Your 10 minute bet paid out – although in fairness I've already been thinking about this a lot this last week!
I particularly like your ideas on path 3 and 4. ChatGPT in its current state can be very useful as a tool for designing thinking and brainstorming. The point here is to hallucinate at least at the beginning! Using AI as a tool to foster a divergent thinking processes is a great application where the quantity is more important than wrong answers to start. All it takes is one idea that can break you out of a local minimum where you are stuck and inspire a new train of thought.
This is excellent. One thing I wonder about is that the level of this querying process of co-creation requires a certain mindset to start with. Those who would do this seem like they would've done something similar elsewhere (using Google and Wikipedia perhaps) and now use better tools. This seems a smaller group than I'd like, especially re the love of whimsy. I hope this expands thuogh.
Pretty much every 4yo kid I've met just keeps asking questions like this until everyone around is tired. Then they go to school, and get punished for straying away from the official curriculum. Nothing kills curiosity like standardized schooling.
Which makes me wonder, what kind of possibilities is AI going to unlock in the education space? How far away are we from a fully personalized and interactive Young Lady's Illustrated Primer?
Option 3 is somewhat concerning - after pointing out that the ChatGPT isn't a domain expert and shouldn't be relied on for (eg) medical advice, you're effectively proposing that people rely on it for financial / entrepreneurship advice?
The first step, coming up with a list of business ideas, is a great use of a tool that lets the user surf across the collected subconscious of the internet circa 2021 and see the output in easily digestible formats. Once you go beyond the point of kicking around ideas, though, you really do need some domain expertise to be confident that you've filtered out any common misconceptions that the AI is repeating.
More general observation: unsurprisingly the more convincing longer pieces of text are beautifully bland and generic. They use elements that could apply in multiple contexts, which are presumably the elements that recur most often in the training data, blend them smoothly, and produce a kind of written equivalent of the "please hold" muzak from a phone queue. The article about the mouse watching Taylor/the Beatles is funny from this perspective, because the concept is unusual but the phrasing is so cookie-cutter.
I write a lot for various reasons and I'm interested in AI but I'm still unsure how much this will change things for me. I don't often want the phrasing and intent that I bring to a specific document to be bland, I want to engage the reader's interest. It would be harder for me to write the kind of letter that I wanted to write if I started out with a version that felt subtly off from my preferred style, so even in a relatively mundane context I suspect I'll find it easier to start from scratch most of the time.
I also do quite a bit of world building but for me the fun of that process is the part where I figure out the details that are specific to the world I'm creating. I think the most likely use case for ChatGPT in that context would be for me to generate fictional bullshitty text blocks from within a fictional world where I wanted them to sound a little generic. Which could be good for fleshing out something like an escape room or game where you want lots of colour but you don't want it to distract people from the core narrative. I'm also wondering about the possibility of creating fictional languages and character naming schemes, or at least prompts for potential character names that sound similar to existing ones.
So for me this still feels like a mildly interesting diversion that I might add to the repertoire of ways to get idea prompts, alongside various existing random generators and a bit more sophisticated - but not necessarily like something I'd use very often.
I am a little concerned that the overall result might be to massively multiply the flood of bullshit in the world, which is not something humanity needs help with tbh.
Do you think that on effect of ChatGPT might be to enable industry insiders to be more successful innovators?
Until now, outsiders have often been more innovative, because they aren’t weighed down by domain knowledge.
As you’ve shown, now with ChatGPT it is crucial to have enough domain knowledge to call out the BS.
But it also reduces the ‘cost’ of experimentation to near zero, and gives insiders the ability to play around with prompts without looking ‘stupid’, plus the AI can inject some fresh ideas & diversity of thought.
One thing that keeps revolving in my head is the need to teach people how to build their own codex/writing/corpus of work to have ChatGPT help me to riff on my own work _in the style of my own work_. I want to use this tool to help me brainstorm but not in the generalized, average way that it does things, but in a creative way that incorporates my writing.
Another outcome that seems likely: the need to detect situations where generative models have been used to create text. For a chatbot and a generic answer, the risk seems low. If you are emailing someone and they respond with generic text, it might be important to know if you got a realistic or generated response.
I was just asking my partner about this because he works in AI—can models be trained to spot AI-generated text—and apparently it's an ongoing arms race because they *can* but as soon as a model figures out the difference, it can then take that in and adjust how to writes to become indistinguishable
Your 10 minute bet paid out – although in fairness I've already been thinking about this a lot this last week!
