDE compiler just can't emit VFP code, nor does CLib support VFP style floats). All but the first version have multiple cores, unused by RISC OS. Hardware acceleration? The ability to switch audio output at runtime? Bluetooth?
Most of this comes down not to incompetence on the part of ROOL but simply a complete lack of people to do the jobs. That's why I feel that 64 bit is a time-wasting delusion. The list of outstanding wants for the current OS (like updated FileCore, for instance) is long. If that work struggles to get done, exactly how the hell is anybody supposed to write a new incarnation from the ground up?
I'll skip over the "two network stacks" thing. That was mind-numbingly stupid given the miniscule size of RISC OS development, but I think largely down to the mentality of nobody wanting to talk about some future project until there's something concrete to show for it. By that point it was too late as there ended up being two of the same thing happening in parallel. Idiotic, possibly avoidable had the head of this place talked informally to the head of that place, but it is what it is. Just lousy communication all around.
But, then, people screaming blue murder on forums is maybe why nobody wants to preannounce anything. Not every good idea results in a viable project. Not every promise can be kept. And sometimes, despite best (or worst) intentions, reality steps in and says "nope".

Consider this. I was looking forward to seeing what Cloverleaf's ArtCube software could do, as I find PhotoDesk's UI to be clunky and confusing and it's often rather overpowered for simply fiddling with photos to make them better. I usually use the editor on my phone to crop, to rotate slightly if it's out of line, to fiddle contrast and brightness (specifically light/mid/dark individually), and to adjust the "ambience" or "vibrance" (which is like saturation, but different as it selectively enhances less saturated colours only). I don't know if PhotoDesk can do all of this - I suspect maybe, but getting to grips with where and how is... way more challenging than it should be. So I was hoping that maybe ArtCube would offer a more user friendly solution.
And... that's about as far as the story ever went. Not enough money was raised to make ArtCube viable. I still edit photos on my phone, and these days I'm trying to wrap my head around Gimp that is every bit as annoying as PhotoDesk, only for a load of different reasons. ☺
My "this worked best for me" image editing software, for reference, was ULead's PhotoImpact 5 that came on a magazine cover-mounted CD-ROM. Worked under Windows 98SE and XP and that software was just intuitive and easy to use. It didn't do weird things like Gimp has a habit of doing. Something sort of like that for RISC OS would have been great. But, alas, that old familiar demon raises its head and says "not enough money, not enough developers".

That right there is the main problem faced by the RISC OS community, moreso than any notion of anybody's malfeasance. Sniping and accusing won't help, that's more likely to cause people to throw in the towel and go use their talents where it'll be appreciated. There are no easy solutions, only imperfect bodges, and one must make the most of what there is in a practical way. That is to say, if you can contribute in some way, then do so without saying "this must happen and this must happen" because it ought to be pretty clear by now that it's probably not going to unless you do it yourself... It is what it is, and that's what it is.

DDE update

Fortran, standing for Formula Translating System was first created in 1955, with the first manual created by IBM in 1956. That's almost seventy years ago. It is, actually, the granddaddy of a lot of modern programming languages. The original Fortran led to Fortran II on one side and Algol (1958) on the other. Algol led to CPL (1963), then BCPL (1966), then C (1972) and all of the C family that we know today. Algol also, by a different path, led to Ada (1983)... and by yet another path led to Pascal (1970).
Fortran II led to more variations of Fortran, but it also led to BASIC (in 1964). Indeed, the often gibberish commands in Fortran along with the bizarre way to specify a loop (DO 10 iter = 1, 100, 1 from the sieve example) and the even weirder IF syntax (IF (J.GT.2) GO TO 53 from the whetstoned example) led to the desire to create a language that was simpler on the programmer. User-friendly, as far as could be achieved using 1960s technology. This was the first version of BASIC. An entire generation of '80s children are thankful.

I am writing the name as Fortran because that's how it was originally stylised. If your browser shows "Fortran" then it doesn't support small capitals. Anyway, the use of small caps is why it sometimes gets called "FORTRAN" (in all capitals). These days it is called "Fortran" (initial F, rest lowercase) but that nomenclature was introduced with Fortran 90. The one supplied with the DDE is F77 (approved in 1978!) which was, I believe, the last version that only supported uppercase keywords.

The F77 standard, though, is ancient old. It looks like the compiler can optionally add some VMS functions (such as bit manipulation). There is a lot that looks and feels like the Norcroft C compiler, and that is because half of it is. The Norcroft compiler worked by translating the source into something they called Jopcodes which were then translated into machine instructions. This means that if you had something else which could create the Jopcodes, then that could be a different front end supporting a different language. In this case Fortran, but Back In The Day there was a Pascal compiler floating around too. I wonder if this will turn up at some point?

According to Acorn's Retail Price Index, September 1989, there was LISP (£228.85), ANSI C (release 3; £171.35), ISO Pascal (release 2; £113.85), Fortran 77 (release 2; £13.85), Assembler (£228.85), and Prolog X (£228.85). I wonder if any of these will turn up in time.
If you think those prices are eye-watering, go look at how much a twenty megabyte harddisc (as in, large enough to hold four or five of the above songs) would cost you. Or, for that matter, an Econet FileStore which was basically a custom device in a Master Compact body that was a tragically slow 6502 machine running a modified version of the Level 3 software in ROM alongside a very cut-down version of the Beeb MOS (only as much as is necessary to get the FileStore to work; along with extensions for the proto-SCSI harddisc). The price of the floating point expansion is why everybody used the FPE module instead.
Anyway, back to Fortran...

When Fortran 77 is ancient old, it turns out that Fortran is actually still in use today, the most recent standard bump being in 2023, which fixes a few issues in the 2018 version. Though, obviously, this compiler won't cope with Fortran code.
The instruction manual implies that one can create modules in Fortran. Hmmm... Has anybody actually ever, in the history of Acorn ARM hardware...? 🤔

Still, it's quirky but it is interesting.
I won't be learning any Fortran though, the syntax looks like it was written to aid the compiler, not the programmer. Given this would have been an era of batch processed punch cards on mainframes (which may well explain the rather odd line formatting), it very likely was indeed designed to help the compiler more than the programmer. There's a reason somebody decided to create BASIC instead, after all.

Well, that took all day to create. Time to feed kitty, make dinner, go to bed... wait... where did my day go?

Your comments:

Please note that while I check this page every so often, I am not able to control what users write; therefore I disclaim all liability for unpleasant and/or infringing and/or defamatory material. Undesired content will be removed as soon as it is noticed. By leaving a comment, you agree not to post material that is illegal or in bad taste, and you should be aware that the time and your IP address are both recorded, should it be necessary to find out who you are. Oh, and don't bother trying to inline HTML. I'm not that stupid! ☺
As of February 2025, commenting is no longer available to UK residents, following the implementation of the vague and overly broad Online Safety Act. You must tick the box below to verify that you are not a UK resident, and you expressly agree if you are in fact a UK resident that you will indemnify me (Richard Murray), as well as the person maintaining my site (Rob O'Donnell), the hosting providers, and so on. It's a shitty law, complain to your MP.
It's not that I don't want to hear from my British friends, it's because your country makes stupid laws.

You can now follow comment additions with the comment RSS feed. This is distinct from the b.log RSS feed, so you can subscribe to one or both as you wish.