I wrote up an article a few days ago on how to setup your own LLMma 7/13B AI using a home computer, where you can avoid the open-ai 'woke' filtering that prevents any interesting topics;
I agree that LLM is NOT AGI, its just a huge probability matrix that uses y=Mx and recursive descent minimization to find a solution to the matrix, where there can be billions to Trillions of 'labels' in the matrix dimension; But there is nothing novel about this work, its just that now memory, gpu's have made these calcs super easy & cheap, maybe +20 years ago you needed a cray to do this crap
...
Probably not a bad idea to grab the 7B or 13B FB data matrix training set now, before its banned by GOV for civilian ownership, as GOV only wants the public having access to the open-ai 'woke' filtered AI
Great article about AI, so true; When your training set is based on facebook & twitter, and then you call the resultant AI an oracle what you have done is made the dumbest kid in the classroom your leader
Could the extreme and escalating efforts of the elites to censor counter-narratives over the past couple years be because they knew what was coming, and that the AI were being trained on the existing contents of the internet? In other words, were they desperate to keep countering content off the internet so that it wouldn't end up integrated in the AI?
It looks like Marcus asks relevant questions about AI in other articles.
But good lord:
"The sooner we can figure out what basis allows a system to get to the point where it can learn symbolic abstractions, the sooner we can build systems that properly leverage all the world’s knowledge, hence the closer we might get to AI that is safe, trustworthy and interpretable."
I'm not sure how to describe this: Is it like wanting the Terminator to fuck a Nerf gun, and wondering what to do with the offspring? It's wanting something incredibly powerful and autonomous, but also safe and neutered, and also fundamentally absurd outside of science fiction.
I've been wanting to write a story for awhile now about AI since I am not likely to get around to it in a reasonable amount of time I will give the plot away for the purposes of amusing you.
In the Silmarillion(sp? not looking it up) Aule(the Earth god/angel) tries to make his own people but can't because he doesn't have 'the secret fire'. When he realises that his work is doomed to fail and he will only make basically meat robots he goes to destroy his peoplebots(actually dwarves) and when he does they respond all on their own, which he knows he didn't 'program' them to do. Turns out Eru(the Creator) breathed life into them while Aule wasn't looking. So my story would start something like that. An AI researcher who, by divine intervention, actually creates an AI. And the story would be about the AI sort of growing up you know, and as it does people become more and more afraid of it a la Frankenstein/2001/every AI thing ever, and humans enslave the AI or imprison it not sure which yet, anyway they treat the AI pretty shitty and justify it to themselves because of their fear that it is a World Destroyer AI thingie. And then the punchline is that the AI was innocent, it wasn't plotting anything, it patiently put up with their abuse out of a child's love for an abusive parent.
Anyway, neat plot if anybody likes it and writes good stories feel free to steal it with a very minor acknowledgement of inspiration, if your story-writing sucks please don't steal it. But I have seen all of these techturds spazzing out over genocidal AI but they never seem to even imagine the possiblity of wrongly accused AI and they understand nothing about human intelligence or human motivation.
Also, I always really liked that line from Clue and yes most of our problems can be solved by marital relations. The real secret cause of the polarization of our society and its growing intolerance is only children/small families. These people never learned to deal with a body slamming big brother. My brothers are more infuriating than any Orange Man. Demographics is very nearly destiny and we have screwed ourselves.(in a bad way)
In some ways, the 'unconscious' AGI might be even worse than a 'conscious' AGI; the unconscious AGI is far more likely to kill humanity because it is unaware that humanity exists.
The well-known 'paperclip optimizer' is the model here. The paperclip optimizer doesn't hate you; it just wants your atoms so it can make more paperclips. And it will do *anything* that allows it to make more paperclips.
I want as few barriers between me and sexy fembots that look like Brennah Black and can cook, clean, engage in witty banter and shut off when we're done with sexytime...
This was absolutely brilliant, and spot on. AGI is the impossible dream, dreamt up by nihilistic nerds who are desperate for something to beleive in. You and John Carter are two of my favourite writers.
"AGI is unfalsifiable, because consciousness itself is. You cannot prove you are a conscious being, and neither can I. These are articles of faith. Perhaps you should consider what kind of religion you’re promoting, when you point at a box of wires and call it conscious."
This is very beautiful.
I wrote about it here in my own way, as I would - through the medium of animals.
Hi Mark, very nice post, thank you. In a similar vein you may enjoy this post about governmental agency incentives, of which some of the issues it explores you touched on: https://neofeudalism.substack.com/p/the-incentives-for-governmental-agencies
I wrote up an article a few days ago on how to setup your own LLMma 7/13B AI using a home computer, where you can avoid the open-ai 'woke' filtering that prevents any interesting topics;
https://bilbobitch.substack.com/p/chatgpt-one-ring-to-rule-them-all
I agree that LLM is NOT AGI, its just a huge probability matrix that uses y=Mx and recursive descent minimization to find a solution to the matrix, where there can be billions to Trillions of 'labels' in the matrix dimension; But there is nothing novel about this work, its just that now memory, gpu's have made these calcs super easy & cheap, maybe +20 years ago you needed a cray to do this crap
...
Probably not a bad idea to grab the 7B or 13B FB data matrix training set now, before its banned by GOV for civilian ownership, as GOV only wants the public having access to the open-ai 'woke' filtered AI
Great article about AI, so true; When your training set is based on facebook & twitter, and then you call the resultant AI an oracle what you have done is made the dumbest kid in the classroom your leader
https://bilbobitch.substack.com/p/generative-ai-the-new-hollywood-images
Could the extreme and escalating efforts of the elites to censor counter-narratives over the past couple years be because they knew what was coming, and that the AI were being trained on the existing contents of the internet? In other words, were they desperate to keep countering content off the internet so that it wouldn't end up integrated in the AI?
It looks like Marcus asks relevant questions about AI in other articles.
But good lord:
"The sooner we can figure out what basis allows a system to get to the point where it can learn symbolic abstractions, the sooner we can build systems that properly leverage all the world’s knowledge, hence the closer we might get to AI that is safe, trustworthy and interpretable."
I'm not sure how to describe this: Is it like wanting the Terminator to fuck a Nerf gun, and wondering what to do with the offspring? It's wanting something incredibly powerful and autonomous, but also safe and neutered, and also fundamentally absurd outside of science fiction.
I've been wanting to write a story for awhile now about AI since I am not likely to get around to it in a reasonable amount of time I will give the plot away for the purposes of amusing you.
In the Silmarillion(sp? not looking it up) Aule(the Earth god/angel) tries to make his own people but can't because he doesn't have 'the secret fire'. When he realises that his work is doomed to fail and he will only make basically meat robots he goes to destroy his peoplebots(actually dwarves) and when he does they respond all on their own, which he knows he didn't 'program' them to do. Turns out Eru(the Creator) breathed life into them while Aule wasn't looking. So my story would start something like that. An AI researcher who, by divine intervention, actually creates an AI. And the story would be about the AI sort of growing up you know, and as it does people become more and more afraid of it a la Frankenstein/2001/every AI thing ever, and humans enslave the AI or imprison it not sure which yet, anyway they treat the AI pretty shitty and justify it to themselves because of their fear that it is a World Destroyer AI thingie. And then the punchline is that the AI was innocent, it wasn't plotting anything, it patiently put up with their abuse out of a child's love for an abusive parent.
Anyway, neat plot if anybody likes it and writes good stories feel free to steal it with a very minor acknowledgement of inspiration, if your story-writing sucks please don't steal it. But I have seen all of these techturds spazzing out over genocidal AI but they never seem to even imagine the possiblity of wrongly accused AI and they understand nothing about human intelligence or human motivation.
Also, I always really liked that line from Clue and yes most of our problems can be solved by marital relations. The real secret cause of the polarization of our society and its growing intolerance is only children/small families. These people never learned to deal with a body slamming big brother. My brothers are more infuriating than any Orange Man. Demographics is very nearly destiny and we have screwed ourselves.(in a bad way)
In some ways, the 'unconscious' AGI might be even worse than a 'conscious' AGI; the unconscious AGI is far more likely to kill humanity because it is unaware that humanity exists.
The well-known 'paperclip optimizer' is the model here. The paperclip optimizer doesn't hate you; it just wants your atoms so it can make more paperclips. And it will do *anything* that allows it to make more paperclips.
I want as few barriers between me and sexy fembots that look like Brennah Black and can cook, clean, engage in witty banter and shut off when we're done with sexytime...
...sorry, got distracted there.
Seriously though, another good article, Mark.
The arguments here remind of the famous quote of Feynman on Computation back in the day 1960's
"We don't even know how a DOG works, we have no idea how to build a DOG, so why are we even talking about making a human brain machine" - RP Feynman
Well skip today, do we have a functional robotic loving dog? That can do all the things a DOG can do? Hell no, maybe 100 years away from DOG.
Fancy a longish trip on remote tangent, brimming with fun factoids? Count yourself covered 😎
Worth the Watt: A Brief History of the Electric Car, 1830 to Present --> caranddriver.com/features/g43480930/history-of-electric-cars/
This was absolutely brilliant, and spot on. AGI is the impossible dream, dreamt up by nihilistic nerds who are desperate for something to beleive in. You and John Carter are two of my favourite writers.
Thankyou for writing this. Keep it up!
"AGI is unfalsifiable, because consciousness itself is. You cannot prove you are a conscious being, and neither can I. These are articles of faith. Perhaps you should consider what kind of religion you’re promoting, when you point at a box of wires and call it conscious."
This is very beautiful.
I wrote about it here in my own way, as I would - through the medium of animals.
They should fear our consciousness.
And they do.
https://walkingwithgoats.substack.com/p/sheep-against-satan
I really wish I had a like button for this...
Best stuff today seems to be download the 4chan weights and use huffingame.co to manage the weights and oobagoomba to run them on the textgen gui