Showing posts with label Sci Tech. Show all posts
Showing posts with label Sci Tech. Show all posts

Wednesday, 29 April 2026

What headphones were made for

Eighteen years after Weller at the BBC comes another compilation of Beeb TV and radio recordings by the Modfather, imaginatively entitled Weller at the BBC Volume 2. It's a 3-CD set of 48 tracks spanning that period, and it's mostly excellent, occasionally just good. His cover of Billie Eilish's What Was I Made For?, recorded in 2024 for Jo Whiley's Radio 2 Sofa Sessions, is a real highlight, and perfect for a certain state of mind. See what you think.

As someone who, many years ago and in a former life, used to sell hi-fi, I was quietly pleased to note Paul's headphones in this video. They are Beyerdynamic DT 770 PRO and, if you look really closely, you can see they are the 80 Ohm impedance variant, as you might expect in a studio. Let me tell you, these will sound absolutely fantastic. I used to love Beyerdynamic headphones, and they were an easy sell to customers looking at high-end kit - all you had to do was get them to take a listen, and they'd be getting their wallets out. Not cheap, but worth it.

Of course I've long since moved on from the world of hi-fi staff discount, so have to content myself with more modest kit. I have some very cheap unbranded earpods for commuting, which are okay. I also have some Aftershokz AS600 bone-conduction headphones for cycling, but don't use those very often, excellent though they are. For more serious wire-free listening, I have a pair of Sony WH-CH720N Bluetooth cans, which perform some pretty amazing noise-cancelling wizardry whilst also delivering proper sound. Best of all, if I'm at New Amusement Towers and don't mind being tethered, I have a pair of cabled Sony MDR-10RC which, though getting on a bit now, might be the best headphones I've ever had, even if they lack the 720N's out-and-about real-world abilities.

What about you, what are your headphone weapons of choice?

Thursday, 26 March 2026

Don't need a weatherman...

...to know which way the wind blows.

So opined Sir Robert of Zimmermanshire. Similarly, I don't need a weather station in my back garden, but I have one anyway because: (a) the weather is interesting; (b) I like data; and (c) I guess I'm just a massive nerd.

Anyway, the night before last, weather was happening fast.

You know when Tomasz Schafernaker gets on the TV in front of his weather map green-screen and talks about a cold front moving in? Well, this is what it looks like when it happens quickly.

Data from my weather station

Yes, this is data from my weather station, which gets uploaded to Weather Underground in near-real-time. Look at that drop in temperature, around half ten. I believe the technical term for that is "fell like a stone". From 10.6°C at 10.28pm to to 7.1°C at 10.49pm. That isn't usual. And the wind, which had been building since 9 o'clock, really ramped up, peaking with gusts of ≈30mph.

I know, this is boring, sorry. I remember learning, in O-level Geography, that a weather front would pass over you entirely in 24 hours. But it can also move quicker. I'll spare you the barometric pressure graph for the day, that pointed to this happening. Let's have an appropriate song instead.

More soon, when I take a deep dive into why the pump don't work (spoiler alert: the vandals took the handles).

Thursday, 25 December 2025

TV times...

Christmas television is not what is was in our younger day, quality content diluted across so many channels and platforms.

So here's an alternate viewing schedule for you, comprising videos I've squirelled away in my YouTube Watch Later list but never really found a reason to post about individually. Start this straight after Christmas lunch and this little lot should see you through to bedtime. Merry Christmas, ya filthy animals...

  1. Alice Roberts | Morals Without Religion: the Unholy Mrs Knight and the Hypocritical Humanist - in case you've had too much religion this week (43:53)
  2. FT Drama starring Stephen Fry | Is AI going to change who we really are? - short thought-provoker, feels very now (13:43)
  3. BBC Archive | Big Jim's Boozy Bike Trip to Braemar - a reminder why Nationwide was better than The One Show (7:05)
  4. The Jam | Danish TV Concert Special - nicely remastered TV special from 1982 (38:00)
  5. CBS Mornings: R.E.M. on songwriting, breaking up and their lifelong friendship - proper Christmas feelgood (41:20)
  6. BBC | "Call My Bluff" S11 E5 (1977) featuring Gabrielle Drake, Tom Baker, Miriam Stoppard, Alan Coren - tea-time telly with (70s sigh) Nick's sister Gabrielle... (29:56)
  7. Mel Smith & Griff Rhys Jones | The Homemade Xmas Video - my concession to the fact that it's Christmas, after all (32:54)
  8. The Royal Institution | The harsh reality of ultra processed food - with Chris Van Tulleken - something to digest as you, er, digest... (57:53)
  9. Documentary | He's Starsky, I'm Hutch - be honest, you're already hearing the theme tune in your head (44:40)
  10. DUST | Limbo - nothing says Christmas like a short film of Black Mirror-esque dystopia (24:22)
  11. Fearne Cotton's Happy Place | Minnie Driver On How The Meaning Of Life Can Fluctuate - I could watch Minnie all day (54:54)
  12. The Diary Of A CEO | Jimmy Carr: The Easiest Way To Live A Happier Life - love him or loathe him, he has some interesting things to say in this long-form interview (1:40:28)
  13. Graham Norton | Robin Williams - unrivalled late-night chat show fare (37:42)
  14. Television Archive | If I Ruled The World - late-night panel show comedy from 1999, if you still don't want to go to bed... (29:23)

Monday, 13 October 2025

It's not what you ask, it's how you ask it

With apologies in advance to the delightful C and all other AI sceptics. I'm a sceptic too, if that helps...

Experimenting with ChatGPT some more, I attempted my usual baseline of getting it to create something you'd never see. But unlike Sora, it's fussy about real people... requiring me, ironically, to get creative in the wording of my prompt.

Prompt to ChatGPT

I like the way it guided me in how to get around its own constraints. And the result?

ChatGPT output

Addendum

John was much smarter than me. He asked, "Can you create a photograph of someone who looks like the singer Morrissey eating a burger in a McDonalds resturant?" And this is what he got back:

Sunday, 22 June 2025

Abject fAIlure

First off, let me promise to you that this is the last post I'll write about AI in 2025. I've got a longer-term AI-related project on the go, but the outcome of that, if indeed there is any kind of outcome, won't be known until May next year. So, honestly, this is the last post on AI for a good while, which is good because, after all...

Keep seeing this pinging round the internet with the creator unacknowledged. Anyone know who did it? Oh God please let it be AI-generated.

[image or embed]

— Ian Martin (@ianmartin.bsky.social) March 21, 2025 at 4:45 PM

With those wise words ringing in our collective ear, cast your mind back to last month when I wrote about Suno, the scary AI music generation platform. And for latecomers, that's not an app for generating scary music, minor keys and Psycho-strings, no, no - Suno is scary because of how quickly and easily it generates anodyne, MOR pap, the sort that fills supermarkets aisles and hold-please phone lines and mid-afternoon commercial radio stations and song contests and more. It's a technological software marvel, no doubt about that, but it's also somewhat worrying, to say the least.

The "success" of my little Sunovision Song Contest last month did make me wonder though. Could Suno create a good song? What if it was fed decent lyrics, rather than merely prompted with a theme to write its own? And what if the music prompt was specifically targeted, with a predetermined end-result in mind?

You can already tell where I'm heading with this, can't you?

A quick copy and paste supplied the lyrics to one of my favourite songs, by one of my favourite bands. I added prompt keywords to Suno for the music - "british accent, male vocal, mod, revival, new wave, beatles-inspired bassline, strong beats, power pop, open chords, electric guitar, bass, drums, weller, foxton, buckler" - and clicked the ironically labelled "Create" button, and this... there's no other word for it, this abomination is what it came up with. Ladies and gentlemen, with profound apologies, I present Thin As Thieves by The Conserve. Give it until the first chorus, at least.

The Conserve · Thin As Thieves

I think we can all agree that is just appallingly, abysmally awful - a festering turd of a pseudo-song. Unadulterated, unmitigated crap, and you are, of course, most welcome to hate me for inflicting it on you. Okay, you might justifiably question my prompting skills, but that's not the real problem here, is it? Either way, I won't do it again.

To redeem myself, here's the untouchable source material, from 1979's almost-a-concept-album Setting Sons - this is a proper song, and a damn good one at that. Think of it as an aural sorbet, after the travesty above. AI is undoubtedly getting more sophisticated at an alarming rate but for now, at least, there's no substitute for fire and skill.

And some final thoughts from a long-time fan of this very blog:

Monday, 19 May 2025

Sunovision results

So Austria's JJ, a counter-tenor at the Vienna State Opera no less, pipped Israel at last knockings to win Eurovision, courtesy of a massive public vote share. The awkwardness of an Israel win in the current geopolitical climate was thus avoided, and JJ has since returned to Austria to a hero's welcome. The UK did reasonably well in the judges' vote but again got absolutely nothing from the public, and finished 19th out of 26. My hat remains uneaten.

More importantly though, there was a winner in the Sunovision Song Contest. This was also close, with the win being secured by the last vote. Without further ado, I can reveal the results were as follows:

6th. Alone Together by The Verdant Sigil, representing Virelia
5th. Lonely Shadows by L'Orée d'Or, representing Elystène
4th. In the Quiet of the Night by Solène Morra, representing Caldovaria
3rd. Terminverzug by Frostwake, representing Drevona
2nd. Symphony of Collapse by Elias Vergrave, representing Ormandia
1st. Dancing Kitty by Silver Vatra, representing Thesskora

What does this tell us? Well, the most bland, most anodyne, most MOR pap occupied the bottom two places. First and second were close, and a good margin ahead of third (though Frostwake's heaviness won them a lot of votes). As for those top two, well, Symphony of Collapse is the one I personally like most, musically and lyrically ... but Dancing Kitty is about a dancing cat, and is slightly earwormy, so what's the world to do?

Conclusions? AI is scary and Eurovision is largely pap. There. I'm glad it took two blog posts to work that out.

