ChatGPT 1844 Edition
6.8 minutes of fun from your friendly Patent Insensible Newsletter Clerk
Dear Readers
Wake up. Look busy. Our new automated overlords are here want to know whether you are “productivity or protein” so they can categorise you for… Oh, come one it’s not that bad. Look – Antonym has arrived a whole day early – like The Last of Us in Super Bowl week…
The Victorian AI writing machine
Amid the frothing and fainting about AI writing this week with Bing’s AI version launch (it’s pretty good by the way) I was delighted to find some historical context on the subject of AI-assisted writing. Jeannette Winterson points out in 12 Bytes – her brilliant collection of essays on artificial intelligence – that the idea of books written by computers dates back at least to a satire of the inventor of the first pre-computer (it was never fully built) Charles Babbage in the top London satirical magazine of the time:
Self-generated novels. Nothing new. The satirical magazine Punch offered this Babbage spoof back in 1844:
Sir... I have been completely successful in the production of a New Patent Mechanical Novel Writer adapted to all styles and all subjects... Babbage Followed by testimonials: By its assistance I am now able to complete a novel of 3 vols in the short space of 48 hours, whereas before, at least a fortnight's labour was requisite for that purpose.
– 12 Bytes by Jeannette Winterson
Novels were the junk copy of its day of its day, which may give hope to the SEO-smiths in 2023.
The writers of Punch must have continued to be amused by the hype around Babbage’s projects, because the same joke was resurrected about lawyers’ clerks (there is now an actual robo-lawyer, by the way, who seems to be doing good business in the US). Below is the letter I found in an archive of Punch proposing Babbage build “an Automaton Lawyer’s Clerk” or “PATENT INSENSIBLE LAWYER’S CLERK”.
There was a proposal to replace MPs with versions of the difference engine that would work out which way to vote. Punch felt this would be a quicker and couldn’t do much worse a job – sentiment I suspect many would agree with.
During the 1990s opinion polls showed that, in contrast to the rising cult of euro-scepticism in the UK, many Italians were warm to the idea of a fully federal Europe, reasoning that rule from Brussels couldn’t be worse than the madness and corruption in Rome. I think we in the UK know what they meant now.
How long have you been talking to that search engine?
According to an blog post from Microsoft about public trials of its new AI-assisted version of its Bing search engine have been going well and teaching them a lot about how to fine tune the thing.
One issue it wants to resolve is:
People are enjoying trying very long chat sessions can confuse the model on what questions it is answering and thus we think we may need to add a tool so you can more easily refresh the context or start from scratch.
How long? Some people are spending two hours in conference with the AI. Wow.
Microsoft Blog: The New Bing & Edge – Learning From Our First Week
But how can we break this?
Let’s remind ourselves of the Chinese joke about an alien spaceship that crashes in the middle of the country, apparently empty and presumably chock full of interesting technology. The response is different in the major cities. In Beijing, sentiment can be summed up as “How can we use this to make China more powerful?”. Shanghai has different priorities: “How can we use this to make money?” is the pressing question. Meanwhile, in Chengdu, the capital city of country’s gourmand heartland Sichuan, the question is “How can we eat this?”.
If internet users were a Chinese city, the collective question about he spaceship, like any other new shiny thing would be “How can we break this?”
Just as consumer packaged goods brands never seem to be able to resist the agency pitch to allow people to make personalised labels for their canned food and drinks and then watch as the meme-ster du jour or culture war battle of the day’s warriors move in to position to take a swing at their massive “kick me” sign: “How can this be? Our SugarChoc bar wrapper appears to be a tribute to a serial murderer!” Or “How did we not foresee that the first letter of the name of each our variety pack of cereals would be used to spell out an until this morning obscure slur against a marginalised community?”
Well, it’s Bing’s turn…
You can choose to be scared if you want to. I for one… etc. etc.
The Emperor’s New Consultants
A very popular article in the FT this week had UCL’s Professor in the Economics of Public Value, Maria Mazzucato, putting the boot into the big consultancies and the extent to which governments are overly. Reliant on them.
“Consultancies and outsourcers, Mazzucato argues, know less than they claim, cost more than they seem to, and — over the long term — prevent the public sector developing in-house capabilities.:
“To highlight the risk of consultants, her current theme, Mazzucato goes back to the Apollo space programme, where Nasa’s director of procurement in the 1960s warned that the agency was at risk of being “captured by brochuremanship”. In recent times, Covid has been a bonanza for consultants: the UK was paying Deloitte £1M a day for its work on testing and contact tracing.”
— FT: Mariana Mazzucato: ‘The McKinseys and the Deloittes Have No Expertise in the Areas That They’re Advising In’, by Henry Mance
The test-and-learn method we’re using at Brilliant Noise could vibe applied anywhere in an organisation, but we come with deep domain expertise, having worked and still working on actual marketing campaigns and operations for clients. Consultants who an do, who stick around and have skin in the game – that’s better way.
One final killer quote from Mazzucato:
Do you support Nasa going back to the Moon? I support Nasa getting back its mission-oriented policies and not simply thinking it’s there to de-risk Elon Musk.
I mean can you imagine having to de-risk that guy?
