Ha! This is both wrong in the inference and in the conclusion.
1 Like
No, it isnât. But thatâs just the first level of how communication works in large companies.
You seem to know your stuff. Okay, letâs move on to level two. I know that you know that I know⊠but weâre all playing along.
I truly hope Scrivener doesnât add any AI features. Theyâre talking about paring down the app to take away distractions. Nothing is more distracting than AI.
The only feature I wish theyâd add is touchscreen scrolling.
2 Likes
Thatâs not in the least bit tru⊠You know what, Iâm not going to get sucked into this one. Yes, youâre absolutely right. Thanks for picking me up on it.
1 Like
Well yes, itâs quite obvious that the primary âadvantagesâ of AI are low cost and lack of autonomy.
2 Likes
Counter: The cost is vast, and they make up stuff randomly.
No argument here. I expect theyâll be a lot less appealing once they start passing the costs on to users, rather than letting VCs subsidize them.
5 Likes
Just remember, weâre at the beginning of whatâs coming.
2 Likes
I summarize what I suggested and I add new ideas for AI in Scrivener:
- Crafting RegExp based on a purpose description (can be done on external tool easily)
- Counting words frequency in a smarter way (sometimes no need for a report on some common word)
- Tag (meta or copy in sub doc) all dialogue lines by character (deducing implicit attribution), so that we can explore the voice of each and check the words being used.
- Get the big picture with a colorful map: parts rich in descriptions, rich in dialogues, in emotion or sensory, etc, and parts deprived of those. And pace, etc.
I hope suggestions donât get lost too much in the flow here (did it become a general discussion on AI?)
1 Like
FWIW, since LLMs burst onto the scene, weâve seen at least as many people saying they donât want AI anywhere near their Scrivener projects as we have requests like this.
Our position remains that the decision about whether to use AI tools at all, and which ones to adopt, is best left to individual users.
Our most recent comment on AI is here:
4 Likes
I donât speak for everyone my writing community, but if AI tools were to be baked into Scrivener, I think many (most?) of our writers would walk away. I donât say that lightly. Itâs a hot topic on our busy wee Discord of known authors and new writers.
5 Likes
That stance is fine by me.
Actually the features I mentionned, which arenât generative AI, would be best provided by some external pluggin (to not add too much to Scrivener code).
1 Like
The author of this article in HuffPost expressed a view I hold:
I donât use AI for anything in my academic or personal life. I value almost nothing more than my ability to think and to freely express myself. Even when I make mistakes, at least they are my mistakes.
Itâs worth a read.
6 Likes
Will Teague (the professor). Using the idea of Trojan horse to trap the cheaters is great! Iâll remember that.
However, he only speaks of generative AI, when just saying âAIâ.
Otherwise, first he would be wrong: he is already using one form of AI or another on a daily basis, even long before chatgpt, without knowing.
And second, where does he draw the line? Like with a classic spell and grammar tool. To the extreme: Is he counting words by hands? Where âintelligenceâ begin here?
And, besides RegExp suggestions (not a human language), all my AI ideas for a writerâs tool are not generative. I guess his views donât apply for this case.
(edit grammar)
It would require pretty sophisticated natural language processing to identify ârich in description, etcâ and to tag dialogue by character. These days, the leading edge of natural language processing is occupied by LLMs.
1 Like
Easier done with GPT / LLM (as a starting point), yes, I guess. Especially if itâs about finding emotions (descriptions would be easier to detect with a classic algorithm).
That said, its task wouldnât be to spurt out words (generate human language). I see that closer to word counting. No fake creativity needed. No intrusion on that side. All doable by people with poor writing skills, given a bit of time to read, so itâs not like the common use of LLM.
I just watched a video that rather makes my point. Predicting the future is always a bit of gazing into a crystal ball, but this one resonates with my own thinking:
(Yes, I know, itâs definitely âout there.â Still, itâs worth watching.)
On the other handâŠ
Confirmation bias is real! The ones that resonate are the ones that should be questioned the most rigorously.
I wonât bore you with the rabbit hole I went down trying to check out Ms. McCoyâs credentials. Iâll just say that most consulting firms make their teamsâ resumes a lot more visible. Thereâs a lot of sizzle, but Iâm not seeing much steak.
4 Likes