I don’t hate AI. It’s a tool, I’m indifferent to it. I hate the people trying to force it on everyone and into everything, and their eagerness to fuck up huge numbers of lives to make a quick buck.
A tool has a deterministic outcome, you hammer a nail you know the outcome, you saw a board you know the outcome. No matter how many times you do the same process you know the outcome. Even dice have a boundary of outcomes you can understand.
You give an AI the same string of prompts multiple times and you will get a different outcome each time. Whether its updates to the model or the fact AI’s own response contributes to its context which small changes will build up over time, there is no way of knowing what the outcome will be, how accurate it is, or how to recreate the same results (especially with multiple prompts involved).
I wouldn’t call it a tool for that reason but maybe I am nitpicking here.
I’d say if a tool produces predictable results, even if not fully deterministic, it qualifies as a tool. It might not be right for jobs were precision is needed, but the current LLMs and GenAI are perfectly suitable for their primary purpose:
Conning idiots into trading their skills and natural intelligence for a promise of convenience, scamming managers into fucking over employees for a promise of saving money, then pulling the rug, cashing in on the desperation of those who can no longer function without it, ruining a generation of students that don’t yet have the expertise to realise the full extent of the damage they’re doing to their own skills (including, as some other post brought up, the skill to not kill people with your MedGPT malpractice), causing unpredictable damage to a host of economies and industries, fucking over residents that don’t get a (democratic) say in whether they want to have a data center chugging their water supply, fucking up the climate, fucking the whole world…
In short: LLMs and GenAI are a tool to sell our future for a quick buck we’ll never see.
I think it could be a tool. Maybe. So long as what you wanted was correct-sounding nonsense, it’s perfect for that.
Of course, when they say “AI is a tool,” what they mean is “AI is not political,” which is obviously ridiculous. Tools have never not been political, and like Icarus, their wings will burn up eventually.
Language modeling is equivalent to a dice roll (given a perfect random number generator). Setting the temperature to 0 removes all randomness from the output, meaning the model always selects the highest probability next word, and the model becomes 100% deterministic. That is, the output of a model is entirely predictable given temperature = 0, you know the model weights, and the seed/prompt.
These technicalities aside, it’s true for both a dice roll event and a specific model/prompt event that, practically speaking, the outputs are treated as probabilistic despite being mathematically/technically deterministic: a human can’t predict with 100% accuracy the output of a die despite the theory (classical mechanics of die positioning, force, velocity, friction, …) proving determinism
You know the boundaries of a dice roll based one the number of sides of the dice. You will never know the boundaries of the AI.
With a D20 I know I have a 5% chance of at a roll of 20 and 50% chance to roll over 10. With AI you don’t even know if the data it was trained on was even accurate or if it will hallucinate and speak nonsense.
You can literally ask an AI how many letter Ts are in the word colonialism and it will tell you two. Now how on earth could anyone have known its probability to say that clearly wrong answer?
Also each AI session its response to your prompts contribute to the context of the session and small alterations in how the AI speaks build up and change the outcome of a session, thus the AIs own responses effect its probabilities, another thing you cannot account for.
Yes, you do know the boundaries of AI. It is purely matrix multiplication: its output distribution is just as intelligible as the distribution of rolls of a dice. We receive a probability distribution for the next token given a sequence of tokens. This is demonstrable; search for softmax online.
To fairly equate a dice roll event to a model prompt event we must understand the technicalities. To say you have a 20 sided die, is equivalent to saying you have a specific model’s architecture and value of every parameter, in the context of qualifying event determinism.
If you can assume your die is fair, and 20 sided, that is an equivalent assumption about a model as to saying it’s llama-3.1-8B-instruct. That is, you do know the specific model weights, corresponding to a functional relationship between input and output which is deterministic. That is, if you know the model weights, which is equivalent to knowing whether a die is fair and n-sided, you can deterministically predict the output of a model as you can deterministically predict which number on a die will land
You’re making specific, technical errors about the mathematical basis of language modeling, and equating things fallaciously to a similar deterministic event.
Despite this, your intuition is right: we can’t perceptually predict the output of a model as we can’t perceptually predict what number will result from a die roll
I mean it is a tool made from stealing other people’s work.
And also we have restrictions on other tools in society. Knifes for instance. They’re pretty neat, but you wouldn’t want to be giving them to kids or pretending that it solves every problem.
How it currently exists, yes in most cases it is trained on stolen cognitive labor. Do you think this is inherent to the technology itself, however? Consider a model trained on entirely public domain data, or non-copyleft liscence not requiring attribution. E.g., talkie
Totally agree that we need strict regulation.