Let's close with a real song by a real band, to cleanse the palate. The Rat by The Walkmen came on the radio at the weekend, and I was quietly delighted to see Amusements Minor enjoying it as much as me. "Who's this?", he asked. A song that's older than he is, still gaining his attention, and later added to his Spotify playlist. Dancing Kitty won't be doing that in twenty years' time.

Douze points for The Walkmen from me.

Friday, 16 May 2025

The Sunovision Song Contest

Does anyone actually care about Eurovision anymore, in any way that doesn't involve irony or drinking games? Turns out that none of the Amusemenents clan had even heard the UK's entry in the build-up to this year's competition, so some deliberate YouTubing took us to What The Hell Just Happened? by Remember Monday. For my money, it's pretty terrible, features one of the worst first lines in pop music history ("Someone lost a shoe"), and can't seem to decide which of three different types of song it wants to be. If it's in the top half of the table at the close of Eurovision proceedings... well, I don't wear a hat but if I did I'd be seasoning it and finding some cutlery.

But anyway, you've read the title of this post and are possibly wondering what the Sunovision Song Contest is? Let me explain.

Last Sunday night I sat alone, nursing a lovely pint, in an unfamiliar pub in an unfamiliar small market town. At the next table, a loud woman with a dog was pontificating about all manner of things, to the rapt attention of three very different men, all of whom seemed to be falling over themselves to pay atttention to her every word, apparently for the sole reason that she was borderline visually attractive. I may be oversimplifying but that's how it seemed to me, over the rim of my glasses. And trust me, I didn't want to listen in, but couldn't help it as she was quite loud in a otherwise quiet pub. After listening to how she would redesign one of the bloke's website for 50 grand, and how her dog was barking at another dog, she finally got onto Suno. "Have you heard of Suno?" she asked her three devotees, "It's incredible, look." And then she proceeded to demonstrate the Suno AI tool that enables users to provide a text prompt and have a song spat back at them. Yep. A fully fledged, multi-instrument, verse/chorus/middle eight/break/fade song with original lyrics, churned out by Suno's AI engine in a matter of seconds. Honestly, it took longer to type the prompt than it took Suno to generate the song... a song that borderline woman proceeded to play loudly from her phone, shattering the peace of the pub even more than her voice and barking dog already had.

God, I sound like a cantankerous old sod, don't I?

Of course, the acolytes all cooed, "That's amazing," or similar. And in a way, it is. Technically, as a programming and computing achievement, it is astonishing, even. But musically? Borderline woman's Suno-song was crap. Anodyne, generic, pointless, meaningless, forgettable... just awful.

And then it occurred to me, after listening to the UK's Eurovision entry yesterday and accidentally sitting through a few minutes of one of the Eurovision semi-finals, that all those words could probably be applied to a lot of the songs that will get churned out in Basel on Saturday night too. And an idea formed.

So here we have the inaugural Sunovision Song Contest. Yes, I used Suno to create the songs, and in keeping with that I used ChatGPT to create fictional countries and their representative artists. But listen up! You have to cast your vote at the end, so strap in and let's take this semi-seriously...

First up, representing Elystène, is L'Orée d'Or ("The Golden Edge") and their track Lonely Shadows. The island nation of Elystène is known for its high art, fashion, and diplomacy, apparently. L'Orée d'Or describe their style as dream power pop, and are known for their ethereal soundscapes, poetic lyrics, and avant-garde fashion.

Next up is Symphony of Collapse by Elias Vergrave, representing Ormandia. Elias is classically trained, would you believe? Of course you would, he's from Ormandia after all, a central European duchy with misty hills, medieval castles, and a strong literary culture.

Getting exciting now, isn't it? Silver Vatra describe themselves as politically charged dance punks mixing traditional sounds with modern beats. No surprise as they're representing Thesskora, a Balkan state with a turbulent history and rich musical tradition. But there's no rebellion in Dancing Kitty, an unabashed Europop anthem!

The Verdant Sigil are from Virelia, and are known for blending traditional forest instruments with modern rock, evoking the mysticism of Virelia’s deep woods and mountain temples. Not that you'd know it from their entry, the cryptic power ballad Alone Together.

Next up, the beautiful Solène Morra, representing the warm, Mediterranean nation of Caldovaria, where Roman ruins and Moorish architecture mix. Solène is a sultry and dynamic solo artist who blends classic Mediterranean rhythms with contemporary pop and Latin guitar in her entry, In the Quiet of the Night.

And finally, the bookies' favourite, Frostwake, representing Drevona, a northern land of fjords and pine forests, with a strong seafaring and folklore heritage. Frostwake combine droning harmonies with ancient Drevonan runic chants and heavy instrumentation on their lively entry, Terminverzug.

So bienvenue and wilkommen, members of the Blogosphere jury, now it's down to you. Who gets your douze points? And who gets the dreaded nul points? You must give every song a score, and you can only give each score value once. Vote now!

