The "here's my prompt only with sentence structure and full grammar" here struck me -- since I don't have any coding to do in my regular life lately, I've found myself mostly using the LLMs a lot for language learning and asking questions where using google would likely involve a lot of combing through auto-translated websites, trying to recognize patterns, etc. The ability to guess the phrase from some best-guess phonetic spelling and, for example, "i heard something like this on a police show when they were trying to get a suspect to stop, any ideas?" is pure magic. Anyway, I found myself asking it if there was a way to quickly switch the mic speech recognition language like you can the keyboard, but... that's kind of a weird feature to want, right? ChatGPT is really the only interlocutor whom I regularly ask to deal with sentences that switch back and forth without warning, conversations that ignore most of its conversational suggestions, etc. I'm rude!
And... it certainly doesn't transmit the full-color personalized rainbow that you get from conversing in your native language, but it's much better at dissecting how each word affects the tone of a sentence than I am.
And admittedly, learning foreign language is mostly a thing my brain thinks is fun rather than a practical skill at this point anyway. My recently purchased cheapo Bluetooth headphones are claiming to do live translation.... I haven't tried it, but even if it's not great yet, I'd bet it's coming.
All rambling to say... I'm not sure where I'm going with this, but I guess it's scary because it can predict patterns within the variable effects in humans, personalize, and split to manipulate more effectively than a human working alone? Again with "to what purpose?" in world where the speed's gotten completely overwhelming for a meat bag.
So I wrote not to long ago where a friend had sent me some Japanese and I threw the kanji into chatGPT and asked it to explain and it was neat how it talked about all the nuance that my friend had mentioned too.
that final bit you talked about is exactly what I'm trying to wrap my head around too - thinks are changing so incredibly fast and it's so...scary good...at some stuff compared to us. I keep dancing around the edges of what does it actually MEAN that a toaster can do so many things we thought were in our exclusive purview just 4 years ago - even if it's just fancy math and semantic modelling...
This has sort of haunted me for a while now: I was introduced to one of my all-time favorite thinkers in public administration, Charles Lindblom, via a textbook structured in chapters about greats in the field with a biography and some selections from their writing in the style of anthologies you more usually see in lit classes. At the end of the biography, it implied that Lindblom had basically invented a lot of the concepts used to think about pluralism in poly sci and PA, then retired after souring on democracy's chances amid the constant bombardment from all sides by messages powered by money. His concern was just at the level of... conventional modern advertising and propaganda, at the level we've lived with all our lives.
And just... how about now?!? And outside the problems of democracy, i can still find plenty of things to use the tool to understand that entertain my brain... but what out there is still worth training myself to do when the tools are so quickly training to do it better? I'm not going to be the one with the big answers here. I guess I'm pretty sure I'll stay entertained, less assured about how big picture humanity and other stories I'm fairly invested in are going to go?
The stuff you're tapping into with Lindblom is exactly what I had in mind when I wrote the "Radical Optimism" stuff...not really an answer but at least how I settle down the same ghosts that are haunting me.
I think another factor differentiating vibe coding from writing is data. Odds are whatever coding problem you are solving can be decomposed into use cases the AI has already been trained on, but much of writing comes from personal experience the AI isn’t aware of. It can fabricate a believable story, but it won’t be *your* story. To get it to write that story, you’d essentially have to tell it the story first.
I think that the the coding stuff has more creativity than you're giving it credit for. For some definition of "creativity?" I also think that an individual's story is...not nearly as unique as lots of folks want to tell themselves. The same with "art." I think, because it takes work and blood/sweat/tears, that we're tricking ourselves into ascribing some higher level of value to the creative activities of writing and art compared to code writing. We aren't as special as we want to think we are.
Me telling the story: me and my friend went to the zoo and saw a flamingo and ate some ice cream.
AI telling that story: me and my brother went to the zoo and saw a parrot and ate a hot dog.
This is the point I was trying to make. Only one of those is your story. The other is just a similar story. If AI doesn’t have all the details, it is going to fabricate your story. Then, it’s not your story.
Coding is often outcome-based and there are many ways to get to that outcome. *Personal* stories and experiences are journeys, and they stop being personal when any of the details of this journey are altered.
Ah okay, I see what you're saying. An interesting lens I'll need to think about - Squirt has a new piece to read today that touches on this a little bit I think but I didn't go down the "personal" road as much as the "novelty" road...they smell similar in some ways though, keep an out.
