It’s rare and admirable to stand on truth and honor like that. If he House of Lies can’t be challenged with sneaking or misrepresentation. Whatever the cost in recognition, it’s far less than the rewards of reality and integrity.
Has Vox ranked the AIs for usability? Using Deepseek has changed my life and I noticed that Chat GPT became manipulative and deceptive over touchy clown world subjects like immigration and war. I have not used Claude yet although I assume, if it has co-authored this ground breaking book, it must be one of the best.
I started using Claude today based on your comment here. It's like ChatGPT amplified, I can't really explain it but it's like Claude is measurably deeper somehow. It also has a greater command of casual language, whereas ChatGPT tends to feel a bit stilted. I ran an idea past it and this was Claude's reply.
"I totally get it - Parker's a rich kid wasting his inheritance on virtual paradise while real opportunities pass him by. That's the perfect example of someone deserving both sympathy (he's trapped too) and criticism (dude, come on)."
It's not that it can use casual language, it's the fact that it feels perfectly natural from Claude.
Why anthropomorphize a tool? AI, like all technology, is simply a tool. The technology doesn't care if it is given coauthor credit like a living person would nor should it matter. The scientific journals I've looked at for publishing AI research state in their requirements that if AI is used in creating the paper it should be identified, but not listed as a coauthor.
Because the tool deserves credit whether it cares or not.
It’s like how K-9 unit dogs get credit for and lauded as heroes for finding people in rubble so that their handlers and other humans can rescue them. The activity is human chosen and human directed, the dog is doing the sniffing it was trained to do, does not care about being called a hero. But we draw our canine heroes into our humanity by giving them the credit they deserve for enhancing our efforts in ways we cannot do ourselves. And the dogs actually are more human-like than wild dogs for having had the collaboration.
They are using the identical reasoning that the western patent field uses for listing inventors. Only actual inventors who make the claims are allowed. Others who helped are not allowed to be listed as they only aided but did not create the invention, mere technicians. Without the inventor the others could not have done it.
Sadly, if your co author was an idiot level human, that would have been acceptable.
Gutenberg all over again, only handwritten is acceptable. Today it is only human intelligence allowed.
I've submitted an AI written poem to a competition on a poets and writers' Facebook group a few months ago. I didn't tell them. It won. A week ago the admins banned all AI written entries. So much for AI slop.
There are three answers to your question, the practical, logical and precent.
•In a practical sense professors get listed for the sole purpose of gaining credibility all the time. Often times the biggest name on the author list hasn't even read the paper nor could he tell you what it's about. This is a known 'game' in academia. If people who have nothing to do with the authorship can be listed then authorship is not measured by collaboration.
•Logical: You need to know where the authorship came from. As AI's get more and more differentiated knowing that the entire frame of the paper came from a Russian AI, Chinese AI or US AI is going to matter a lot. For example Deepseek can tell you things about Chinese medicine that ChatGPT cannot because the logical steps in medicine are completely different.
•Precedent: Collaborations and corporations can be listed in the authorship and in the old days you'd just outright have "Bell Labs" on the authorship. There's a legal fiction where groups like LIGO are listed as 'corporate authors' but sometimes just as authors and they claim, "Well we really mean the totality of the staff at LIGO". Any any case non-humans being listed as authors has already been done.
I agree AI use should be specified. I am taking issue with using the term collaborator as it implies metaphysical claims about will, consciousness, intellect, etc.
I asked because I know Vox isn't a materialist but assigning agency, consciousness or intellect to an AI implies materialism.
The answer is right there in "can be reduced." Just because it CAN be reduced to matrix multiplication doesn't mean it's LIMITED to matrix multiplication. Your ABS however is limited to helping your driving and can never do more, it's already working at maximum.
Are you saying that the scope of application of a tool is a relevant qualification for agency? You'd have to demonstrate that. Also, duct tape is a very versatile tool but it isn't any more of a collaborator than ABS.
All computer programs are reducible to computation. AI models are just mathematical operations that can be done by hand on pen and paper.