As ever, I'm playing with all the AI stuff so you don't have to. And let's be clear, the six tracks produced here are all pretty awful by any objective measure. Laughably so, in places. But really, are they that much worse than some of the pap that will get served up at Eurovision in Basel? Possibly not. And here's the other thing. All the AI work, the Suno track generation, the Chat GPT fictional countries, bands and bios, took less than ten minutes all told. Writing this post took me nearly an hour, what with uploading MP3s to Soundcloud, drafting my devastating prose, formatting it all and pressing "Publish". And therein lies the rub. If you want originality and quality, you might be prepared to wait (and pay for) carbon-based creatives. But increasingly customers are just going to want something cheap and quick. AI is coming for us.

Come back soon to see which track won Sunovision 2025, and what prompt I used to create it...

Tuesday, 11 February 2025

D'you want a bag on your head?

Way back when, before my so-called career (ha!), I used to work in a hi-fi shop. It wasn't quite like this, though there were occasions when the customer's lack of technical knowledge could be advantageous. Upselling, I'd guess you'd call it these days. Don't judge me, it was a very low paid job with a commission element. Anyway, what I can say is that it was easy to sell expensive kit as long as it was demonstrably better. As long as I could demo a Sony CCD-TR805 camcorder I would sell it, more often than not, even though it was hideously expensive at £1,099.99, just because it was obviously so much better than everything else. Ditto the brilliant TCK-611S cassette deck, again expensive at £299.99 but an easy sell once demonstrated. Likewise the WM-DD33 Walkman (£99.99).

I was a good salesman, I think, because I made sure I knew my stuff, and used that knowledge to find something that matched the customer's needs. I enjoyed the job too, more often that not, and met a good mate there in the form of Tim, a friendship that endures to this day, despite rarely seeing each other.

I was interested in it all too, which is why I still knew those models numbers and prices instantly off the top of my head, despite my time there being more than 30 years ago. The brain's a funny thing, I guess. For the record, the 805 was a better camcorder because it had an optical image stabiliser (basically a clear gel between two lenses that acted as a dynamic prism) rather than the digital efforts of other brands, which tended to give grainy, pixelated results because they just didn't have the necessary processing power back then. And the 611S brought Dolby-S noise reduction to the domestic market, so much better than B and C; with a good quality blank tape (Sony Metal-XR, for example) the 611S would be the best way of recording CDs until MiniDisc and CD-Rs came along. As for the WM-DD33, well, DD stood for direct drive - no drive belts to stretch over time - plus it had heft! A sign of quality components, back then. Terry Hall had a WM-DD33, fact fans (and so did I).

I'll end with a track from a CD we used for demos, mostly because it was a DDD recording - digital recording equipment, digital mastering, digital media (rare then). With a good amp and decent speakers you could really appreciate the sound quality however much you liked the music ... and Sting irked the musos, even then. Here's the closing track from Ten Summoner's Tales, in all its 21st Century, compressed, low-bitrate, embedded YouTube misery. Does anyone care about sound quality any more?

That's a very Beatlesy outro there, isn't it?

Monday, 3 February 2025

Speak!

What do you think about podcasts? Do you listen to them? Do you subscribe? Any that you'd recommend?

I don't listen to many - most of my offline listening is downloaded radio shows from BBC Sounds, and that's not quite the same thing. But I do try to keep up with the usual suspects (Adam Buxton and Louis Theroux, for example). I've also stumbled across the odd gem - The Old Fools, featuring David Quantick and Ian Martin, is a joyous, if occasionally sweary, gem. But then maybe I like that so much because I am, increasingly, an old fool myself. I am the target market. I am the demographic.

Have you ever tried podcasting yourself? If so, what platforms or technologies have you used? How did you find the experience? Would you recommend it? Do it again? Are you still doing it?

And if all this makes you think, oh god, isn't it enough that he already writes a blog that barely anyone reads, now he wants to launch a podcast that ever fewer people will listen to, well ... the thought has crossed my mind. It probably won't happen - it's a lot more effort than just typing a couple of paragraphs, adding the odd hyperlink and then embedding a YouTube video, after all. And I'm not so great with effort, these days. So, rest easy...

Speaking of embedding a YouTube video, I should probably end with one. What is this blog without a bit of appropriate music, after all? (That's a rhetorical question, though the answer would probably be "Not very much.") Anyway, if the only song you're aware of by Sultans of Ping F.C. is the brilliance of Where's Me Jumper? then I'm about to double your knowledge. You can thank me later.

Just one more thing: having asked for your recommendations it seems only fair to reciprocate. I use AntennaPod to consume anything that I don't already get from BBC Sounds. It's brilliant, free, open source, has no ads and, best of all, doesn't require you to create an account. I don't know if something similar exists in the Apple ecosystem but you can get AntennaPod at Google Play and you should because it's a proper piece of software. Although I appreciate that when I say "...and you should" I sound like Patrick Bateman imploring people to listen to Huey Lewis's lyrics.

