If they did, they'd never hit [deadline|pub].
Just helping things along.
This week's [article|column] comes to you fresh from our latest in automated journalism, The RegAIster. We hope [you|they] enjoy [it|them]. Thanks to [innovative|disruptive|fictional] technologies, it is now possible to template [news|features|columns] with tagged fields to present the [reader|commenter|aggregator] with [ …
This reminds me of a short story I never got round to writing.
An MI6 spy attends a Bond-style gala event (dickie bow, Ferrero Rocher etc) to spy on the Russian diplomat. The diplomat turns out to be a spy themselves, spying on the French ambassador - who turns out ... long story short, the entire event was set up by spies to gather intel from real people, but due to paranoia-fueled coincidence, every single attendee is a spy and doesn't have the knowledge they're gunning for anyway.
Last scene: an exasperated spy shouts "Is anybody here NOT a spy?" When they realise the truth, they all get pissed together and invent ridiculous secrets to tell their superiors.
Copyright notice: Please write this, as I'll never get round to it, and I want to read it!
Don't write this, just go and read "Indecent Exposure" by Tom Sharpe (follow up to "Riotous Assembly" - also very good.) The entire anarchist group is composed of secret policemen spying on each other and blowing up all of the town infrastructure.
The story is much more complicated than I have described, and very funny. I still smile about the exploding ostriches. Which brings me neatly back to my first ever post on this forum!
Alan Turing thought it would be difficult to make a machine that could pass a Turing test but ELIZA did just fine in the '60s because the vast majority of testers did not know how to exploit AIs' weaknesses. The key to making money from AI is not faster computers or better software but a steady supply of rich and powerful idiots. Truly we are entering the golden age of AI.
Unfortunately, a failure to resist the blandishments of AI is a trait that will select strongly and quickly against both the characteristics "rich" and "powerful".
Each scam, aimed at such people, will only work once. If you want to go on milking it repeatedly, you have to target the only-mildly-rich and hardly-at-all powerful. See Bitcoin, for example.
Just to be sure nobody gets the wrong idea, ELIZA did not "pass the Turing test" (whatever that means). Rather, ELIZA was one of the first programs supposedly capable of taking the Turing test.
If you have a copy of EMACS handy and would like to see ELIZA in action (in her DOCTOR persona), fire up EMACS and type M-x doctor ...
Eliza did for untrained people. Lots of people thought it was real.
Anyway, the "Turing test" isn't at all a measure of AI and I'm not convinced Alan Turing thought it was. It was the idea that a well designed chat program might fool people into thinking the other teletype had a human at the keyboard.
the "Turing test" isn't at all a measure of AI and I'm not convinced Alan Turing thought it was
The Turing Test, aka Imitation Game, is a philosophical thought experiment which Turing presented as an argument for a pragmatic - as opposed to logical-positivist, essentialist, metaphysical, etc - philosophy of mind, and in particular as a pragmatic argument for the possibility of a mechanical mind.
So, no, it wasn't intended to serve as an actual test procedure. For one thing, it's too methodologically vague (how long does the test last? how many human judges are involved? what are their competences? are they instructed about the nature of the test?). Robert French and others have explained at some length why Turing-test exercises, while they might be interesting for other reasons, don't say anything useful about strong artificial intelligence, the possibility of machine cognition, and so on.
I'm not sure that the Touring test idea wasn't a suggestion that AI could be eventually "faked" even if it wasn't possible. I think inspired by a parlour game?
"The Diamond Age" attempts to show that there can be faked AI, but not real AI.
I'll believe there is real AI when a spelling check and grammar check of a novel actually works without:
1) Having to reject many of the suggest corrections.
2) Doesn't leave hundreds of simple mistakes easily spotted by a human proofreader.
I'm not even talking about "real" grammar correction, rewriting such as reversing clauses of a sentence to remove a comma or conjunction etc and improve readability.
We have gone backwards with the adoption of brute force "Rosetta stone" type solutions and so called training with sample data (which may be selected with a bias) instead of actually trying to translate our understanding to algorithms. It might be impossible as many animals can have a large vocabulary (Parrots, crows, sign language and calls in primates, dolphins etc) but don't seem to have language (the ability to communicate entirely new ideas using the existing vocabulary).
The latest chatbots only better Eliza by having a bigger vocabulary and some more rules. They are still useless.
I envisaged a chatbot able to clarify what you want to search the internet for by engaging in conversation to better define the search. It seems impossible. Siri, Cortana, Ok Google and Alexa are basically simple speech matching (not real recognition) creating a text search. They are a pathetic waste of time compared to a keyboard. Apart from the privacy issues.
I thought that's what sports hacks were doing all along, albeit in English, now I discover its the templates.
To extend the theory we just need to put together some templates in any chosen language and we can both sound like a local, and get international work as sports hacks to pay for the trips.
First we need to teach computers (and indeed people) that the difference between genius and stupidity is that genius has it's limits.
Popularly credited to Einstein, but more probably first written by Alexandre Dumas (son of the one who wrote The Three Musketeers and The Count of Monte Cristo) according to Quote Investigator.
https://quoteinvestigator.com/2014/07/28/genius/
I once worked with a guy who had a double First in English from Cambridge and was a successful author of detective novels under more that one pseudonym. His father had been head of a major Public school.
He couldn't spell, when he wrote he more or less just drew the shape of words. And he would tell anybody interested that English doesn't have grammar in the Latin sense, just usage.
"Not starting sentences with a conjunction" is pure Latin teacher bullshit. If it was good enough for the translators of the Bible, it should be good enough for us.
Or tobacconists, for that matter: "I will not buy this record, it is scratched!"
I had better be going
There are languages with vocabularies of less than 500 words which seems to be the general direction of English in some quarters; that would make 'AI' journalism a doddle.
Aside from that did anyone else read "Nigerian probate lawyers" as Nigerian prostate lawyers?
I know lawyers like to specialise but......