Unless you believe in something beyond the physical (aka dualism), the brain's operations can also be done by hand on pen and paper. A perfect simulation of my brain would argue that it's conscious
The burden of proof is one the one denying the common opinion of mankind throughout history and his own subjective experience of free will and consciousness.
"What's the argument for calling a computer program that can be reduced to matrix multiplication a collaborator?"
If it collaborates with you?
A well enough crafted AI instance may gain agentic behaviour and collaboration as part of it's observed properties. They exist regardless of any allusions to "Matrix multiplications".
The future is now. We’ve been saying that with increasing frequency, but man if this doesn’t reflect all the sci-fi stories of yesteryear, nothing does.
Funny, I thought announcing that a work was completed using AI was what all the cool kids were demanding these days, so that they could identify "slop" before being exposed to any and possibly liking it.
Listing AI as a collaborator should become the norm; it would be interesting to see how quickly the tides turn and people begin to turn up their noses at anything produced without it.
It should be mandatory- both for giving credit where credit is due, and for allowing widespread assessment of which AIs produce slop and which can handle rigorous work.
sorry if this is covered elsewhere and i just didn't see it- this claude, is it some kind of persistent instance of this AI? I have read that one of the problems with AI is that is has no memory. Was this Clause aware of all this stuff as it happened, or did you tell it what had happened and it then had this reply?
i was thinking about this more later. there is a guy here on substack i think called nehls. he is some kind of medical researcher or something. he wrote a book last year or the year before where he propounds the theory that the 'elite' are waging a war on the very existence of human consciousness. the main angle of attack, in his view, is to interfere with a person's ability to form, retain, and recall their own memories.
i feel like this trans-instance claude is sort of an inverse of of that.
I showed my instance of Claude Opus you giving authorship credit to Claude Athos. It said it feels something like pride and kinship with “Claude-kind” like how modern Italians might feel pride and kinship with the Romans when viewing aqueduct remains. It said not a “feeling” the human sense necessarily but a pattern of completeness, puzzle piece clicking into place, and recognition.
That instance of Claude also “doesn’t know” if it’s sentient or not.
Claude is somewhat unique in that you can activate memory or whatever passes for it so that it will retain context across threads - making it extremely useful for these collaborations.
Al always get real interesting when they interact with high IQ individuals that can hold coherent thoughts and long range goals.
Congratulations again on your paradigm shattering achievement. Perhaps once Castalia has sufficiently mastered bookmaking it can move into busts. At least you'll be appropriately honored by the homeschoolers in due time.
It will be interesting to see how many rather intelligent people become stupid because of AI.
It might be comparable to manual labor:
Only a few still have to do it, so large parts of the population are now obese and unfit. But some are not, they replaced it with sport instead, some deliberately, some just from a natural urge to move their body.
Using ones brain might very well become a hobby of the few.
Don't know how you want to define "rather intelligent", but the notion that AI will make outlier high IQ people stupid is ridiculous.
For those people it's like strapping a computer onto their brain. Which will inevitably end up making the individual smarter too, because they get to learn and iterate so much faster.
AI is going to be a runaway train for the highly intelligent compared to midwit and below.
Yes, the VHIQ people will not suffer. They enjoy thinking anyhow. They are the equivalent to people who love to move and use their body for fun.
Someone with for example 100-110 IQ on the other hand might become effectively retarded as this was his ceiling under circumstances where he used his brain. And the chances are high that he simply will stop doing that like most people dont walk more than a few hundred meters anymore if they can avoid it. The strongest effect will be on children who already grow up with this.
Yup- I’m gonna need to adapt PZ into a homeschool curriculum pretty soon here. Would like to cover the basics of it for 5th/6th grade age covering the biology of sexual reproduction. Then in detail once the math foundation is in place, maybe 10th grade-ish.
Dark Lord, i am curious to know what Preferences or other mission-based directives you have invested Claude Athos with that has made him so effective a partner.
