When you said "I hope people are forgiving of my spelling errors", I thought "I've never seen an error in your newsletter." So I was reassured when the final paragraph includes what is a probably a typo for "but": "Cut that doesn’t mean that the growth of AI abilities will continue indefinitely".
Here is the key sentence: "I am stuck on a paragraph in a section of a book about how AI can help get you unstuck."
I find the same.
But, there is more!
I find that getting it to write something fires me to write things that I was not planning to or that escaped me for the moment *even in the area of my core competence.*
Similar to Cunningham's Law which states "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."
The best way to get unstuck may very well be to ask for a bunch of suggestions for the next sentence or idea from an LLM, which will highlight all the ways you COULD go that you really know you don't want to. I'm describing a process of elimination where I think "nope, not it" until I find one that I can work with and make my own. I just now realized that I've been doing this a lot lately. It's really helpful!
Writing is not the hard part. Deciding what is worth writing is the hard part. Even very smart people can write for their whole lives without contributing anything worthwhile.
I have a similar interpretation of the role of AI for white collars.
In my industry I call it "roboconsulting": the integration of AI methods and tools to amplify the human aspect of management consulting.
Consultancies are under pressure, and won't be able to enjoy the "other side" unless they are able to increase their productivity by a significant margin.
The way to do it involves AI and an appropriate operating model.
I've written a 3-part series on it in my newsletter... Who knows, maybe it'll grow into a book some time in the future (helped by AI, of course)!
The singularity has occurred. There is no you and AI as separate entities. You and AI wrote this book so make sure you add your AI cowriter as the coauthor/ghost writer (Gemini, Copilot, ChatGPT) to your book.
Your words (or AI words) provide the evidence:
"As you will see later in this post, AI played a critical role in writing the book"
"Then, naturally, people also ask me if I use AI to write for me, and the answer, again, is no, for a reason that I think is useful to understand."
"When I am doing an analysis with the help of AI, I will decide on what statistical approaches to do but then let the AI handle producing graphs"
"Can you help me rewrite the paragraph and finish it by giving me 10 options for the entire paragraph in four professional styles?"
"While I rarely used any of the text the AI produced, it gave me options and pathways forward."
"Make this better, in the style of a bestselling popular book about AI, or add more vivid examples."
Note: Gemini prompt and answer:
What is it called when a book author using chatgpt or other AI tools as a resource in writing a book?
There isn't one universally accepted term for an author who uses AI tools like ChatGPT in their writing process. Here are a couple of terms you might encounter:
AI-assisted author: This is a general term that describes an author who uses AI tools to some degree in their writing.
Co-author with AI: This term emphasizes the collaborative aspect of using AI. The AI tool is seen as a partner in the writing process, not just a resource.
Other terms you might see include "human-in-the-loop author" or "augmented author," but these are less common.
I pre-ordered the audiobook so I can listen while I take walks and “touch grass”. Though I suspect I’ll get the physical book eventually too as that is how I refer back to things.
Have you thought about creating Copilot GPT versions of the book GPT companions? GPTs built using Microsoft Copilot Pro can be accessed by free Copilot users who have the link so it removes the ChatGPT Plus constraint. I suspect you will have people who buy your book who do not subscribe to a premium AI model (students perhaps)
What computer hardware do you use? this is a practical question, as I am upgrading my PC and I want to make sure my new CPU, GPU, RAM, etc will be at a level that enables me to Centaur or Cyborg.
Also, how much data storage should I plan for -- both SSD and HDD? Hard drives are huge now, 14 TB and up! Is this discussed in your book, or in a past post?
You are obviously not in a position to recommend specific hardware, but it would be useful for your readers to know minimum and desirable specifications when buying or building a computer: CPU, GPU, RAM, hard disk storage capacity.
I see a relationship between your cyborg analogy and Claude 3's mixture-of-experts architecture.
