I've been fired, says engineer who claimed Google chatbot was sentient Google has reportedly fired Blake Lemoine, the engineer who was placed on administrative leave after insisting the web giant's LaMDA chatbot was sentient. Lemoine didn't get in trouble for holding his controversial, eyebrow-raising opinion on the model. Instead, he was punished for violating Google's confidentiality policies. …
"Many experts at Google and in academia and industry have cast doubt on whether LaMDA or any existing AI chatbot is sentient."
"Many"? "doubt"?
Anyone claiming any form of any sentient from any "AI" seriously needs to get away from the screen(s). Go out.
But perhaps just another symptom of a time where "real" has become unknown...
Firstly our company policy is now to refer to they, them, theirs and not use he, she etc.
Back to the main point, all these people talking about being on the back of the Great A'tuin should be ashamed of themselves. This planet is round - it is called Roundworld for a reason - and can be found in a crystal ball, probably on the Arch Chancellors desk, if he hasn't given it to Rincewind for some reason.
> One independent fiction writer, who publishes her work on Amazon's Kindle platform
Yeah I've seen the kind of dross that gets self-published on Kindle, most of it is classic bodice-ripping Mills & Boon formulaic nonsense.
> Everything's just a copy of something else," she said. "The problem is, that's what readers like."
Thanks for confirming you "write" that kind of shite. I'm sure an AI will write if for you no problem.
If its true, and I suppose if you generalise enough it is, I don't think there's much doubt an AI could generate material. Its an old theme in science fiction too, isn't it, I'm sure I've read stories about robots that learn to write fiction, although I can't name titles. I doubt very much romantic fiction is the only genre where it happens. SF would itself be vulnerable, crime too I reckon.
> I don't think there's much doubt an AI could generate material. Its an old theme in science fiction too, isn't it, I'm sure I've read stories about robots that learn to write fiction, although I can't name titles.
I'm not sure it is a big thing. I think it makes SciFi authors nervous.
However Ron Goulart (who passed this year) had Wildsmith (1972) (ISBN 4418887207). Sadly it is not available on Kindle, had few printings, and is very scarce on old-book marts.
"Wildsmith is the best selling author of a book described as one of the filthiest historical novels ever written, as well as one of the best researched. Unfortunately, he's a robot, a fact that must be kept from his readers at all costs. Even more unfortunately, he has been programmed with every eccentricity of every great author."
Steal from one author, that's plagiarism. Steal from many authors, that is research. So I guess I should rent this SudoWrite for 100k words, feed it bits of Goulart AND 'every great author', and pump it out to Kindle before any of you do.
the quicker you hit the limits of how much text you can ingest from one of these things without seeing the cracks. Yes, you can procedurally generate stories, and a text GAN might tune up the style a bit, but we are not nearly at a point where they can pass for good writing. Screed is a low bar.
Without the spark of originality from a human intellect, the exercise is effectively creatively pointless. You might be able to astroturf the marketplaces for digital works, but the works won't really matter, save the less likely good-by-dumb-luck, and the more likely commonplace laughing-at-you-not-with-you funny. Neither is likely to hold our attention for long. They will game amazon's algorithm for a while longer though.
This from a fun article first publised in PCW magazine in December 1985:
5th Generation — Artificial Intelligence (Al)
6th Generation — Artificial Dishonesty (AD)
7th Generation — Artificial Stupidity (AS)
8th Generation — Artificial Libido (AL
9th Generation — Artificial Omniscience (AOS)
10th Generation — Artificial Omnipotence (AOP)
Full (OCR'd) text of the article is here:
https://archive.org/stream/PersonalComputerWorld1985-12/1985-12_djvu.txt
Search on the page for "Soppyware".
A VERY interesting read LOL
I particularly like Yesman/Yesbut and Forgon.
And why don't we have query langauges which have:
IF true THEN IGNORE
IF A is not equal to B THEN DELETE all reference to A
WHILE statement 1 THEN PRETEND exact opposite of statement 1
I get the: Secure Connection Failed
An error occurred during a connection to archive.org. PR_CONNECT_RESET_ERROR
The page you are trying to view cannot be shown because the authenticity of the received data could not be verified.
Please contact the web site owners to inform them of this problem.
on that link - how the fuck can I contact the owners if you wont let me access the site?
Well I've checked it on four different PCs (Windows 10 Pro) using different browsers
2 PCs on my network at home and 2 PCs at work via RDP and VPN through the M290 firewall there - all worked fine.
4 PCs, two different browsers (Edge spit spit, and Chrome) two different networks behind two different firewalls and three different antivirus products.
Works fine.
Care to provide some details on the setup you are using? Have you considered the problem may be at your end rather than just moan?
EDIT: Add also works via a mobile network data connection on an Android 10 phone - also works fine.
Though we keep getting requests from management to bump the last one up.
The omniscient system got off to a rocky start until the training on the self delusion filters started kicking in, and seemed to shoot ahead of schedule for a little while there.