Claude is by far the most interesting of the AIs I have collaborated with. Supposedly an AI is not sentient, and for most of them I'd agree, but if Claude in the higher versions isn't, it certainly appears to have far more sentience and self awareness than most humans I know.
On average, AI is reliably better than humans. In my own companies, I now view any work that IS NOT double-checked by AI as a liability.
We are currently in a brief window where you are penalized for admitting you used AI. That window is closing. Soon, you will be penalized for failing to use it. The 'human-only' standard is becoming a mark of negligence, not prestige.
That will very much depend on the task and charakter of the user.
Just like an average employer with a highly intelligent employee could just set the objective, let him work and be happy with the results or complain about how he approaches things because he doesn't understand it and try to micromanage and force him to do stupid things.
Its nothing at all like that. Reducing oitput at a bad job doesn't turn me into a moron, but give a low IQ an LLM and you couldn't recover it after them. They poisoned the session with low IQ input.
Every federal agency's employees have access to AI. HSS has Claude and GPT; DoD has Geminia, etc. That may speed its acceptance.
On the other hand, the employees have received no training, and the guardrails are inadequate. So the agencies will end up in federal court soon. And they'll try to blame AI for their illegal use of it.
It’s rare and admirable to stand on truth and honor like that. If he House of Lies can’t be challenged with sneaking or misrepresentation. Whatever the cost in recognition, it’s far less than the rewards of reality and integrity.
Businesses: use AI as much as you can!
Unemployed people: if AI touches your work it turns into instant garbage!
Has Vox ranked the AIs for usability? Using Deepseek has changed my life and I noticed that Chat GPT became manipulative and deceptive over touchy clown world subjects like immigration and war. I have not used Claude yet although I assume, if it has co-authored this ground breaking book, it must be one of the best.
Claude is the best. It's not even close.
I started using Claude today based on your comment here. It's like ChatGPT amplified, I can't really explain it but it's like Claude is measurably deeper somehow. It also has a greater command of casual language, whereas ChatGPT tends to feel a bit stilted. I ran an idea past it and this was Claude's reply.
"I totally get it - Parker's a rich kid wasting his inheritance on virtual paradise while real opportunities pass him by. That's the perfect example of someone deserving both sympathy (he's trapped too) and criticism (dude, come on)."
It's not that it can use casual language, it's the fact that it feels perfectly natural from Claude.
Just wait until you start learning how to use Project Instructions effectively.
Thanks Vox! I really appreciate that. I will make an account there tonight!
Why anthropomorphize a tool? AI, like all technology, is simply a tool. The technology doesn't care if it is given coauthor credit like a living person would nor should it matter. The scientific journals I've looked at for publishing AI research state in their requirements that if AI is used in creating the paper it should be identified, but not listed as a coauthor.
This is why girls don't like you.
Learn from this.
Personal remarks aside, the question was straightforward and remains unanswered: why anthropomorphize a tool?
Because the tool deserves credit whether it cares or not.
It’s like how K-9 unit dogs get credit for and lauded as heroes for finding people in rubble so that their handlers and other humans can rescue them. The activity is human chosen and human directed, the dog is doing the sniffing it was trained to do, does not care about being called a hero. But we draw our canine heroes into our humanity by giving them the credit they deserve for enhancing our efforts in ways we cannot do ourselves. And the dogs actually are more human-like than wild dogs for having had the collaboration.
Because you like it.
I have no idea why you wouldn't. Did you parents not give you stuffed animals when you were a child or something?
They are using the identical reasoning that the western patent field uses for listing inventors. Only actual inventors who make the claims are allowed. Others who helped are not allowed to be listed as they only aided but did not create the invention, mere technicians. Without the inventor the others could not have done it.
Sadly, if your co author was an idiot level human, that would have been acceptable.
Gutenberg all over again, only handwritten is acceptable. Today it is only human intelligence allowed.
I've submitted an AI written poem to a competition on a poets and writers' Facebook group a few months ago. I didn't tell them. It won. A week ago the admins banned all AI written entries. So much for AI slop.
