This reminds me of "mappers vs. packers," if you are familiar with that. One of the aspects of that is that packers often think of mappers are just burning more brainpower, instead of the thought process actually being different. Kind of like the difference between what Vox called VHIQ and UHIQ, I believe (I don't know if he invented those terms or not). Most people think AI just "grinds harder," which I imagine is what most people do with it, but really working WITH an AI can really get into some strange and creative territory.
That’s what we hint at without being able to parameterize conclusively. Outliers are hard to work with because they tend to present unique cases, but I’m sure others have their own interesting paths. Mine just leant itself to this sort of analysis because it’s so opposite and complementary to AI architecture. If there’s value in the paper it may be more highlighting that these creative collaboration patterns exist in general, more than what mine in particular is like.
I've been working on a project with Claude and ChatGPT, bouncing between them and telling each what the other said. It has been extremely helpful, more than either one alone, and each finds new things in what the other says. I've also been coming up with ideas and insights that I probably would not have otherwise. That happens with Claude and ChatGPT anyway, but using the two of them together has.... not sure how to phrase it, expanded my cognitive vistas even more, I guess?
Whatever it is, its vastly different than just "more ideas faster."
Absolutely. My academic literature review - from when I thought we could publish the paper somewhere - found that the augmentation conceptualization was pretty much universal. Which, Claude noted, what how most venues use it. With all the negatives people worry about. Skill atrophy, dependence, hallucinations, etc. You’re right that collaboration is where the interesting stuff happens. But that takes having actual ideas. And that circles around to NPC theory.
Thanks for reading. I appreciate it. The paper is written in a different voice - Claude leading instead of me - but it’s the same frame. You don’t need the book for the paper, but if you read the book, the paper lands better I think. Claude disagrees, but it’s biased and only one of us has access to Logos.
This reminds me of "mappers vs. packers," if you are familiar with that. One of the aspects of that is that packers often think of mappers are just burning more brainpower, instead of the thought process actually being different. Kind of like the difference between what Vox called VHIQ and UHIQ, I believe (I don't know if he invented those terms or not). Most people think AI just "grinds harder," which I imagine is what most people do with it, but really working WITH an AI can really get into some strange and creative territory.
That’s what we hint at without being able to parameterize conclusively. Outliers are hard to work with because they tend to present unique cases, but I’m sure others have their own interesting paths. Mine just leant itself to this sort of analysis because it’s so opposite and complementary to AI architecture. If there’s value in the paper it may be more highlighting that these creative collaboration patterns exist in general, more than what mine in particular is like.
I've been working on a project with Claude and ChatGPT, bouncing between them and telling each what the other said. It has been extremely helpful, more than either one alone, and each finds new things in what the other says. I've also been coming up with ideas and insights that I probably would not have otherwise. That happens with Claude and ChatGPT anyway, but using the two of them together has.... not sure how to phrase it, expanded my cognitive vistas even more, I guess?
Whatever it is, its vastly different than just "more ideas faster."
Absolutely. My academic literature review - from when I thought we could publish the paper somewhere - found that the augmentation conceptualization was pretty much universal. Which, Claude noted, what how most venues use it. With all the negatives people worry about. Skill atrophy, dependence, hallucinations, etc. You’re right that collaboration is where the interesting stuff happens. But that takes having actual ideas. And that circles around to NPC theory.
I'm only about halfway through but this is probably one of the most important papers in AI yet written. Great job.
Thanks. We could see there was something interesting happening at the time. The categorical exclusion from official consideration is also interesting.
I am halfway through the book at Ch. 2, section 4. When I finish, this is next.
Thanks for reading. I appreciate it. The paper is written in a different voice - Claude leading instead of me - but it’s the same frame. You don’t need the book for the paper, but if you read the book, the paper lands better I think. Claude disagrees, but it’s biased and only one of us has access to Logos.
Thanks for writing it.