Yesmap Information Center as a prompted AI encyclopedia
Posted: Sat Mar 29, 2025 11:17 am
With the advancement of AI editing, I am thinking it would make more sense to maintain Ipce, NewgonWiki, the NAMBLA archive, and Yesmap's research collections as user-generated fodder for a prompted AI encyclopedia.
To do this, we would need a model for content generation, then all we would have to do is put in the title of the article we wanted it to write, and review/fine tune the output. This could be done by telling it how to approach the topic, i.e. the slant and how much content it should output, and how much it should devolve to sources.
The model would need to integrate internet archive sources, and potentially do it's own archiving. There needs to be some legacy protection, i.e. if it updates every 2 weeks, then a source goes offline, it would still be able to use the source if it were archived somewhere. We'd need to protect ourselves from libel accusations as well, meaning our model would have to be fine tuned.
On legal topics, the potential to compile information would be far greater with AI, but we would have to leave disclaimers on almost every article.
If a model were set in place, MAPs would become the first group to fully utilize AI editing to present a full history and present of their group as a social actor. In our community's case the upside potential is considerable, as an awful lot has been written about the topic, but no one person has been able to fully summarize that information.
This could streamline the efforts of a limited volunteer workforce. For example, instead of working out how to use old archived material, we could focus on menial tasks such as scanning in old papers collected by groups such as NAMBLA and Ipce while they were active.
To do this, we would need a model for content generation, then all we would have to do is put in the title of the article we wanted it to write, and review/fine tune the output. This could be done by telling it how to approach the topic, i.e. the slant and how much content it should output, and how much it should devolve to sources.
The model would need to integrate internet archive sources, and potentially do it's own archiving. There needs to be some legacy protection, i.e. if it updates every 2 weeks, then a source goes offline, it would still be able to use the source if it were archived somewhere. We'd need to protect ourselves from libel accusations as well, meaning our model would have to be fine tuned.
On legal topics, the potential to compile information would be far greater with AI, but we would have to leave disclaimers on almost every article.
If a model were set in place, MAPs would become the first group to fully utilize AI editing to present a full history and present of their group as a social actor. In our community's case the upside potential is considerable, as an awful lot has been written about the topic, but no one person has been able to fully summarize that information.
This could streamline the efforts of a limited volunteer workforce. For example, instead of working out how to use old archived material, we could focus on menial tasks such as scanning in old papers collected by groups such as NAMBLA and Ipce while they were active.