The "here's my prompt only with sentence structure and full grammar" here struck me -- since I don't have any coding to do in my regular life lately, I've found myself mostly using the LLMs a lot for language learning and asking questions where using google would likely involve a lot of combing through auto-translated websites, trying to recognize patterns, etc. The ability to guess the phrase from some best-guess phonetic spelling and, for example, "i heard something like this on a police show when they were trying to get a suspect to stop, any ideas?" is pure magic. Anyway, I found myself asking it if there was a way to quickly switch the mic speech recognition language like you can the keyboard, but... that's kind of a weird feature to want, right? ChatGPT is really the only interlocutor whom I regularly ask to deal with sentences that switch back and forth without warning, conversations that ignore most of its conversational suggestions, etc. I'm rude!
And... it certainly doesn't transmit the full-color personalized rainbow that you get from conversing in your native language, but it's much better at dissecting how each word affects the tone of a sentence than I am.
And admittedly, learning foreign language is mostly a thing my brain thinks is fun rather than a practical skill at this point anyway. My recently purchased cheapo Bluetooth headphones are claiming to do live translation.... I haven't tried it, but even if it's not great yet, I'd bet it's coming.
All rambling to say... I'm not sure where I'm going with this, but I guess it's scary because it can predict patterns within the variable effects in humans, personalize, and split to manipulate more effectively than a human working alone? Again with "to what purpose?" in world where the speed's gotten completely overwhelming for a meat bag.
this is neat!
So I wrote not to long ago where a friend had sent me some Japanese and I threw the kanji into chatGPT and asked it to explain and it was neat how it talked about all the nuance that my friend had mentioned too.
https://read.ambivalent.dad/i/182947197/narau-yori-narero
that final bit you talked about is exactly what I'm trying to wrap my head around too - thinks are changing so incredibly fast and it's so...scary good...at some stuff compared to us. I keep dancing around the edges of what does it actually MEAN that a toaster can do so many things we thought were in our exclusive purview just 4 years ago - even if it's just fancy math and semantic modelling...
This has sort of haunted me for a while now: I was introduced to one of my all-time favorite thinkers in public administration, Charles Lindblom, via a textbook structured in chapters about greats in the field with a biography and some selections from their writing in the style of anthologies you more usually see in lit classes. At the end of the biography, it implied that Lindblom had basically invented a lot of the concepts used to think about pluralism in poly sci and PA, then retired after souring on democracy's chances amid the constant bombardment from all sides by messages powered by money. His concern was just at the level of... conventional modern advertising and propaganda, at the level we've lived with all our lives.
And just... how about now?!? And outside the problems of democracy, i can still find plenty of things to use the tool to understand that entertain my brain... but what out there is still worth training myself to do when the tools are so quickly training to do it better? I'm not going to be the one with the big answers here. I guess I'm pretty sure I'll stay entertained, less assured about how big picture humanity and other stories I'm fairly invested in are going to go?
The stuff you're tapping into with Lindblom is exactly what I had in mind when I wrote the "Radical Optimism" stuff...not really an answer but at least how I settle down the same ghosts that are haunting me.
I think another factor differentiating vibe coding from writing is data. Odds are whatever coding problem you are solving can be decomposed into use cases the AI has already been trained on, but much of writing comes from personal experience the AI isn’t aware of. It can fabricate a believable story, but it won’t be *your* story. To get it to write that story, you’d essentially have to tell it the story first.
Hmm, I think we just disagree?
I think that the the coding stuff has more creativity than you're giving it credit for. For some definition of "creativity?" I also think that an individual's story is...not nearly as unique as lots of folks want to tell themselves. The same with "art." I think, because it takes work and blood/sweat/tears, that we're tricking ourselves into ascribing some higher level of value to the creative activities of writing and art compared to code writing. We aren't as special as we want to think we are.
Me telling the story: me and my friend went to the zoo and saw a flamingo and ate some ice cream.
AI telling that story: me and my brother went to the zoo and saw a parrot and ate a hot dog.
This is the point I was trying to make. Only one of those is your story. The other is just a similar story. If AI doesn’t have all the details, it is going to fabricate your story. Then, it’s not your story.
Coding is often outcome-based and there are many ways to get to that outcome. *Personal* stories and experiences are journeys, and they stop being personal when any of the details of this journey are altered.
Ah okay, I see what you're saying. An interesting lens I'll need to think about - Squirt has a new piece to read today that touches on this a little bit I think but I didn't go down the "personal" road as much as the "novelty" road...they smell similar in some ways though, keep an out.