It isn't perhaps so much a sharing of responsibility with the AI as much as it is requesting a richly informed perspective from which to make decisions. (Skip the next paragraph if you're already familiar with MoE)
--
The mixture-of-experts architecture is a machine learning approach that combines the outputs of multiple specialized models (experts) to make a final prediction, guided by a gating mechanism that determines the relevance of each expert's input based on the given data.
--
In other words, you, the human, are the gating mechanism or higher-level intelligence which guides AI output. A human using an AI like Claude 3 is then the higher intelligence to Claude's higher intelligence, which itself resides over a group of expert AI systems. So imagine using multiple instances of Claude to roleplay different readers, each with their own considerations, perspectives, and domains of expertise--- what feels like a reader group quickly becomes a village.
From the human perspective, we feel cyborg-like in having one AI companion-- but in reality and as system architectures change, it might look more like a grove - human in the middle, surrounded by branching trees in each desired direction.
Thank you for sharing the above! I had never heard of the MoE architecture of Claude 3 but recently started using it and have been really impressed so far.
I used your earlier writings from last year to help me write my novel about AI and what it means to be human. People like to ask whether AI wrote the book and I can say no. But it was a fantastic co-creator and helped in many ways, including challenging my own biases (did you know you can have it make counter-arguments better?)
It was also super useful for coming up with names, scene generation, adding or removing emotion, and basically anything you can think of.
But it can't write very well.
Here's a super short (2.5 min) video talking about it:
When you said "I hope people are forgiving of my spelling errors", I thought "I've never seen an error in your newsletter." So I was reassured when the final paragraph includes what is a probably a typo for "but": "Cut that doesn’t mean that the growth of AI abilities will continue indefinitely".
Looking forward to the book!
Here is the key sentence: "I am stuck on a paragraph in a section of a book about how AI can help get you unstuck."
I find the same.
But, there is more!
I find that getting it to write something fires me to write things that I was not planning to or that escaped me for the moment *even in the area of my core competence.*
I think this is a good point, you need to separate out productive and unproductive struggle.
Similar to Cunningham's Law which states "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."
The best way to get unstuck may very well be to ask for a bunch of suggestions for the next sentence or idea from an LLM, which will highlight all the ways you COULD go that you really know you don't want to. I'm describing a process of elimination where I think "nope, not it" until I find one that I can work with and make my own. I just now realized that I've been doing this a lot lately. It's really helpful!
Love it. I get stuck lots. AI has helped no doubt. Preordered in November. Truly can't wait.
Hope you enjoy!
OK, I pre-ordered the paperback.
I look forward to reading your new book, good luck with the book launch!
Pre-order - done!
Writing is not the hard part. Deciding what is worth writing is the hard part. Even very smart people can write for their whole lives without contributing anything worthwhile.
I have a similar interpretation of the role of AI for white collars.
In my industry I call it "roboconsulting": the integration of AI methods and tools to amplify the human aspect of management consulting.
Consultancies are under pressure, and won't be able to enjoy the "other side" unless they are able to increase their productivity by a significant margin.
The way to do it involves AI and an appropriate operating model.
I've written a 3-part series on it in my newsletter... Who knows, maybe it'll grow into a book some time in the future (helped by AI, of course)!
Ethan, Ethan, Ethan!
The singularity has occurred. There is no you and AI as separate entities. You and AI wrote this book so make sure you add your AI cowriter as the coauthor/ghost writer (Gemini, Copilot, ChatGPT) to your book.
Your words (or AI words) provide the evidence:
"As you will see later in this post, AI played a critical role in writing the book"
"Then, naturally, people also ask me if I use AI to write for me, and the answer, again, is no, for a reason that I think is useful to understand."
"When I am doing an analysis with the help of AI, I will decide on what statistical approaches to do but then let the AI handle producing graphs"
"Can you help me rewrite the paragraph and finish it by giving me 10 options for the entire paragraph in four professional styles?"
"While I rarely used any of the text the AI produced, it gave me options and pathways forward."