I particularly like your ideas on path 3 and 4. ChatGPT in its current state can be very useful as a tool for designing thinking and brainstorming. The point here is to hallucinate at least at the beginning! Using AI as a tool to foster a divergent thinking processes is a great application where the quantity is more important than wrong answers to start. All it takes is one idea that can break you out of a local minimum where you are stuck and inspire a new train of thought.
This is excellent. One thing I wonder about is that the level of this querying process of co-creation requires a certain mindset to start with. Those who would do this seem like they would've done something similar elsewhere (using Google and Wikipedia perhaps) and now use better tools. This seems a smaller group than I'd like, especially re the love of whimsy. I hope this expands thuogh.
I think that is a really great question. Does it require a particular mindset? How much of this is a teachable skill? How do we help create that?
Pretty much every 4yo kid I've met just keeps asking questions like this until everyone around is tired. Then they go to school, and get punished for straying away from the official curriculum. Nothing kills curiosity like standardized schooling.
Which makes me wonder, what kind of possibilities is AI going to unlock in the education space? How far away are we from a fully personalized and interactive Young Lady's Illustrated Primer?
Funny you should ask about the education space...🙂
Please say more! 😊
Option 3 is somewhat concerning - after pointing out that the ChatGPT isn't a domain expert and shouldn't be relied on for (eg) medical advice, you're effectively proposing that people rely on it for financial / entrepreneurship advice?
The first step, coming up with a list of business ideas, is a great use of a tool that lets the user surf across the collected subconscious of the internet circa 2021 and see the output in easily digestible formats. Once you go beyond the point of kicking around ideas, though, you really do need some domain expertise to be confident that you've filtered out any common misconceptions that the AI is repeating.
More general observation: unsurprisingly the more convincing longer pieces of text are beautifully bland and generic. They use elements that could apply in multiple contexts, which are presumably the elements that recur most often in the training data, blend them smoothly, and produce a kind of written equivalent of the "please hold" muzak from a phone queue. The article about the mouse watching Taylor/the Beatles is funny from this perspective, because the concept is unusual but the phrasing is so cookie-cutter.
I write a lot for various reasons and I'm interested in AI but I'm still unsure how much this will change things for me. I don't often want the phrasing and intent that I bring to a specific document to be bland, I want to engage the reader's interest. It would be harder for me to write the kind of letter that I wanted to write if I started out with a version that felt subtly off from my preferred style, so even in a relatively mundane context I suspect I'll find it easier to start from scratch most of the time.
I also do quite a bit of world building but for me the fun of that process is the part where I figure out the details that are specific to the world I'm creating. I think the most likely use case for ChatGPT in that context would be for me to generate fictional bullshitty text blocks from within a fictional world where I wanted them to sound a little generic. Which could be good for fleshing out something like an escape room or game where you want lots of colour but you don't want it to distract people from the core narrative. I'm also wondering about the possibility of creating fictional languages and character naming schemes, or at least prompts for potential character names that sound similar to existing ones.
So for me this still feels like a mildly interesting diversion that I might add to the repertoire of ways to get idea prompts, alongside various existing random generators and a bit more sophisticated - but not necessarily like something I'd use very often.
I am a little concerned that the overall result might be to massively multiply the flood of bullshit in the world, which is not something humanity needs help with tbh.
I teach violin at a nonprofit and I am getting set to use this to write a grant request. Wish me luck!
Um, it is the Beatles, not the Beetles :-)
Can't believe there isn't somebody on TikTok yet who is documents starting his own business all with the help of AI
You make me want to start a business just so I can do it... or maybe I can document it here on Substack instead?
Nothing much to lose when trying
Do you think that on effect of ChatGPT might be to enable industry insiders to be more successful innovators?
Until now, outsiders have often been more innovative, because they aren’t weighed down by domain knowledge.
As you’ve shown, now with ChatGPT it is crucial to have enough domain knowledge to call out the BS.
But it also reduces the ‘cost’ of experimentation to near zero, and gives insiders the ability to play around with prompts without looking ‘stupid’, plus the AI can inject some fresh ideas & diversity of thought.
One thing that keeps revolving in my head is the need to teach people how to build their own codex/writing/corpus of work to have ChatGPT help me to riff on my own work _in the style of my own work_. I want to use this tool to help me brainstorm but not in the generalized, average way that it does things, but in a creative way that incorporates my writing.
Another outcome that seems likely: the need to detect situations where generative models have been used to create text. For a chatbot and a generic answer, the risk seems low. If you are emailing someone and they respond with generic text, it might be important to know if you got a realistic or generated response.
I was just asking my partner about this because he works in AI—can models be trained to spot AI-generated text—and apparently it's an ongoing arms race because they *can* but as soon as a model figures out the difference, it can then take that in and adjust how to writes to become indistinguishable