If only we lived in a society where people could be freely able to produce cognitive labor while also being guaranteed a dignified life with universal basic services and income, regardless of what they produce. Then, like with piracy, LLM training, in my opinion, could be trained on anything without harming original authors.
I mean that’s a wonderful “what if” situation. Universal basic income and an expanded commons.
But I’m going to pull you up on the notion of cognitive labour.
Most artists and writers are not just coming up with ideas but actually executing them.
Creating pictures, even digitally, usually requires the artist to place their body in a location for a duration and perform (often repetitive) tasks.
This is manual labour too. They chose to exchange their physical existence for the creation of this work.
The same is true of most writers too. Not just the typing up of their ideas, but frequently the physical process of acquiring the material to express those ideas. Travelling for interviews. Talking to people.
So yeah, don’t let them get away with suggesting that theft was purely “cognitive labour” because that finishes the crime.
Imagine the head of a factory declaring, “Why am I paying all these people to assemble products when I could give 20% of them hammers and fire the rest.”
I dunno about this analogy, I think if a hammer just got invented then for some trades 20% of the workforce with hammers will dramatically outperform the full workforce without.
AI is just not a hammer-calibre tool to begin with, but honesty I’d rather argue about where we go even if we imagine AI really is that useful. People being laid off en masse should be much more concerning than which technology their employers dramatically overestimed to get to that point. I think I’d be just as upset about mass layoffs in a fictional world where they were a sound business decision. I actually don’t care that those decisions will tank some business after the next quarter.
It was mostly a joke because AI is getting shoehorned into everything in ways that makes as much sense as a hammer at every station of an assembly line; at the very least, AI is not actually replacing the laid off workers who were fired for lack of projects and work.
What I also think is pretty great about the AI-backlash is that it’s effecting the educated “middle class”, which has built society around gate keeping degrees. Nobody generally gave a shit when modernization displaced factory workers (“just the price of progress!”), because working with the BRAIN is obviously morally superior (/s), but now knowledge work is effected, they’re feeling a drop of class consciousness.
I don’t hate AI. It’s a tool, I’m indifferent to it. I hate the people trying to force it on everyone and into everything, and their eagerness to fuck up huge numbers of lives to make a quick buck.
A tool has a deterministic outcome, you hammer a nail you know the outcome, you saw a board you know the outcome. No matter how many times you do the same process you know the outcome. Even dice have a boundary of outcomes you can understand.
You give an AI the same string of prompts multiple times and you will get a different outcome each time. Whether its updates to the model or the fact AI’s own response contributes to its context which small changes will build up over time, there is no way of knowing what the outcome will be, how accurate it is, or how to recreate the same results (especially with multiple prompts involved).
I wouldn’t call it a tool for that reason but maybe I am nitpicking here.
I’d say if a tool produces predictable results, even if not fully deterministic, it qualifies as a tool. It might not be right for jobs were precision is needed, but the current LLMs and GenAI are perfectly suitable for their primary purpose:
Conning idiots into trading their skills and natural intelligence for a promise of convenience, scamming managers into fucking over employees for a promise of saving money, then pulling the rug, cashing in on the desperation of those who can no longer function without it, ruining a generation of students that don’t yet have the expertise to realise the full extent of the damage they’re doing to their own skills (including, as some other post brought up, the skill to not kill people with your MedGPT malpractice), causing unpredictable damage to a host of economies and industries, fucking over residents that don’t get a (democratic) say in whether they want to have a data center chugging their water supply, fucking up the climate, fucking the whole world…
In short: LLMs and GenAI are a tool to sell our future for a quick buck we’ll never see.
I think it could be a tool. Maybe. So long as what you wanted was correct-sounding nonsense, it’s perfect for that.
Of course, when they say “AI is a tool,” what they mean is “AI is not political,” which is obviously ridiculous. Tools have never not been political, and like Icarus, their wings will burn up eventually.
Language modeling is equivalent to a dice roll (given a perfect random number generator). Setting the temperature to 0 removes all randomness from the output, meaning the model always selects the highest probability next word, and the model becomes 100% deterministic. That is, the output of a model is entirely predictable given temperature = 0, you know the model weights, and the seed/prompt.
These technicalities aside, it’s true for both a dice roll event and a specific model/prompt event that, practically speaking, the outputs are treated as probabilistic despite being mathematically/technically deterministic: a human can’t predict with 100% accuracy the output of a die despite the theory (classical mechanics of die positioning, force, velocity, friction, …) proving determinism
You know the boundaries of a dice roll based one the number of sides of the dice. You will never know the boundaries of the AI.
With a D20 I know I have a 5% chance of at a roll of 20 and 50% chance to roll over 10. With AI you don’t even know if the data it was trained on was even accurate or if it will hallucinate and speak nonsense.
