Can I include an asshole AI? Or one that nitpicks my grammar? Also need someone to explain how I am wrong about nearly everything and should give up writing? Otherwise it would feel super fake.
Perhaps this is a joke but I'd truly love to sic a fleet of asshole AIs on my drafts.
Grant us the ability to reply to bots in a public and timestamped manner so that, if ever a human makes a similarly ridiculous response, we can just point at our response there. It'd free up space in the piece itself that'd otherwise go to asterisks and other preemptions of armchair dweebery.
Haha yeah I'm experimenting with a few personas. Should add a good grumpy troll hermit library of commenters.
Hey team! I find journaling for a fictive audience to be more effective personally; since it forces me to try digest my thoughts for an external listener.
Then why not add the actual fictive audience through LLMs? That's how this was born. Feel free to leave your thoughts/feedback here.
My condolences for building something cool to share with the world, just to have HN folks commenting on it.
The purpose and utility of this seem obvious to me, but I can already see the stream of typical HN responses coming in.
Godspeed.
Gotcha, why?
So you didn’t “get him”?
I don't understand why this is important, I understand he likes the feedback, but it is artificial. It can help with brainstorming I guess, but I was hoping to hear more about his personal desires and what is lacking right now...
As some one who constantly chats to AI about an array of topics, I can immediately see the value in this. Thank you for building this. As other commentators have already said, I also predict a frosty reception but please don't lose heart. This is cool and will make people become better writers.
> Hey team! I find journaling for a fictive audience to be more effective personally; since it forces me to try digest my thoughts for an external listener.
Okay, but I don't understand the benefit of writing to an entirely fictitious AI construct instead of writing to the ideal of the kind of reader you'd eventually like to have.
I mean, I get that it's frustrating to pour effort into writing something that effectively nobody reads (i.e. you never connect with a wider audience), but engaging with an entirely fictitious audience seems hollow to me.
> I mean, I get that it's frustrating to pour effort into writing something that effectively nobody reads
And if that was the issue this clearly doesn’t solve it since nobody is reading this.
I might be the wrong audience for this considering I have a public blog but this to me sounds like an insane product.
But hey, if someone finds it useful, good for them.
Given that LLMs have no understanding of the text, what is the point of it?
To shuffle it up in semi-random ways that make you think. If you're determined to hate LLMs for any reason or any purpose, just think of them as an elaborate game of Exquisite Corpse or Ultimate Mad-Libs.
You don't have to think LLMs are smart or real people to think of them as useful. I love it when I can make an idea clear enough in text that an LLM can completely regurgitate it and build upon it. I also love it when an LLM trips over and misses the one real novelty that I've slipped into something; what better for an originality test than trying to choke an automatic regurgitator?
Transistors have no understanding of what I'm doing, but somehow I still find them useful.
I feel people talk to LLMs in chat format just so they feel there's someone listening. This puts thats in a journaling/blogging context, hopefully delivering the same value in a unique context.
What do you mean by "no understanding of the text"?
I think it's pretty neat. It'd be cool to have recurring commenters with different personas who focus on different things. I wrote a test post and almost all of the comments focused on one detail to the exclusion of others. With different personas/focuses you'd get more variety in responses. I also think responding to comments would be a natural next step. Or if those commenter personas are persistent, you could @-mention them in a follow up.
This is actually genius and obvious and also feels really, really weird, all at the same time.
Might as well integrate it with an AI boy/girlfriend, have a very hot partner who is also capable of listening...
I do the opposite of this (release stuff publicly and don't have comments enabled) but this looks really nice for a specific kind of audience. I'm curious to see how it'll evolve and be received, good luck!
This seems like a surefire way to build a private echo chamber. I doubt a simulated audience is going to challenge your thoughts or help find what you are trying to say.
That's (in my view) also not the point of journaling, but rather to blow off some steam and digest internal dialogue.
It might be interesting to have some of the "commenters" challenge or critique, as an option?
Yes definitely, I'd love to have a balance of empathy and alternative perspectives
Do the commenters reply to things now? It might be nice to have personas that usually reply, others that reply rarely, etc, and that way you might get long discussion chains, or drive-by comments, or anything in between.
Very interesting idea, in general!
Public social networks are already building echo chambers. At least this one could be made to not have supremacist spam all over it like Twitter/X. Is it really any worse? I guess it wouldn’t have real life impact with an audience of humans.
Echo chambers seem pretty popular tho.
Very much like most of the substacks I see.
Given the rapidly diminishing quality of discourse on the open internet the last decade or so, I understand.
I recently restricted comments on my blog or 15 years to existing subscribers only. It took me a while to accept that after removing the random spam, then the racist, misogynist, homophobic, and other lowlife commenters, I was left with 10-15% of discussions worth reading.
Pretty sad state of affairs, and it’s clear it’s degrading faster every day.
I appreciate you sharing this and it looks made with care. It also seems like a cool take on how to use LLMs.
This isn’t for me, but thanks for sharing.
Ofcourse! Thanks for the kind comment :)
This feels like one of those moments where you're witnessing history be made in real-time. And I hate it.
It takes approximately zero effort to see how this could be both monetized and used for harm. This episode of Black Mirror is writing itself and congratulating itself in the comments.
At what point is the internet just ... dead? AI posts with AI comments generated for the sole purpose of spamming affiliate links to drop-shipped products we don't need to generate ephemeral satisfaction in fulfilling our purpose as an economics unit of consumption? Where do I sign up for the beta?
A good extra use-case could be to test an article out and tease out any flaws before releasing it to a live audience.
I think that’s the main use case.
this could be good to help someone steelman their arguments, highlight implicit assumptions, identify communication that's confusing, etc. could be interesting to get reactions from a random sampling of synthetic personalities from within a target audience set.
What happens when you trigger the models censorship function?
Does it comment that it can't comment on that?
Honestly, fine. I think some commenters have a point (on both sides). But really y'all should just try talking to yourself if you think you'll like this.
OP, pay no attention to the haters. Brilliant idea, especially given the slop out there. For my journaling, I've been experimenting with similar.
That being said, whatever I do I'm going to keep completely in-house, as in offline with local LLMs only, because privacy.
[flagged]
"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."
I love how you see this super introverted thing for super introverts and your advice is "maybe meet a human".
Well, here's what this human has to say: Maybe you just aren't the target audience.
There may, in fact, be limits.
Also, the typical misapplication of what introversion actually refers to..
Using AI for companionship is poison.
Absolutely. And as we are increasingly coming to realise, can actually cause dangerous delusional states.
I didn’t see a super introverted thing for super introverts. I saw something that, from a place of very genuine personal belief, is at once useless, probably counterproductive to developing good writing, and possibly dangerous - particularly if, as some people have suggested, it’s about journaling and introspective practice. It just feels like a way to accelerate your way to a state of AI induced psychosis.
If you want to write, write.
If you want to write for an audience write and make your writing public.
If you want to write for an audience and get feedback, join a writing group - in person or virtual.
If you want to use AI to workshop ideas, craft characters or build frameworks to write around, do that.
But a cosy little claque of glad handing AI simps? Honestly, what’s the point?
I don't understand why you feel the need to judge it. If someone wants to use it, they should use it. Why do you need to understand their reasons, or why do they need to justify it to you?
It literally does not concern you at all what people choose to do privately. Why feel the need to get involved?
Good tip, will add this to a future FAQ section
This is sarcastic, right? I refuse to believe people are doing this for real.
Thanks. I hate it.