Monday, 6 January 2025

Take AIm

Or TIWHBALYIIHBOS (Things I Would Have Blogged About Last Year If I Hadn't Been On Sabbatical) #2

With all apologies in advance (especially to C, who I know gets understandably hot under her exquisite collar about this), one more AI experiment. As ever, I'm doing this crap so you folks don't have to, okay?

So I've previously played with text-based AI, and image generation. Now onto video, courtesy of Minimax. For consistency, I've stuck with the same idea of trying to generate something you would never see for real. Specifically, my prompt to the AI was "Singer Morrissey eating a hamburger in an empty McDonald's restaurant, at night". This is what it came up with.

All very impressive, I'm sure. He looks nothing like Morrissey, and doesn't actually eat the burger, just holds it whilst his jaw is moving. Plus, it's very short and has no sound but other than that ... well, it looks like real, carbon-based creatives can rest easy ... for now, at least.

Anyway, with apologies (again) to all those who don't any more, here's the real McCoy with his northern outsider pals, when they were still pals, most decidedly not eating a burger.

Forty years ago though. Stone me.

Saturday, 4 January 2025

Remember when "compact" mattered?

Or TIWHBALYIIHBOS (Things I Would Have Blogged About Last Year If I Hadn't Been On Sabbatical) #1

I had to get a new phone last year, unfortunately. And it seems that if you want a decent spec, compact phones are no longer an option. My new phone is a full 20mm longer than the one it replaced - so much for progress. Add on the new protective case I also had to buy and it all adds up to a ridiculous slab. I may or may not be pleased to see you, but it really is a phone in my pocket.

Anyway, it's time to belatedly update the mobile timeline:

Because these photographs are inexplicably popular (in web searches, at least) they have their own label so, for completists (!), here are the previous posts in the series.

Thursday, 2 January 2025

Simplification (or, time for it all to go)

It's time to accept the Internet is never going to make me rich.

Likewise, I'm never going to win the world snooker championships, lead the Tour de France, or write an international bestselling novel. But back to the Internet and, more specifically, this blog.

For the past two years of so, I've ended every post with this, a kind of Patreon-esque way of letting readers tip me any amount, big or small, for something I've written.

Tip the author

In all that time, no-one has ever tipped me anything, which is fair enough. Patreon and the like are clearly best left for high-volume, mass-market, original content creators. None of those adjectives apply to me, or my "content". Besides, if you wanted to pay to read something, newspapers are available (for now).

Never mind, eh? For the past eighteen months I've also had this in my sidebar, allowing readers to buy me a cuppa.

Buy Me A Coffee

In all that time, no-one has ever bought me a coffee, much less a notional cup of tea that I would actually enjoy. And that's fair enough too - I'm probably over-caffeinated anyway.

So it's all going, from today. New Amusements is eschewing third-party advertising wherever possible, and becoming a public service broadcaster. With apologies to the actual BBC, whose sign I've photoshopped...

The British Blogging Corporation

All of which is my tenuous double-link to a song, Go by Public Service Broadcasting.

I especially love the very enthusisastic "Go!" from Steve Bales on Guidance. He was only 26 at the time, can you believe?

Anyway, in the unlikely event that you do want to somehow make me pence richer, you can always buy my book. Or a t-shirt! What a time to be alive.

Monday, 16 September 2024

Wiped out

Disclaimer: this post was written in December 2023, and scheduled for future posting. Its contents may no longer be accurate or appropriate.

It's getting on for thirty years (okay, it's 29) since I first played Wipeout with my excellent friend Tim. The basic premise was ... well, you raced anti-gravity ships around tracks, picking up weapons and other enhancements as you went, and tried very hard to be faster than the computer-driven opponents. The graphics and speed of play were literal game changers back in 1995. As was the soundtrack, which included a selection of high-octance dance tracks. We almost always raced to Chemical Beats by The Chemical Brothers, over and over again, for hours at a time, into the small of the night.

There's a moment from about 2:10 in this where Ed and Tom introduce a sort of protracted phasing of the top-line melody, producing something akin to a Doppler shift. On our favourite race track, this would usually line up with an echoey section of tunnel, in a way that just seemed so very ... right.

Two years later, the sequel came out (Wipeout 2097) and from then we raced almost exclusively to an instrumental mix of Firestarter by The Prodigy. A better track, in most objective terms, but not quite as good for anti-gravity racing (although the intro was great on the starting grid). This little clip gives you a sense of late 90s gameplay, although I can't help but point out the player here, whoever they are, misses the turbo start off the line. Rookie error.

Tip the author

Friday, 4 August 2023

Yesterd-AI

I watched Yesterday again recently. It was on late on the Beeb, nothing else remotely watchable was on and, besides, it has a charming if slightly unoriginal premise that makes it worth a look - imagine what would happen if the rest of the world collectvely forgot something monumental, and only you could remember. Hence struggling singer-songwriter Jack Malik (Himesh Patel) is the only person who can remember The Beatles after waking up in an alternate reality where they never existed. Cue lots of cameos from Ed Sheeran (the best of which involves Ed comparing Jack to Mozart and himself to Salieri), lots of songs everyone knows to ramp up the feel-good, some good in-jokes (like Oasis not existing either - "Makes sense," says Jack), and even Lily James as the smalltown love interest. And since the whole thing is a Danny Boyle and Richard Curtis creation, you're in safe, if slightly too safe, hands.