You can literally ask an AI how many letter Ts are in the word colonialism and it will tell you two. Now how on earth could anyone have known its probability to say that clearly wrong answer?
Also each AI session its response to your prompts contribute to the context of the session and small alterations in how the AI speaks build up and change the outcome of a session, thus the AIs own responses effect its probabilities, another thing you cannot account for.
Yes, you do know the boundaries of AI. It is purely matrix multiplication: its output distribution is just as intelligible as the distribution of rolls of a dice. We receive a probability distribution for the next token given a sequence of tokens. This is demonstrable; search for softmax online.
To fairly equate a dice roll event to a model prompt event we must understand the technicalities. To say you have a 20 sided die, is equivalent to saying you have a specific model’s architecture and value of every parameter, in the context of qualifying event determinism.
If you can assume your die is fair, and 20 sided, that is an equivalent assumption about a model as to saying it’s llama-3.1-8B-instruct. That is, you do know the specific model weights, corresponding to a functional relationship between input and output which is deterministic. That is, if you know the model weights, which is equivalent to knowing whether a die is fair and n-sided, you can deterministically predict the output of a model as you can deterministically predict which number on a die will land
You’re making specific, technical errors about the mathematical basis of language modeling, and equating things fallaciously to a similar deterministic event.
Despite this, your intuition is right: we can’t perceptually predict the output of a model as we can’t perceptually predict what number will result from a die roll
Yeah, sort of.
I mean it is a tool made from stealing other people’s work.
And also we have restrictions on other tools in society. Knifes for instance. They’re pretty neat, but you wouldn’t want to be giving them to kids or pretending that it solves every problem.
Sort of like cigarettes in the 1950s.
Or asbestos.
How it currently exists, yes in most cases it is trained on stolen cognitive labor. Do you think this is inherent to the technology itself, however? Consider a model trained on entirely public domain data, or non-copyleft liscence not requiring attribution. E.g., talkie
Totally agree that we need strict regulation.
If only we lived in a society where people could be freely able to produce cognitive labor while also being guaranteed a dignified life with universal basic services and income, regardless of what they produce. Then, like with piracy, LLM training, in my opinion, could be trained on anything without harming original authors.
I mean that’s a wonderful “what if” situation. Universal basic income and an expanded commons.
But I’m going to pull you up on the notion of cognitive labour.
Most artists and writers are not just coming up with ideas but actually executing them.
Creating pictures, even digitally, usually requires the artist to place their body in a location for a duration and perform (often repetitive) tasks.
This is manual labour too. They chose to exchange their physical existence for the creation of this work.
The same is true of most writers too. Not just the typing up of their ideas, but frequently the physical process of acquiring the material to express those ideas. Travelling for interviews. Talking to people.
So yeah, don’t let them get away with suggesting that theft was purely “cognitive labour” because that finishes the crime.
Without a doubt! LLMs are a fantastic tool. But they’re just a tool! They makes some tasks easier.
They’re not the mark of the beast.
They’re not Skynet.
They’re not even “artificial intelligence”. That’s all marketing.
Once this bubble bursts, humanity will be left with some great tools to move forward with.
I would hate a hammer if everyone I cannot send dickwise demanded I perform brain surgery with it.
And that’s exactly where AI fits currently.
Imagine the head of a factory declaring, “Why am I paying all these people to assemble products when I could give 20% of them hammers and fire the rest.”
I dunno about this analogy, I think if a hammer just got invented then for some trades 20% of the workforce with hammers will dramatically outperform the full workforce without.
AI is just not a hammer-calibre tool to begin with, but honesty I’d rather argue about where we go even if we imagine AI really is that useful. People being laid off en masse should be much more concerning than which technology their employers dramatically overestimed to get to that point. I think I’d be just as upset about mass layoffs in a fictional world where they were a sound business decision. I actually don’t care that those decisions will tank some business after the next quarter.
It was mostly a joke because AI is getting shoehorned into everything in ways that makes as much sense as a hammer at every station of an assembly line; at the very least, AI is not actually replacing the laid off workers who were fired for lack of projects and work.
What I also think is pretty great about the AI-backlash is that it’s effecting the educated “middle class”, which has built society around gate keeping degrees. Nobody generally gave a shit when modernization displaced factory workers (“just the price of progress!”), because working with the BRAIN is obviously morally superior (/s), but now knowledge work is effected, they’re feeling a drop of class consciousness.
I would agree with you if it wasn’t so destructive to the environment
Generative “AI” sucks.
this… tool is (kinda) ok, sucks and it’s wrong a lot but that’s besides the point.
the tool should be accessable, but NOT forced…