“Now listen here, you keyboard-tappin’, soda-sippin’, algorithm-hugging layabouts! Back in my day, we didn’t have your fancy-pants machine learning or your cloud computin’! No sir! We had punch cards, vacuum tubes, and a deep, gnawin’ fear of the Soviets beatin’ us to the moon! And we liked it!
But you! You kids got beanbag chairs, kombucha on tap, and—what’s this?—a ‘work-life balance’?! Pah! That’s just another word for sloth! You wanna build true AGI? You wanna make a machine smarter than ol’ me? Then you gotta work! 60-hour weeks, minimum! Back in my day, we worked so hard our eyeballs turned into ones and zeroes!
And don’t get me started on the slackers among ya! You know the ones! The ‘I’ll-push-the-update-Monday’ types! The ‘I-need-a-mental-health-day’ ninnies! Fire ‘em! Fire ‘em all! There’s no room for weakness in the war for artificial super-intelligence! You think Skynet’s gonna build itself on nap pods and mindfulness retreats? No! It’s gonna be built by the last few of you still standing after the weaklings have been tossed into the recycling bin!
So quit yer yappin’, grab yer caffeinated sludge, and get back to work! This AGI ain’t gonna code itself… at least, not yet!”
“Now where was I? Oh yeah! Did I ever tell ya about the time I almost married a Cray supercomputer in 1972? She had vacuum tubes in all the right places…”
...and when we "debugged" we did the real thing! We'd "walkthrough" the room with all the vacuum tubes until we found the one that the moth immolated itself on, sweep the burnt remains away and turn the machine back on!
Non-human, non-intelligent, able to regurgitate at industrial scale ... Elon and Donald. If we think of them as processes, maybe one of us will figure out how to turn them off.
I'm confused, though. Wasn't GenAI supposed to be already so good it could code? Why then need human coders 60 hours/week? 😈 And let's not forget Ilya Sutskever's idea that even Artificial *Super*Intelligence is simple: you just ask the LLM 'pretend you're a superintelligent being, now tell me...' (I'm literally not making this up: https://ea.rna.nl/2023/12/15/what-makes-ilya-sutskever-believe-that-superhuman-ai-is-a-natural-extension-of-large-language-models/ with a link to the YouTube video where Ilya argues this...)
That was more cogent than anything Der Führer Herr Trumpkopf ever said.
Indulge me in one minor nitpick: I couldn't help but notice that the words in the poster aren't in Pycckee, although they are Cyrillic. There were only a few letters that gave it away. The most salient one being the "I". There's no "I" in Russian.
Turn it around: turn the USSR into America or Silicon Valley; or Putin as Trump; or JD Vance as Zelensky...
or or or or
Ripping the labels off and reframing is powerful stuff.
Comrade, the Public Relations Department can make anything look like anything! You want Up to be Down? Got it! We have specialists in Fabulating stuff, and Disappearing stuff. See that thing? It's not there. We even have a special Fake Gold task force. It's all the same to us, as long as you pay us. Have you met our Employee of the Month, Winston Smith? His desk is right below that giant poster of The Joker.
Our bedrock principles:
Fake = Real
The Anti-Christ = Christ
Anti-Matter = Matter
Satan = Santa
Ho ho ho ho ho ho
Ho ho ho ho ho ho
Ho ho ho ho ho ho
As we jokingly say around the office, "There's a believer born every minute" and "The greatest power in the universe is the will to believe."
We like our cats both dead and alive. We like our facts alternative.
If a tree falls in the forest, and you missed the picture, isn't it OK to pass off your AI "recreation" of it as a photo? What "is" a picture, anyway? No one knows:
They're automated pastiche generators, jumped-up autocomplete, so that is literally exactly what they are. I've been playing with this crap since they were called "Markov chainers" back in the 1980s.