Greek deity orders Tesla recall
Imagine a powerful Greek god who blessed a mortal with fortune and insight about machinery that could match the magic of the gods. The mortal tried out his new talents and made millions as a young man, which he then parlayed into billions and started building ever more ambitious projects that would change the world: a tower that would reach the gods, ships that would sail into the heavens and a range of bland but very fast and possibly intelligent range of electric cars with iPads for dashboards. Imagine then that the God lost interest and withdrew, but the man didn’t care. He had so much money, technology, followers and – most precious of all – reputation as the greatest inventor and business man in the world. He went on doing what ever he wanted. When a single big bet on a social network everyone loved but couldn’t make profitable, went wrong he cursed the name of the god and swore he would destroy him.
In revenge the God took the man’s greatest achievement – the cars that he told people – were intelligent, and ripped away the illusion of autonomy.
In a dream, Elon (for it is he) was confronted by the tech god who thundered at him:
You have been a fool for a year and now 1,000 of your cars will be returned by customers for every day of that year (bar a couple of important Olympian public holidays, so 363 total). And if you don’t repent, all your spaceships will go on the fritz too. You didn’t listen when the hyperloop turned to shit, but you will listen now human!
“Bed wetter,” snarled the sleeping Elon, and turned over to enjoy the rest of his 4.3 hours oxygen-optimised sleep in a micro-yurt pitched in the Twitter server room.
Cut to last Thursday’s FT:
This week
Reading
I’ve started the Bernie Gunther books again from the beginning, which is nicely complementing my Babylon Berlin obsession. It must be about 25 years since a friend got me hooked on Philip Kerr’s stories of the hardboiled Berlin private eye scraping a living in the shadow of the Nazi takeover.
I’d forgotten how very Raymond Chandler-inflected the prose was (no bad thing at all, as these three samples of colourful prose show from March Violets
The voice was fastidious, suave even: soft and slow, with just a hint of cruelty. The sort of voice, I thought, that could lead you into incriminating yourself quite nicely, thank you.
At this point, the butler cruised smoothly into the room like a rubber wheel on a waxed floor and, smelling faintly of sweat and something spicy, he served the coffee, the water and his master’s brandy with the blank look of a man who changes his earplugs six times a day.
That this goddess should be married to the gnome sitting in the study was the sort of thing that bolsters your faith in Money.
Writing
Couple of things I’m working on:
Writing with AI: I’m writing a long-form, monograph type thing on writing and AI. Intention is to calm the hype and look at what’s actually going on from the perspective of a writer. Should be finished next week. Send me a note or leave a comment if you want a version 1.0 – I’m not sure where it will end up being published.
Test–Learn–Lead™ part three: Finishing the third in a series of articles about the framework and process for test-and-learn as a way of working in all of marketing (not just the A/B tests on ad campaigns and websites). Should be published on the BN blog soon. Parts one and two are already available there.
Watching
I recommended it last week, but it is so good it’s getting a second LOUDER recommendation in for form of some Gifs – early candidate for film of the year RRR (Netflix):
Other interesting things
These Startups Hope to Spray Iron Particles Above the Ocean to Fight Climate Change, by James Temple, MIT Technology Review
Straight off the pages of Neal Stephenson’s Termination Shock (see Antonyms passim) is news of a company sending tiny particles into the atmosphere to lower the planet’s temperature – yes, it’s actual real work
“Blue Dot Change hopes to determine whether the particles will accelerate the destruction of methane, one of the most powerful greenhouse gases in the atmosphere. If it works, the four-person company hopes to begin spraying the particles on commercial scales within a year after that, says David Henkel-Wallace, the founder and chief executive.”
Not very encouragingly, an expert asked for comment said:
“We have no idea what will happen there,” says Natalie Mahowald, an atmospheric scientist at Cornell and an expert on iron aerosols.
Fiction-checking update
We talked about the need to fiction-check for AI’s “hallucinated facts” last week. Professional journalists are seeing lots of slips (although none as expensive as Google’s £100BN stable last week.
In “AI-generated Writing Highlights the Need for Editing and Fact-Checking”, by Lisa McLendon of the the non-profit media institute Poynterssaid:
CNET has had to retract or significantly fix numerous AI-generated articles because of factual errors and plagiarism. Even though CNET claimed its AI articles had been fact-checked, the mistakes that got through showed that fact-checking needed to be much more thorough than it actually was, if indeed it happened.”
McLendon also notes other major publications in the report who have fallen down using cheap labour from AI to write articles without the more expensive labour of good subeditors and fact-checkers to see if any of it is true.
For more on the dangerous confusion of Men’s Health and its bots see this article by Jon Christian in Neoscope.
Grainy spy photos
I mean, you know what it’s like. You come back from a conference in a hyper-surveilled dystopian mega-resort and you’re finding cameras in your bags and pockets for weeks. Those darn grain-sized cameras just get everywhere!
Read about the size-of-grain cameras on the Princeton University website. Found via Schneier on Security blog, which notes these will be “changing the scale and scope of surveillance.” Oh my – they will won’t they?
That’s all for this week folks. Hope you enjoyed the ride! Now go and watch RRR!
Antony