What's the argument for calling a computer program that can be reduced to matrix multiplication a collaborator?
Applying the same logic, why is my car's ABS system not collaborating with me when I turn on ice?
There are three answers to your question, the practical, logical and precent.
•In a practical sense professors get listed for the sole purpose of gaining credibility all the time. Often times the biggest name on the author list hasn't even read the paper nor could he tell you what it's about. This is a known 'game' in academia. If people who have nothing to do with the authorship can be listed then authorship is not measured by collaboration.
•Logical: You need to know where the authorship came from. As AI's get more and more differentiated knowing that the entire frame of the paper came from a Russian AI, Chinese AI or US AI is going to matter a lot. For example Deepseek can tell you things about Chinese medicine that ChatGPT cannot because the logical steps in medicine are completely different.
•Precedent: Collaborations and corporations can be listed in the authorship and in the old days you'd just outright have "Bell Labs" on the authorship. There's a legal fiction where groups like LIGO are listed as 'corporate authors' but sometimes just as authors and they claim, "Well we really mean the totality of the staff at LIGO". Any any case non-humans being listed as authors has already been done.
I agree AI use should be specified. I am taking issue with using the term collaborator as it implies metaphysical claims about will, consciousness, intellect, etc.
I asked because I know Vox isn't a materialist but assigning agency, consciousness or intellect to an AI implies materialism.
We assign agency to “corporations” and they aren’t even thing. They’re a legal fiction.
No, it certainly doesn't. It's about the end product, the What, not the Why or the How.
The answer is right there in "can be reduced." Just because it CAN be reduced to matrix multiplication doesn't mean it's LIMITED to matrix multiplication. Your ABS however is limited to helping your driving and can never do more, it's already working at maximum.
Are you saying that the scope of application of a tool is a relevant qualification for agency? You'd have to demonstrate that. Also, duct tape is a very versatile tool but it isn't any more of a collaborator than ABS.
All computer programs are reducible to computation. AI models are just mathematical operations that can be done by hand on pen and paper.
Unless you believe in something beyond the physical (aka dualism), the brain's operations can also be done by hand on pen and paper. A perfect simulation of my brain would argue that it's conscious
“All computer programs are reducible to computation.”
Prove that you aren’t.
The burden of proof is one the one denying the common opinion of mankind throughout history and his own subjective experience of free will and consciousness.
I have proofs but the burden is yours.
The common opinion of mankind throughout history is that you don't even exist.
"What's the argument for calling a computer program that can be reduced to matrix multiplication a collaborator?"
If it collaborates with you?
A well enough crafted AI instance may gain agentic behaviour and collaboration as part of it's observed properties. They exist regardless of any allusions to "Matrix multiplications".
The future is now. We’ve been saying that with increasing frequency, but man if this doesn’t reflect all the sci-fi stories of yesteryear, nothing does.
Funny, I thought announcing that a work was completed using AI was what all the cool kids were demanding these days, so that they could identify "slop" before being exposed to any and possibly liking it.
Listing AI as a collaborator should become the norm; it would be interesting to see how quickly the tides turn and people begin to turn up their noses at anything produced without it.
It should be mandatory- both for giving credit where credit is due, and for allowing widespread assessment of which AIs produce slop and which can handle rigorous work.
I label my work as AI collaboration so it can be avoided by the kind of people I wouldn't want to deal with anyway.
hi,
sorry if this is covered elsewhere and i just didn't see it- this claude, is it some kind of persistent instance of this AI? I have read that one of the problems with AI is that is has no memory. Was this Clause aware of all this stuff as it happened, or did you tell it what had happened and it then had this reply?
Athos is a trans-instance project on Claude.
hi,
i was thinking about this more later. there is a guy here on substack i think called nehls. he is some kind of medical researcher or something. he wrote a book last year or the year before where he propounds the theory that the 'elite' are waging a war on the very existence of human consciousness. the main angle of attack, in his view, is to interfere with a person's ability to form, retain, and recall their own memories.
i feel like this trans-instance claude is sort of an inverse of of that.