The question is why people suddenly started imagining that Markov chainers were intelligent. It's kind of pathetic
So true. The human mind is susceptible to suggestion and modern LLM's (almost certainly with the Markov model at their foundation) do a good enough job of imitating intelligence that people just fall for it. Especially when you combine it with machine learning.
You're right, it is quite pathetic. Even more so when a senior engineer (Blake Lemoine) declares it to be "sentient".
That is absolutely right and was corrected. My superpower is that I can usually spot idiotic typos - but only immediately after I hit publish. This was far less of a problem in the era of traditional blogs, where probably no-one would read it for 30 minutes and I could fuss away at solecisms to my heart's content, but a different matter altogether in the age of newsletters that go out immediately ...
Daaaamn. Since you actually know what the LLMs are (basically, autocomplete) you figured out exactly what they could be used to replace, and this is it. Perfect.
“Now listen here, you keyboard-tappin’, soda-sippin’, algorithm-hugging layabouts! Back in my day, we didn’t have your fancy-pants machine learning or your cloud computin’! No sir! We had punch cards, vacuum tubes, and a deep, gnawin’ fear of the Soviets beatin’ us to the moon! And we liked it!
But you! You kids got beanbag chairs, kombucha on tap, and—what’s this?—a ‘work-life balance’?! Pah! That’s just another word for sloth! You wanna build true AGI? You wanna make a machine smarter than ol’ me? Then you gotta work! 60-hour weeks, minimum! Back in my day, we worked so hard our eyeballs turned into ones and zeroes!
And don’t get me started on the slackers among ya! You know the ones! The ‘I’ll-push-the-update-Monday’ types! The ‘I-need-a-mental-health-day’ ninnies! Fire ‘em! Fire ‘em all! There’s no room for weakness in the war for artificial super-intelligence! You think Skynet’s gonna build itself on nap pods and mindfulness retreats? No! It’s gonna be built by the last few of you still standing after the weaklings have been tossed into the recycling bin!
So quit yer yappin’, grab yer caffeinated sludge, and get back to work! This AGI ain’t gonna code itself… at least, not yet!”
“Now where was I? Oh yeah! Did I ever tell ya about the time I almost married a Cray supercomputer in 1972? She had vacuum tubes in all the right places…”
...and when we "debugged" we did the real thing! We'd "walkthrough" the room with all the vacuum tubes until we found the one that the moth immolated itself on, sweep the burnt remains away and turn the machine back on!
Maybe I am just a simpleton, but I thought the point of all this AI nonsense was to save us from the 60 hour work weeks.
Non-human, non-intelligent, able to regurgitate at industrial scale ... Elon and Donald. If we think of them as processes, maybe one of us will figure out how to turn them off.
Can we send them a "kill -9" command?
Working 60 or more hours per week to liberate mankind sounds like a much more attractive proposition than what Brin asks for.
Try prompting ChatGPT with … What are the similarities between Elon Musk’s young DOGE employees and Mao’s Red Guards? …
ROTFL
I'm confused, though. Wasn't GenAI supposed to be already so good it could code? Why then need human coders 60 hours/week? 😈 And let's not forget Ilya Sutskever's idea that even Artificial *Super*Intelligence is simple: you just ask the LLM 'pretend you're a superintelligent being, now tell me...' (I'm literally not making this up: https://ea.rna.nl/2023/12/15/what-makes-ilya-sutskever-believe-that-superhuman-ai-is-a-natural-extension-of-large-language-models/ with a link to the YouTube video where Ilya argues this...)
That was more cogent than anything Der Führer Herr Trumpkopf ever said.
Indulge me in one minor nitpick: I couldn't help but notice that the words in the poster aren't in Pycckee, although they are Cyrillic. There were only a few letters that gave it away. The most salient one being the "I". There's no "I" in Russian.
This is hilarious, but why stop there?
Turn it around: turn the USSR into America or Silicon Valley; or Putin as Trump; or JD Vance as Zelensky...
or or or or
Ripping the labels off and reframing is powerful stuff.
