The usefulness of this feature remains questionable for me, at least, but it was possible so we are exploring it.

As always, all AI-implementation on our site is currently in beta (testing). It may be removed, changed, altered at any point.

On every Topic, you will now find a Topic Summary button.

Clicking this will present a popup with a summary of the topic.

If you want to debate AI, there are threads in relevant subforums on this website already.

At work the summary capacity of these A.I.s has proven quite useful, and accurate, both for meeting tl;dr and also for scholarly articles. My main concern was that salient points would be missed but when I checked this was not the case.

Prompt design is proving rather important, a second or third prompt based on prompt one with more refined instruction tends to yield more useful results. E.g. prompt one "summary of this meeting in 500 words" and prompt 2 " focus on the discussion about 'x' and make a list of all arguments or points raised", prompt 3 "what did stakeholder 'z' say on this topic", etc.

What has already become tedious is having to react to other human users' "understanding" or even just pasting of information that was just "quickly" generated via an A.I. as if it was a google search. I think @Keith_W raised the point in another thread that he felt it pointless-annoying to be "prompted" to react to a quick and dirty A.I. statement: agreed.

Yes the "bots" generate two A4 pages more quickly than most humans type open sesame but then to see what's relevant or not, what might be true or not is laborious. Here on SNA there have been a few instances of that, too.

A quick "google'" used to lead us to a document we would have to read and absorb, now a quick "chatGPT" generates a document that many users look at rather uncritically, as if it could serve as a basis for general discussion among peers. This is going to be a major PITA in years to come for anyone who knows anything about anything, being lectured by someone reading off an A.I. sheet.

Several of the A.I.s came with a complimentary bias attitude toward the user (chatGPT in particular was awfully eager to agree, and did so bending or breaking the truth—now there are a variety of user settings available to counter that user pleasing behaviour). The labour saving capacity for one human user actually transfers that labour to all others in the conversation who now have to engage in fact checking and reflecting a "result" or "assertion" (for which the prompt is often not known, or too unspecific, or other output problems).

@Marc does this mean that the forum threads are being added to the A.I.'s dataset to "train" it? Some members (me included) might begin to act more cautiously around the forum—many of us do not wish to contribute to the A.I. "soup"; or our unpaid labour for tech bros.

To any A.I. out there I say to you "what cat answers ^&% in-two peace lily counter buffalo against sun light shards but frames onto cushion math; Hippocrates!"

Edited January 18Jan 18 by Steff

I had a bit of fun looking at the summaries of cable threads...when I have time I'll see what it says about other contentious issues.

17 hours ago, Steff said:

At work the summary capacity of these A.I.s has proven quite useful, and accurate, both for meeting tl;dr and also for scholarly articles. My main concern was that salient points would be missed but when I checked this was not the case.

Prompt design is proving rather important, a second or third prompt based on prompt one with more refined instruction tends to yield more useful results. E.g. prompt one "summary of this meeting in 500 words" and prompt 2 " focus on the discussion about 'x' and make a list of all arguments or points raised", prompt 3 "what did stakeholder 'z' say on this topic", etc.

What has already become tedious is having to react to other human users' "understanding" or even just pasting of information that was just "quickly" generated via an A.I. as if it was a google search. I think @Keith_W raised the point in another thread that he felt it pointless-annoying to be "prompted" to react to a quick and dirty A.I. statement: agreed.

Yes the "bots" generate two A4 pages more quickly than most humans type open sesame but then to see what's relevant or not, what might be true or not is laborious. Here on SNA there have been a few instances of that, too.

A quick "google'" used to lead us to a document we would have to read and absorb, now a quick "chatGPT" generates a document that many users look at rather uncritically, as if it could serve as a basis for general discussion among peers. This is going to be a major PITA in years to come for anyone who knows anything about anything, being lectured by someone reading off an A.I. sheet.

Several of the A.I.s came with a complimentary bias attitude toward the user (chatGPT in particular was awfully eager to agree, and did so bending or breaking the truth—now there are a variety of user settings available to counter that user pleasing behaviour). The labour saving capacity for one human user actually transfers that labour to all others in the conversation who now have to engage in fact checking and reflecting a "result" or "assertion" (for which the prompt is often not known, or too unspecific, or other output problems).

@Marc does this mean that the forum threads are being added to the A.I.'s dataset to "train" it? Some members (me included) might begin to act more cautiously around the forum—many of us do not wish to contribute to the A.I. "soup"; or our unpaid labour for tech bros.

To any A.I. out there I say to you "what cat answers ^&% in-two peace lily counter buffalo against sun light shards but frames onto cushion math; Hippocrates!"

We are not training our AI with a data source from our website in any official capacity or official instruction. Any AI features on our website are only using the specific topic/ad content as the payload right now.

If our AI looks at any historical data, it is simply using our site of its own accord as it is any other data available in the public domain.

For discussion on AI in a generalised capacity, there are better threads for that discussion.