@ismh86 / @MacSparky What are your thoughts on Discourse AI, in respect to this forum?
Personally I’m not a fan of AI, but I know many members here are. I’m curious though how good summaries of some of the longer topics here would be.
I do “worry” about “privacy” / “content ownership” (did not read yet whether the processing of posts will be on your Discourse server or on an AI company’s server), but on the other hand I think those AI companies will probably already scrape this forum anyway.
Like you, I am not a fan of this constant drive to infantalise users - have we really reached the stage that a forum post or thread can be too long, so that we need an Algorithmical Idiot nanny to save our poor little brains the effort of reading the words in context to judge who’s talking sense or not?
I’m not quite as robust as @brookter in my wish to keep AI away from this forum 
The Solution feature allows an answer to a question the OP raised brings it to the top next to the question.
I like to read the whole discussion, summarisation by an AI or a person would remove a lot of the nuance and valid points which I like to consider, even i they don’t match my own.
I’m here to learn as much as to help and summarisation (IMHO) would remove those opportunities.
Summarisation by AI usually means ”shortening” the text, without the ability to pick up on the most important findings, topics or opinions. So, basically quite useless imo. (Fully respecting that others may have other experiences.)
As for ownership, I have always considered any text I post on public Internet discussion forums like this as becoming part of the public domain.
I “discovered” this feature because the Trakt forum has this feature enabled, although they never asked for consent to process our posts (as far as I know).
I don’t think the potential value (if there is any) is in the summary of a single post, but might be in summaries of multi-page threads.
I was exaggerating (slightly) for comic effect 
I agree completely!
Any thread over a certain length (for some forums it’s about 3 posts…) inevitably has:
Joking apart (and a serious answer to @rob’s point, which I do take seriously): long threads rarely stay on topic, and often they contain wildly inaccurate information – well-meaning or not – from posters whose knowledge it’s very difficult to assess.
We, as humans, have a chance of ignoring most of it to pick out the reliable information, but how could an AI? It’s just pattern matching, and when the majority of info in the thread is likely to be dross, won’t it highlight the dross on frequency grounds? If we hive off our responsibility for making these judgements to AI, how do people learn to judge well in future?
If they’re using it (with AI) for basically the same purposes they’re already using your data for to host the forum, it’s not a change of use (on the assumption that unlike Reddit they’re not selling the posts to a 3rd party. And with Discourse being self hosted rather than centrally, Discourse as an originations should not have access to or rights to the data on each forum.
Of course all this depends on Software licensing and user agreements.
If it’s a user-driven AI, I could see value. Consider:
“Hey dingus, read through this 500-reply thread called ‘suggestions for a task manager’ and list all options given, along with the number of people that are suggesting each one.”
That would bring value.
If it’s just a random summary, I agree that it’s probably not a great representation of the details of the thread.
I hope MPU won’t enable this.
Just think how much more efficient our forum users would be if we ditched all the reading and let AI write everything! We could all not-read hundreds of forums every day!
This.
LLMs are at there ‘best’ when user invoked.
You can still read the threads, even if an AI summarises them.
I’ve been doing this for years, btw, using WI - Wifely Intelligence. My wife reads incredibly quickly, and I’m a slow reader, but we have similar tastes in books. For the last 25 years she has read books then summed them up and told me whether she thinks I would like them or not. It’s an incredible time saver, and confirms to me that I married well
I think it might be interesting and educational to try it out for maybe a month and see what people think. It might be great or it might be a dud, but if we don’t try it we won’t really know.
That’s probably the best type of “agentic intelligence” currently available 