But this isn't a film review. There's a scene two thirds of the way through the film where Jack gets to meet John Lennon. Who is in his seventies, living by the sea, quietly creating art, having had a happy, normal life. Well, here it is.

Which is nice, isn't it? Nice to imagine John's life without the events of December 1980.

Did you recognise who was playing John, in that clip, by the way? Behind the de facto round glasses and under the docker's cap? An uncredited cameo from regular Boyle collaborator Robert Carlyle, but I digress. The point is, I guess it was easy to imagine John looking like that: the glasses, the hat, the jeans. It's how I might have drafted him, given the brief. But how, I wondered, would AI imagine him? Like this, it turns out:

My prompt to the AI engine for this was "John Lennon as a pensioner", rather than "...at 78". But you get the idea. Aside from one side of his glasses being on the wonk, it's quite good. Or is it? I mean, it's well rendered. But does it capture the essence, the spirit of the man? Maybe I should ask a real, human artist to draw John as an old man (any volunteers, C?) and see how it compares.

Tip the authorGiven the artifice of AI, it feels like there's only one song to end this post with. And for non-Beatles obsessives, that guitar solo from 49s is George.

Thursday, 15 June 2023

Last one on AI for a while

I've written a few posts on AI lately. They have proved surprisingly popular, more so than my usual guff about some old song or other that I like. Although when I say popular, what I mean is they've had a lot of traffic and comments. I'm not trying to suggest the coming omnipotence of AI is popular, because it is far from that amongst my modest readership. But indulge me one more time. I read about a very basic limitation of one AI engine, and so had to try it out for myself. How's your counting, because ChatGPT's isn't always very good...

Oh dear. The correct answer is four, as any fule kno. So what's gone wrong? If I ask how many "z" are in "zzzzzz" it correctly answers six, so it's probably not to do with repeated letters. Maybe for words that it doesn't immediately recognise (and there are many variant spellings of pizzazz) it looks to the dictionary definition and counts from that? Or from the pronunciation pɪzæz?

Maybe I could teach ChatGPT to count...

Which, to my programmer's mind, suggests it is probably missing the last "z". Think for loops versus while loops, maybe. Either way, seems I'm still asking the question wrong. Let's try again. If I was programming this task, let's say in JavaScript, it's a doddle to iterate through a string and count how many times a certain letter appears. Look, I did it here if you want to learn some basic code - as you can see, it correctly returns 4. But what does ChatGPT make of it?

Which is really, really disappointing. For whilst it correctly understands and describes the purpose of the function, rather than run through the code it shortcuts straight to what it thinks it already knows...

Tip the authorThis is a very human response though - to have a preconception of what is right and assume it to be true, rather than retest one's understanding. Confirmation bias, in other words, and we all do it at times. I hope, for our sake, that ChatGPT is not exhibiting egocentric bias... Whatever, it seems that the real skill in these early days of AI is asking the question correctly. And then, just maybe, being sceptical about the answer.

Post script

ChatGPT, backed by Microsoft, struggles, as per the above. Bard, developed by Google, gets it right first time, as below. Not only that, it also offers a snippet of code to prove its answer, inadvertantly demonstrating that my programming skills will soon be obsolete too...

Monday, 5 June 2023

Art for AI's sake

In which I continue to experiment with AI so you don't have to...

I read recently about a website called Night Cafe that enables users to generate all manner of pictures, in all manner of styles, without ever picking up a pencil or paintbrush. I had to have a play, of course, in my continuing but most likely futile attempt to be ready for the eventual takeover of our robot overlords.

It's pretty good, actually. You provide it with a text prompt of what you want to draw (the irony here being that you don't actually do any drawing), select a rendering model and processing algorithm, and click 'Create' (the irony here being that ... oh, you get the idea by now).

I thought I would try to create something that would never be seen otherwise, so entered a prompt of "Morrissey eating a hambuger". Here is the result:

There are, of course, a couple of things to note here. I cannot draw this well, for starters. But it's not perfect. For a start, it hasn't followed the brief: SPM is holding the burger here, not eating it. And finally, from this and other experimentations I can tell you that Night Cafe, at least, struggles with hands, their angle, their positioning, even (sometimes) the number of fingers and having two right hands and no left. Look at SPM's left hand here, for example. So it struggles with complexity ... for now, at least. I might try this again in a year's time... Tip the author

Wednesday, 22 March 2023

Nothing like the sun (or more about AI)

Estelle Skornik Thomas Thorne William Shakespeare Bill Gates
What do these four have in common? Nothing, other than all appearing in this post.

As you'll know from my last post, I've been playing with ChatGPT, the better-than-you-might-expect AI chatbot that Microsoft are getting behind (I'm on the waiting list to try Google's Bard AI).