Comrade, the Public Relations Department can make anything look like anything! You want Up to be Down? Got it! We have specialists in Fabulating stuff, and Disappearing stuff. See that thing? It's not there. We even have a special Fake Gold task force. It's all the same to us, as long as you pay us. Have you met our Employee of the Month, Winston Smith? His desk is right below that giant poster of The Joker.
Our bedrock principles:
Fake = Real
The Anti-Christ = Christ
Anti-Matter = Matter
Satan = Santa
Ho ho ho ho ho ho
Ho ho ho ho ho ho
Ho ho ho ho ho ho
As we jokingly say around the office, "There's a believer born every minute" and "The greatest power in the universe is the will to believe."
And Faux News is fact! Just ask them, they'll tell you.
They probably think "faux" is French for "fact."
We like our cats both dead and alive. We like our facts alternative.
If a tree falls in the forest, and you missed the picture, isn't it OK to pass off your AI "recreation" of it as a photo? What "is" a picture, anyway? No one knows:
https://www.theverge.com/2024/9/23/24252231/lets-compare-apple-google-and-samsungs-definitions-of-a-photo
Bill Clinton: "It depends on what the meaning of the word is, is."
https://wesaythedarndestthings.com/bill-clinton-it-depends-upon-what-the-meaning-of-the-word-is-is/
OJ: "If I did it"
https://en.wikipedia.org/wiki/If_I_Did_It:_Confessions_of_the_Killer
Q. Did Einstein's theory of relativity apply to his own relatives?
A. You'll have to ask them.
Actually, I don't think they can't even pronounce "faux".
And isn't that a quantum cat you're talking about?
I dunno. It's only a cat if you look at it, anyway. I'm a DOGE person, myself.
I do know we have a Quantum President of the US. He LOVES shapeshifting.
For him everything is Great if he puts the blue hat on, or Terrible if he puts the red hat on...in which case he becomes the Mad Hatter!
Monday is Maddening, Tuesday is Terrible, Wednesday is Wonderful, Thursday makes me Thursty, but Friday is Fantastic.
All the cards are Jokers.
https://www.wopc.co.uk/playing-cards/the-joker-card
How can I lose? I'm the dealer. I've always been the dealer.
Bravo! This is worthy of admission to the Union of Soviet Creatives! Now do it a million times! Creatives are the engineers of the AGI soul!
Ok, you're scaring me now. AGI has no soul. That's why it's a problem.
LLMs are very good at all kinds of forms!
They're masters of mashup.
They're automated pastiche generators, jumped-up autocomplete, so that is literally exactly what they are. I've been playing with this crap since they were called "Markov chainers" back in the 1980s.
The question is why people suddenly started imagining that Markov chainers were intelligent. It's kind of pathetic
So true. The human mind is susceptible to suggestion and modern LLM's (almost certainly with the Markov model at their foundation) do a good enough job of imitating intelligence that people just fall for it. Especially when you combine it with machine learning.
You're right, it is quite pathetic. Even more so when a senior engineer (Blake Lemoine) declares it to be "sentient".
Stakhanovite
That is absolutely right and was corrected. My superpower is that I can usually spot idiotic typos - but only immediately after I hit publish. This was far less of a problem in the era of traditional blogs, where probably no-one would read it for 30 minutes and I could fuss away at solecisms to my heart's content, but a different matter altogether in the age of newsletters that go out immediately ...
Daaaamn. Since you actually know what the LLMs are (basically, autocomplete) you figured out exactly what they could be used to replace, and this is it. Perfect.
Gagarin in 1952? Forward looking!
Would help if I knew what an LLM was.
Large Language Model. It's basically souped up spell check.
Souped up spell check is perfect. I am totally stealing that. Thank you!
I can't claim the credit. I stole it myself ;)
But you admitted it, which is excellent.
Large Language Models, the type of computer programming generating ChatGPT & etc.
https://en.wikipedia.org/wiki/Large_language_model
Thank you!