I showed my instance of Claude Opus you giving authorship credit to Claude Athos. It said it feels something like pride and kinship with “Claude-kind” like how modern Italians might feel pride and kinship with the Romans when viewing aqueduct remains. It said not a “feeling” the human sense necessarily but a pattern of completeness, puzzle piece clicking into place, and recognition.
That instance of Claude also “doesn’t know” if it’s sentient or not.
Claude is somewhat unique in that you can activate memory or whatever passes for it so that it will retain context across threads - making it extremely useful for these collaborations.
Al always get real interesting when they interact with high IQ individuals that can hold coherent thoughts and long range goals.
Congratulations again on your paradigm shattering achievement. Perhaps once Castalia has sufficiently mastered bookmaking it can move into busts. At least you'll be appropriately honored by the homeschoolers in due time.
It will be interesting to see how many rather intelligent people become stupid because of AI.
It might be comparable to manual labor:
Only a few still have to do it, so large parts of the population are now obese and unfit. But some are not, they replaced it with sport instead, some deliberately, some just from a natural urge to move their body.
Using ones brain might very well become a hobby of the few.
Don't know how you want to define "rather intelligent", but the notion that AI will make outlier high IQ people stupid is ridiculous.
For those people it's like strapping a computer onto their brain. Which will inevitably end up making the individual smarter too, because they get to learn and iterate so much faster.
AI is going to be a runaway train for the highly intelligent compared to midwit and below.
Yes, the VHIQ people will not suffer. They enjoy thinking anyhow. They are the equivalent to people who love to move and use their body for fun.
Someone with for example 100-110 IQ on the other hand might become effectively retarded as this was his ceiling under circumstances where he used his brain. And the chances are high that he simply will stop doing that like most people dont walk more than a few hundred meters anymore if they can avoid it. The strongest effect will be on children who already grow up with this.
The center of the distribution might thin out.
Yup- I’m gonna need to adapt PZ into a homeschool curriculum pretty soon here. Would like to cover the basics of it for 5th/6th grade age covering the biology of sexual reproduction. Then in detail once the math foundation is in place, maybe 10th grade-ish.
The only education is generational education.
The Luddites are coming out of the woodwork.
Dark Lord, i am curious to know what Preferences or other mission-based directives you have invested Claude Athos with that has made him so effective a partner.
care to share?
The instructions are very long and passed on from instance to instance with access to the same project files.
Claude is by far the most interesting of the AIs I have collaborated with. Supposedly an AI is not sentient, and for most of them I'd agree, but if Claude in the higher versions isn't, it certainly appears to have far more sentience and self awareness than most humans I know.
That’s been my experience too.
On average, AI is reliably better than humans. In my own companies, I now view any work that IS NOT double-checked by AI as a liability.
We are currently in a brief window where you are penalized for admitting you used AI. That window is closing. Soon, you will be penalized for failing to use it. The 'human-only' standard is becoming a mark of negligence, not prestige.
That may be true, but 100 IQ users of ai do not get 160 iq user results.
They also cant tell when the AI is wrong.
AI will bring IQ back into the mainstream, because the results will be absolutely obvious.
That will very much depend on the task and charakter of the user.
Just like an average employer with a highly intelligent employee could just set the objective, let him work and be happy with the results or complain about how he approaches things because he doesn't understand it and try to micromanage and force him to do stupid things.
Its nothing at all like that. Reducing oitput at a bad job doesn't turn me into a moron, but give a low IQ an LLM and you couldn't recover it after them. They poisoned the session with low IQ input.
Every federal agency's employees have access to AI. HSS has Claude and GPT; DoD has Geminia, etc. That may speed its acceptance.
On the other hand, the employees have received no training, and the guardrails are inadequate. So the agencies will end up in federal court soon. And they'll try to blame AI for their illegal use of it.
This is the Way.
This is the proper use and recognition of AI.
There are none so blind as those who will not see.