I'm not going to pretend that any of these questions have actually been asked yet, but the day is young
For really good reasons, AO3 is extremely defensive about random sites automatically downloading works on a regular basis, or at all. Rather than attempt to get around their systems to block me, I'm optioning to stick with things that have an official way of getting at them.
I am Nicholas 'Aquarion' Avenell, I've been in fandom for a while. You shouldn't, though. I'm only requiring enough data to log you in and give you access to the things you ask for - I don't even keep or display copies of the works you want to track for longer than it takes to count the words in them.
Physically, in a database in the EU. Passwords are encrypted, emails aren't. Canary warning: we've never had a data breach. Legally, Novelathon is produced by Istic Networks ltd, a Limited Company registered in the UK.
It isn't. I'm only counting the words in it.
I'm still not hosting it, just monitoring it. Take it up with the author or, better still, don't look for things you don't like.
Oh god. I need to write a full article about this, but in brief:
Using generative AI trained on works not explicitly out of copyright or given explicit permission to be included in the corpus is a modal desolation.
All the current publically available models for Text and Image generative AI have been found to be using data not included in the above definition (Except Adobe Firefly, on a technicality)
Therefore the use of any current LLM to generate "creative" works is ethically bankrupt.
It's environmentally destructive. So are any servers, but generating LLM models is more
so. There's
a lot of misinformation about environmental impact (Any study that doesn't split the difference
between
There are valid and useful things to be found in LLM use, including moderation (Humans who moderate social media are psycologically damaged by the experience, and it's a thing technology should be able to do), image study (LLM's uses for pattern matching can help medical research, and LLM models have been used to identify areas in need of disaster relief) and such, but the general public availablity of generative AI has been incredibly bad for society in general.
Full transparency, I have used generative AI to help write some of the code that runs this site, mostly because it's infecting a lot of code development jobs, which is my day-job, and I need to understand its strengths and weaknesses better than I do in order to decide where to fight my battles on it. For this I'm using Local models as far as I can (LLMs hosted on my own hardware, not calling out to a third party), to minimise the ethical impact. This is not a "Vibe Coding" situation, more advanced auto-complete.
I have no way of stopping them. I'm looking into being able to automatically guess if something is AI generated along with the wordcount, but automated systems for that currently have a cost-per-submission and I do not want to pay for that while this site doesn't make any money. Maybe later.