I was hoping for slightly more than "oh you can see the quirky phrases and em dashes!" First of all, I disagree that all that can write an email can write well. That's just devaluing the complex...
I was hoping for slightly more than "oh you can see the quirky phrases and em dashes!"
First of all, I disagree that all that can write an email can write well. That's just devaluing the complex art of writing well.
Secondly, right after listing the AI quirks the video goes "oh and there have been plenty of novels written like this. I mean - where did the AI learn from?" which just goes against the thesis entirely.
As for "they all write the same way!" - yeah, that's cause you're picking examples written by the same model with the same prompt that just says "write me whatever". You can control the way it writes both by prompting and by switching to a different model. Obviously whatever chatgpt serves by default isn't made or system prompted to write good prose.
He didn't say that if you can write an email you can write well, he said you can write. I believe his point is basically that anyone can write as long they put in the effort, and it will come out...
He didn't say that if you can write an email you can write well, he said you can write. I believe his point is basically that anyone can write as long they put in the effort, and it will come out better than if they used an AI. As he mentions after that part, you might not be the next Shakespeare, but you likely don't aspire to be nor do you need to.
As for "they all write the same way!" - yeah, that's cause you're picking examples written by the same model with the same prompt that just says "write me whatever". You can control the way it writes both by prompting and by switching to a different model. Obviously whatever chatgpt serves by default isn't made or system prompted to write good prose.
I have to respectfully disagree with this. I've been around the block and tried a myriad of models, both cloud frontier as well as local. They all have their own quirks, but they also share some mannerisms. For example I had to rewrite this paragraph because I started with "I have to pushback on this" and realised that both Gemini and Claude write this whenever they want to go a different direction than mine.
Prompting and instructions will only get you so far. The model might respect your directions until a certain point, then it will inevitably fall back to its own way of writing. Claude Opus 5 for example is infamous for this. Speaking for myself, it ignores my instructions to be concise after 2 or 3 prompts, and starts writing its trademarked nonsensical verbose claudese.
Also, I don't think he would know what models people are using when they submit their games to him. That's an assumption that you can't really make about him in this case
Randomly found this video while browsing. I found it refreshing for giving reasons why you're a better writer than AI by using pragmatic reasons and examples, instead of the usual buzzwords that -...
Randomly found this video while browsing.
I found it refreshing for giving reasons why you're a better writer than AI by using pragmatic reasons and examples, instead of the usual buzzwords that - regardless if you're think those buzzwords are right or not - will not bring anyone to your side if they already aren't. He does mention some very real issues like the impact it has on its users capacity for work, but for the purposes of the video it's left ignored.
I recommend giving it a watch, the guy has an uplifting vibe and seems to genuinely enjoy helping others. He also has a good sense of humor, so it was 20 minutes well spent.
TL;DW:
LLMs have quirks and love to repeat them over and over again (like "it's not X, it's Y", or lists of 3 things, etc)
They love writing a whole lot of nothing, and the actual text with substance gets buried in the middle of all of it.
They all generally sound the same, so your work will sound the same as everybody else's using LLMs.
You will pay one way or another if you try to cut corners (e.g. using AI for editing instead of a proper human editor). Be it in time, quality, your own voice, etc.
I was hoping for slightly more than "oh you can see the quirky phrases and em dashes!"
First of all, I disagree that all that can write an email can write well. That's just devaluing the complex art of writing well.
Secondly, right after listing the AI quirks the video goes "oh and there have been plenty of novels written like this. I mean - where did the AI learn from?" which just goes against the thesis entirely.
As for "they all write the same way!" - yeah, that's cause you're picking examples written by the same model with the same prompt that just says "write me whatever". You can control the way it writes both by prompting and by switching to a different model. Obviously whatever chatgpt serves by default isn't made or system prompted to write good prose.
He didn't say that if you can write an email you can write well, he said you can write. I believe his point is basically that anyone can write as long they put in the effort, and it will come out better than if they used an AI. As he mentions after that part, you might not be the next Shakespeare, but you likely don't aspire to be nor do you need to.
I have to respectfully disagree with this. I've been around the block and tried a myriad of models, both cloud frontier as well as local. They all have their own quirks, but they also share some mannerisms. For example I had to rewrite this paragraph because I started with "I have to pushback on this" and realised that both Gemini and Claude write this whenever they want to go a different direction than mine.
Prompting and instructions will only get you so far. The model might respect your directions until a certain point, then it will inevitably fall back to its own way of writing. Claude Opus 5 for example is infamous for this. Speaking for myself, it ignores my instructions to be concise after 2 or 3 prompts, and starts writing its trademarked nonsensical verbose claudese.
Also, I don't think he would know what models people are using when they submit their games to him. That's an assumption that you can't really make about him in this case
Randomly found this video while browsing.
I found it refreshing for giving reasons why you're a better writer than AI by using pragmatic reasons and examples, instead of the usual buzzwords that - regardless if you're think those buzzwords are right or not - will not bring anyone to your side if they already aren't. He does mention some very real issues like the impact it has on its users capacity for work, but for the purposes of the video it's left ignored.
I recommend giving it a watch, the guy has an uplifting vibe and seems to genuinely enjoy helping others. He also has a good sense of humor, so it was 20 minutes well spent.
TL;DW: