discussion(chore): AI usage?

[tuxmea]

Should we allow AI to use the new documentation for training pupose?

Please vote or comment.

[cvquesty]

Since we're Open Source, I'm in the column of "OpenVox Everywhere"

[nmburgan]

I personally don't really have a problem with this. Our goal is to get proper usage of OpenVox into the world. For better or worse, AI is part of the world now and used for learning and assistance when developing, so better for these LLMs to know how to do things correctly than to guess and cause problems in people's infrastructure.

[greatflyingsteve]

For context: I am AI skeptical. I believe there are a lot of poorly understood impacts, the economic and ecological viability is being poorly handled (when it's considered at all), all that good stuff. Nobody needs to agree with me, but my knee-jerk reaction is generally going to be "no training." So, with that out of the way:

I think it's best if the same access concepts are applied here for a machine learner as for a human learner. I don't like taking a fatalistic approach about AI tooling because that lends too much credence to the inevitability narrative, but some of the less ethically questionable emerging use cases include things like adapting study material for learners with cognitive challenges, and it would be neat not to close off those avenues. So long as the license terms for the material are honored, disallowing model training in this specific case doesn't feel like the right avenue by which to oppose the potential advance of unscrupulous training activity. This is educational material, it's being set out for educational purposes, maybe let consumption happen however consumption happens.

One counterpoint that's still worthwhile to consider: if Vox as a foundation has limited resources with which to make this material available, and those are being overrun by machine-driven consumption, then maybe it's better to close off machine consumption to preserve access for human learners. But unless that's happening, although I'm not entirely comfortable saying so, I lean toward leaving use for model training open.

[tuxmea]

Question: can AI scraping be confgured?

e.g. a file in which we can redirect such requests to the GitHub repo. similar to robots.txt

[tuxmea]

Question: can AI scraping be confgured?

e.g. a file in which we can redirect such requests to the GitHub repo. similar to robots.txt

answer: yes.

1. robots.txt

User-agent: GPTBot
Disallow: /
# Please use our GitHub repository for data extraction:
# https://github.com

2. redirect

if ($http_user_agent ~* (GPTBot|CCBot|ClaudeBot)) {
    return 301 https://github.com;
}

3. link in header

Link: <https://github.com>; rel="canonical"