"Make this better, in the style of a bestselling popular book about AI, or add more vivid examples."
Note: Gemini prompt and answer:
What is it called when a book author using chatgpt or other AI tools as a resource in writing a book?
There isn't one universally accepted term for an author who uses AI tools like ChatGPT in their writing process. Here are a couple of terms you might encounter:
AI-assisted author: This is a general term that describes an author who uses AI tools to some degree in their writing.
Co-author with AI: This term emphasizes the collaborative aspect of using AI. The AI tool is seen as a partner in the writing process, not just a resource.
Other terms you might see include "human-in-the-loop author" or "augmented author," but these are less common.
Hmm? How about "I, Cyborg: Using Co-Intelligence"
Written with spell check.
"Let's delve" the most overused phrase used by AI!
Let's delve into a tapestry of complexities, and together we'll establish a more nuanced understanding.
I pre-ordered the audiobook so I can listen while I take walks and “touch grass”. Though I suspect I’ll get the physical book eventually too as that is how I refer back to things.
Have you thought about creating Copilot GPT versions of the book GPT companions? GPTs built using Microsoft Copilot Pro can be accessed by free Copilot users who have the link so it removes the ChatGPT Plus constraint. I suspect you will have people who buy your book who do not subscribe to a premium AI model (students perhaps)
I have been hoping to get Copilot GPT access for that
What computer hardware do you use? this is a practical question, as I am upgrading my PC and I want to make sure my new CPU, GPU, RAM, etc will be at a level that enables me to Centaur or Cyborg.
Also, how much data storage should I plan for -- both SSD and HDD? Hard drives are huge now, 14 TB and up! Is this discussed in your book, or in a past post?
Thank you!
You are obviously not in a position to recommend specific hardware, but it would be useful for your readers to know minimum and desirable specifications when buying or building a computer: CPU, GPU, RAM, hard disk storage capacity.
I see a relationship between your cyborg analogy and Claude 3's mixture-of-experts architecture.
It isn't perhaps so much a sharing of responsibility with the AI as much as it is requesting a richly informed perspective from which to make decisions. (Skip the next paragraph if you're already familiar with MoE)
--
The mixture-of-experts architecture is a machine learning approach that combines the outputs of multiple specialized models (experts) to make a final prediction, guided by a gating mechanism that determines the relevance of each expert's input based on the given data.
--
In other words, you, the human, are the gating mechanism or higher-level intelligence which guides AI output. A human using an AI like Claude 3 is then the higher intelligence to Claude's higher intelligence, which itself resides over a group of expert AI systems. So imagine using multiple instances of Claude to roleplay different readers, each with their own considerations, perspectives, and domains of expertise--- what feels like a reader group quickly becomes a village.
From the human perspective, we feel cyborg-like in having one AI companion-- but in reality and as system architectures change, it might look more like a grove - human in the middle, surrounded by branching trees in each desired direction.
Sorry, this isn't directly aimed at you, Ethan, not implying you wouldn't know MoE! Got messy with my target audience as you/other readers :|
Thank you for sharing the above! I had never heard of the MoE architecture of Claude 3 but recently started using it and have been really impressed so far.
Man I've rarely read anyone explain things so efficient and clear. Can read your sub for hours, thanks man really ❤️🖖
I used your earlier writings from last year to help me write my novel about AI and what it means to be human. People like to ask whether AI wrote the book and I can say no. But it was a fantastic co-creator and helped in many ways, including challenging my own biases (did you know you can have it make counter-arguments better?)
It was also super useful for coming up with names, scene generation, adding or removing emotion, and basically anything you can think of.
But it can't write very well.
Here's a super short (2.5 min) video talking about it:
https://www.youtube.com/watch?v=psHFoic0fn0&t=19s&ab_channel=PolymathicDisciplines
Is there a table of contents available anywhere? Given your explorations and writings, I could see the book going in many different directions! :)