Since yesterday was World Poetry Day, I asked ChatGPT if it could write a sonnet about French actress Estelle Skornik. And if you're thinking that names rings a vague bell, she was briefly famous in the 90s for playing Nicole in those Renault Clio ads. Anyway, here's what the bot wrote, literally in a couple of seconds:

Fair Estelle, with eyes so bright and bold,
A beauty that inspires both awe and grace,
Your presence on the screen is a sight to behold,
A talent that few can ever hope to trace.

With every line you speak, your voice doth flow,
Like honey, sweet and rich and full of life,
Your passion for your craft doth clearly show,
And stirs within us all a yearning strife.

Your beauty, like a rose in full bloom,
Shimmers with a radiance that cannot be ignored,
And captivates us all in its sweet perfume,
A sight that we shall always adore.

So here's to you, Estelle, with love and admiration,
May your star forever shine with such brilliance and dedication.

So what do we think? I mean, yes, technically it is a sonnet - fourteen lines, rhyming, expressing romantic love or infatuation - and it is about Estelle (an Estelle at least, rather than necessarily Mlle Skornik) ... but it's awful, isn't it? I mean, I'm no expert but this feels like the kind of "moon, June and coffee spoons" rhyming that you'd find in a greetings card, or at best something Thomas from Ghosts would write. But is it any better or worse than arguably the most famous sonnet in the English language?

My mistress' eyes are nothing like the sun;
Coral is far more red than her lips' red;
If snow be white, why then her breasts are dun;
If hairs be wires, black wires grow on her head.

I have seen roses damasked, red and white,
But no such roses see I in her cheeks;
And in some perfumes is there more delight
Than in the breath that from my mistress reeks.

I love to hear her speak, yet well I know
That music hath a far more pleasing sound;
I grant I never saw a goddess go;
My mistress, when she walks, treads on the ground.

And yet, by heaven, I think my love as rare
As any she belied with false compare.

Yeah, that's better, isn't it? That Shakespeare bloke was alright, wasn't he?

The thing is, though, this AI lark: it's in the kindergarten right now. Think how quickly we went from the Wright brothers to Apollo XI. Think how quickly we went from a Nokia 3310 to the iPhone. Now think how quickly AI will go from Hallmark poetry to Shakespeare...

From one Bill to another. Bill Gates is in the news today, suggesting that AI is the most important tech advance in decades. For what my opinion is worth (nought), I think he's right. And whilst Gates cautions that governments should work with his industry to "limit the risks" of AI, he is broadly optimistic about the benefits of this new tech. I worry more though. For every positive application (research, medicine) I see more negatives: you think algorithms are bad now, just wait 'til some commercial AI thinks it knows you better than you know yourself. And worse than the monetisation will be the militarisation. Gah.

Let's end with a song. The aforementioned Sonnet 130 gave this post its title, but also gave Sting a line for a song and the title for an album in 1987. Now Sting irks a lot of people, almost as much as Bono, but he's done some good stuff, and I'm not just talking about his time in the Rozzers. Anyway, to prove that I don't have a hip bone in my body, other than my pelvis, here's an excellent version of that song, Sister Moon, by Gordon Sumner and his jazzy mates. Until such time as AI can play the drums with the intense joy and precision of Billy Kilson in this, maybe we can rest easy.

Sunday, 19 March 2023

About AI

So, I've finally got to the top of the waiting list to have a go with ChatGPT, the AI chatbot from OpenAI. From my admittedly limited experimentation thus far, I'd say it's scarily good at some things, not so good at others. Examples, you say? Well, I started off with a vaguely work-related question:

Q. Will AI replace data managers?

A. It's unlikely that AI will completely replace data managers in the near future. While AI can automate many tasks related to data management, such as data cleaning, preprocessing, and analysis, human oversight is still necessary for many aspects of the data management process.

For example, data managers play a critical role in defining the data needs of an organization, selecting appropriate data sources, and ensuring the accuracy and quality of the data. They also need to interpret the results of data analyses, communicate findings to stakeholders, and make recommendations for action based on the data.

Furthermore, data management involves many non-technical skills such as communication, problem-solving, and project management, which are difficult for AI to replicate. While AI can assist with some aspects of data management, it is unlikely to replace the need for skilled human professionals in the field.

That's not too bad, is it? On the money, I'd say, if a bit bland. So what about something more taxing? Maybe a question about the future wife, also requiring some creativity, perhaps...?

Q. Can you write a limerick about Kylie Minogue?

A. There once was a pop star named Kylie,
Whose music was loved far and wily,
From "Can't Get You Out of My Head"
To "Spinning Around" instead,
Her fans danced with joy so spryly.

Which I think we can all agree is terrible. However, could any of us do any better, with that brief, and in less than twenty seconds? Not sure I could (and that's being generous to myself). Next up, a question that requires opinion and morality to answer, and has vexed many of us for some time.

Q. Is Morrissey racist?

A. There has been controversy surrounding Morrissey and accusations of racism. Some of his comments and behavior have been criticized by individuals and groups as being insensitive or offensive towards certain racial and ethnic groups.

