Never analyse comedy. Analysts have no sense of humour
AI is no resolution, it is a revolution
This week you find me in a vibrating skin.
At least, you would if I actually had one to put on. But it exists! I am, of course, referring to OWO’s haptic gaming system built into a $500 figure-hugging (i.e. flab-revealing) cycle jersey. Ubisoft says it will bundle one with a special edition of the imminent Assassin’s Creed Mirage this October.
Personally I cannot wait to set my torso a-buzz with haptic feedback every time I get wounded with an arrow, slashed with a sabre or shot in the back. Let’s hope it is an outrageous success this autumn and encourage OWO to develop some haptic cycling shorts to go with it. Then I’ll really feel I’m in the game next time I fall awkwardly on a chimney-pot while parkouring across medieval rooftops. Ouch, that might put a strain on the old marital relationship.
Now that I think of it, Ubisoft has a big office in Canada… maybe that’s why Justin Trudeau is separating from his wife: all those hours of ACM and Justin’s preproduction vibrating shorts rubbed her up the wrong way. I understand Mr Trudeau now plans to spend more time on his other personal interests, such as quantum computing, boxing and face-painting.
This kind of story used to be the backbone of my weekly chronicles of the computer industry, here at Autosave is for wimps and before that at The Register. In between tales of office bafflement and Fortean error messages, I would often return to simple reports of unlikely tech products threatening to lead us further down the path to dystopia. Since so many businesses are keen to roll out increasingly loopy applications for computerised electronics, most of which eventually fail because they all are utterly mad, they’re an easy win for a writer of mildly amusing weekly tech opinion columns.
Since the beginning of the year, however, the tech agenda has been dominated by large language model chatbots which, let’s face it, are not particularly funny. That said, there is so much going on in the nutty world of Chat GPT and its ilk that I could very easily just cherry-pick the nuttiest AI chatbot stories each week for your mild amusement and not bother with anything else.
I could do this. It is tempting – not least because the topic is so unimaginatively repetitive that I could teach a LLM chatbot to write the columns for me. It would be a shame, though, because there remains plenty of potential in the barrel of dystopia that I keep by the back door next to the compost bin.
For example, there’s a crowdfunder here in France for a smartphone case designed to dispense contraceptive pills.
I had wondered initially under what circumstances one might have one’s smartphone literally in hand for it to realise one might need reminding to take the pill. Does the device’s accelerometer detect a regular rocking motion? Does an image recognition routine peek through your camera lens and identify that some, er, jiggery pokery is taking place?
Actually, none of these things. I had it womansplained to me. The Lunes smartphone case sounds a daily alarm to remind you to take a pill, pops it out and that’s it. As the blurb says: Une sonnerie, je l’attrape, je l’avale, fin. In English, that’s: “A ringtone, I grab it, I swallow it, end of story.” Of course, the French is ambiguous in that ‘le’ in this context means “it” but could just as easily mean “him”, which certainly makes the slogan more colourful in its native language.
No less colourful is the accidental fact that the French word for a smartphone case is a coque. They’re missing a trick here. What could be more natural than to think about contraception every time you grab a coque?
If vibrating gameskins and pill-popping coques don’t sound dystopian enough for you, try my third product of the week: a $1600 domestic robot pet with a creepy infant voice (think: ‘Noddy’s Eurasian cousin’) and looks like a decapitated whippet. Things get even more uncomfortable when it offers to “pleasure you” by dancing.
No? Oh come on, at least it’s funny. I could resort to some AI stories if you prefer but while they strongly lean towards dystopia, they’re more likely to make me sweat in terror than make me snigger.
For example, remember a few weeks ago when I reported how ex-US Air Force pilot Missy Cummings was making waves in – that is, incurring death threats from – the autonomous vehicle industry? Having taken time out from her post-Top Gun career in academia to work in government regulation of transport systems, she learnt a thing or two about how unreliable and dangerous AVs can be.
Well, Missy got invited to write a longer piece for IEEE Spectrum last month to explain her findings. It does not make happy reading. You know all those daft and irresponsible things that LLM chatbots do? That’s your next autonomous vehicle, that is, and it’s already on the road. “While a language model may give you nonsense,” she writes, “a self-driving car can kill you.”
A laugh a minute, she is. But that’s the problem: AI is the promise of dystopia without punchlines or giggles. No amusing asides. No mirth. Not a knob-gag in sight. AI is not funny. But then, they say, you should never analyse comedy.
Aaaand… off your balls again, Hugh.
Maybe the reason AIs aren’t particularly funny is because nobody has properly taught a chatbot what comedy is. You can teach it to repeat jokes, of course, but does it understand them?
A brave attempt to do this, albeit in a round about way, was published in a research paper last month by the Association for Computational Linguistics in Canada (all dystopian trails lead back to Canada, have you noticed?). Engagingly entitled Do Androids Laugh at Electric Sheep? the paper reported how its research team trained AI models to attempt New Yorker magazine’s notorious weekly cartoon caption contest.
They pruned 14 years’ worth of the contest to leave some 650 sample cartoons of sufficient digital image resolution for the AIs to ‘see’ what is going on in them, then challenged the AIs to pick the winning caption in each case from a multiple choice offering that included irrelevant one-liners. The chatbot then had to explain its choice and summarise why it is considered funny in conjunction with the cartoon.
Just to be sure that image recognition wasn’t an issue, they also tested the AIs by describing in text form what certain cartoons depicted before getting them to pick the right caption. And, naturally, they ran the same tests by volunteer humans.
TLDR; the AIs guessed plenty, but not most, of the captions right. You must admit this is a good start, especially given that even the human volunteers floundered over some of the captions. Humour is not universal, is it? However, even when the AIs picked the right gag, often its explanation did not make sense: it didn’t get the joke and nevertheless determined the correct caption through some other process – possibly by eliminating the alternatives as being utterly irrelevant first.
So, no digital giggles from the AIs just yet. Despite this, the authors of the paper suggest AIs may yet be useful for helping cartoonists and humorists brainstorm potential jokes.
I can hardly wait.
So let me finish on a caption competition of my own. Mme D took this photo while out and about this morning:
Have a go writing a caption. Feel free to type it in the Comments. The winner might win something but only if your caption is funny. I’ll let you know the result next week.
Alistair Dabbs is a freelance technology tart, juggling IT journalism, editorial training and digital publishing. According to this report, Chat GPT can only tell Dad jokes. “What do you call an alligator in a vest? An investigator.” Are your sides splitting yet?
”Philip always crossed his T:s.”
(If the walker had had tennis balls stuck on its feet, ”Off yer balls again, Hugh!” would have the obvious caption.)
Part of her disability was the inability to park inside the lines.