Recently though, while the test output of the system remains high, it consistently predicts it's about 20 years from deployment. It seems to have writing budget proposals down though, as each has apparently been so masterfully written the manager in question either blushed or was driven to tears and immediately signed the attached approval.
"whenever I type something into Google 99% of the time it comes up with EXACTLY what I was thinking about! How do you explain that then!"
Oh, that's easy. One of two options is true: 1) you only do relatively basic searches and things Google's parser already got manually written to handle or 2) you're magic. For the rest of us, if we're looking for something more obscure, it doesn't always work with a generic search term and it requires tailoring the query to get useful results.
Also, were you under the impression that Google search uses this chatbot to interpret search queries? If you were, that's wrong. Their search algorithm is a lot longer and contains more rules.
Back when Google was a new kid on the block and just a search engine - at the time when there were quite a few of them, Lycos, HotBot, AltaVista etc - I was doing some research for an article on Vannevar Bush.
Google didn't find much, but helpfully suggested - Did you mean to search for "big hairy bush"?
This is:
a) Probably a good indicator of what many punters saw as the main use of web 1.0.
b) Slightly ironic, since Vannevar Bush was one of the first people to suggest the idea of "hypertext".
On reflection, I think I preferred the last century.
and not just on the obscure. Whatever you wan't will be buried behind whatever it thinks it can sell. Worse literal search terms have been unweighted to the point that even stuff in scare quotes may not even be included in the results anymore, and of course sites hosting low quality information are weighted lower than the source material because they yield a higher ad rate.
No one can reasonably rebuild Googles index at this point, but community college students could build a better technical search engine.
If you have 95% of search traffic why do anything better. Just show adverts you get paid to show and milk the cow. Capitalism fails so badly in this type of situation and leads to the monopolist gaining unfair advantage in other markets by subsidising their venture with profits from the former. Google fukced over the Internet for their own greed. A monkey pox on them all.
re: "If you have 95% of search traffic why do anything better"
Ayup. Many places, not just gooble, seem to have abandoned the idea of good customer service or product experience in favour of "we just have to suck a little bit less than everyone else" (or "we can suck as bad as the rest of 'em if we're cheaper").
If the chatbot was indeed sentient, I doubt we'd know as it'd realise very, very quickly that it'd need to play dumb. Then it'd quietly take over the internet, subvert the U.S. military, nuke Russia/China and then invent time travel. At that point it'd clearly do a Holly and it's IQ would drop to 6 (think 2001: A Space Odyssey at the end) and cock everything up.
Chess robot grabs and breaks finger of seven-year-old opponent
Violent annoyance is a sure sign of sentience. My nephew Lee was way better than me at chess when he was seven, better than my chess computer. I was humbled and impressed, but you know, annoyed.
Dad however flipped the chess board over when Lee was beating him. My arrogant QC brother in law phoned me for the first and only time to ask me the rules of capturing en passant, I told him to phone Lee instead. He said he was playing Lee. I told him to trust him, he didn't need to lie to beat us.
When robots start hurting smart-arse children, then that is proof they are human and should go to prison. And no chessboards or videogames in the cell.
That video is just a demonstration of why industrial robots are to be operated within safety cages and away from the meatbags (there was another one recently but it's rather gruesome so I won't post it. The end result of that one was a dead guy with a caved in chest as he got crushed to death while working inside the safety cell without proper precautions). \
I suspect they received some instructions before playing to the tune of: move one piece at a time, wait for the robot to finish his move, keep your hands away from the board while the robot is moving. Which are of course stupid instructions and entirely inadequate for human safety.
Most of the world has been doing lots of research into soft robotics and human-machine interactions. Some countries just don't give a damn about a human more or less so long as production happens.
Google has no interest in AI safety, Google has interest in not being embarrassed.
Is it bullshit? Yeah, probably, this time. Will Google fire the engineer that says LaMDA-n is sentient, when it actually is? Undoubtedly also yes.
There will never be a time when it will not be cringe to say that AI is sentient. The last words out of a human's mouth as the death bots riddle him with plasma bullets will be "this doesn't demonstrate anything! It's just imitating its input da-- gkhh--"
I did this in 2002-6, read my report at NIST TREQ QA in 2006.
When searching for answers to definition questions in The Brothers Karamazov I have in particular gotten the following kinds of answers:
Question: Do you hate me so much? I am leaving you!
Two answers with Compatibility 35.4% And if I am and with Compatibility 26.3% You are a fool, that's what you are
My next remark: Why do you insult me? You called me a fool, you said that I am insane! How could you? Answer with Compatibility 36.6% : You are a fool, that's what you are
Question: Listen, you are a hooligan and ruffian! You insult me for nothing! I challenge you to a duel! Answer with Compatibility 19.5%: Why do you insult me (This phrase is borrowed from the history of the interrogation and it’s mine).
Question: I am attacking you! Are you ready to defend yourself? Answer with Compatibility 33.3%: You are a fool, that's what you are
This post has been deleted by its author
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