For example, Morrissey has been known to make comments in interviews that some have interpreted as being anti-immigrant or anti-Muslim. He has also been criticized for wearing a badge with the logo of the far-right political party For Britain during a performance on a US talk show.

However, Morrissey has also denied accusations of racism and has spoken out in support of animal rights, veganism, and the LGBTQ+ community. He has also collaborated with artists from diverse backgrounds and included diverse musical influences in his work.

Overall, the issue of Morrissey and racism is complex and controversial, with differing opinions and interpretations. It is up to individuals to form their own opinions based on the available evidence and their personal values.

Which is more a collection of statements than an answer, and which is as admirably on the fence as even the most die-hard SPM fan.

So, let's do one more, just the one as I know this is getting boring for you. There's a skit in The Simpsons (series 8, episode 6) where Milhouse's parents' marriage is breaking down; playing Pictionary at a dinner party, Milhouse's dad Kirk has to draw a representation of "dignity" and his wife Luann can't guess what it is. Because how do you draw dignity in a doodle? Except when challenged by Kirk, Luann does exactly that, the joke being that we don't get to see the mystery picture. Why have I described the scene, when you could just watch it on YouTube right here? But anyway, that leads me to my final question for ChatGPT, and the one that caused it to think long and hard...

Q. How could you represent dignity in a picture?

A. Dignity can be represented in various ways in a picture, depending on the context and the intended message. Here are some possible ideas:

A portrait of a person who exudes confidence and self-respect. This could be a close-up shot of their face, capturing their expression and body language that conveys a sense of dignity.

A scene of a person standing tall and strong in the face of adversity. This could be a silhouette against a dramatic background, such as a stormy sky or a cityscape.

A group of people holding hands or linking arms, symbolizing solidarity and unity. This could be a diverse group of people from different backgrounds, genders, ages, and cultures, showing that dignity transcends individual differences.

A photo of a majestic animal, such as a lion or an eagle, that represents strength, courage, and nobility. This could be a close-up of the animal's face or a wide shot of it in its natural habitat.

A picture of a grand monument or a majestic landscape that inspires awe and reverence. This could be a panoramic shot of a mountain range, a waterfall, or a cathedral.

A simple, minimalist image that conveys a message of dignity through symbolism. This could be a single flower standing tall amidst a barren landscape, a candle burning brightly in the darkness, or a hand reaching out to help someone in need.

Ultimately, the best way to represent dignity in a picture is to capture the essence of what it means to you and convey that message through visual storytelling.

Which is actually not a bad answer, if a bit clichéd. A noble lion, an eagle, a mountain... It's almost like motivational posters have been designed by AI for years.

Okay, I know I said that was the last question but I'm human, not an AI, so I'm prone to error and changing my mind, so indulge me one last time. This is how I ended my session on ChatGPT.

Q. Is ChatGPT any good?

A. As an AI language model, I am programmed to provide helpful and informative responses to the best of my ability based on the training data that I have been given. However, like any technology, there are limitations to what I can do and areas where I may not be as knowledgeable or accurate. That being said, I strive to provide the best possible answers to your questions and constantly work to improve my capabilities. Ultimately, it is up to you to decide whether or not my responses are helpful and meet your needs.

Which basically translates as "I'm good but I'm modest with it." And that has to be a good thing - the more modest the machines are, the less chance we will all end up in a dystopian wasteland, ruled by mechanical overlords from the Cyberdyne factory. On that note, let's end with a video that is supposed to highlight some of Boston Dynamics' robot fails... except it really just shows us how good they are getting. I'm not sure how I feel about all this, but I'd be interested to know what you all think.

Monday, 9 January 2023

About bookmarks

I have a complicated relationship with Twitter. I mostly dislike it, especially the chaotic version that has been in place since the Musk takeover... but I can't seem to tear myself away completely, because just occasionally it throws something up that I perhaps wouldn't otherwise have seen.

Like most people, I retweet things that I think need or deserve greater attention, and like things that I want to show support for (whether I actually "like" them or, rather, feel some other emotion in response, like sympathy). I guess most people do something similar. But how do you use bookmarks? I tend to use them as a reminder to come back to something when time or circumstance allow. By circumstance, I most often mean when I'm somewhere I can play a song or video without issue. Anyway, the upshot of this approach (and the raison d'être for this post) is that I end up with a backlog of bookmarked tweets that I often don't get around to ever actually catching up with; in other words, I have a virtual to-read pile almost as big as my real one.

All of which is an overly wordy and slightly boring introduction to a cheap post that relies solely on embedding other people's content. I know, the shame... but regardless, in no particular order here's a small selection of the many, many tweets I've bookmarked in the last three months. Enjoy, or not.

And that's only a fraction of it, but there - I've done Twitter for you. You can thank me later (or not).

Thursday, 26 May 2022

The Unewsual XI - the UK produces the second highest amount of e-waste per capita in the world

So we should make our devices last longer, not get caught in the upgrade spiral, and repair things went they fail. But also, at least, this: