A plea to the moderators to prohibit posting of screenshots of AI-generated results on this forum

Faith & Science Conversation
1

Hi all,
Feel free to express your own opinion. Personally, I find my eyes and mind glazing over when faced with a screenshot of a long wall of AI-generated content within a forum thread.
Not only is it

  1. visually tedious to wade through, but
  2. AI generated results are not always reliable and need to be fact checked. Chat GPT is known to make sh*t up. So I’m inclined to be skeptical of the generated results at the outset and often just skip over such screenshots. At the base of it, I am interested in what human interlocutors have to say on this forum, not what a chatbot may tell me
  3. If I want to query Chat GPT I can do it myself. I don’t need to sort through a screenshot of someone else’s search. It sucks up real estate needlessly
  4. If a participant wants to use AI to do some background research before presenting an idea to the forum, I think that is fine. But then I would suggest they should summarize in their own words what Chat GPT told them, and offer a specific point for comment.

thoughts?
K.

6 Likes
2

On a personal note.

I do not read or take notice of Ai as an authority. Perhaps i am missing something but a computer has never been more than its programing and can only access information that it has been given or has access to. It will therefore reflect the views of its programmer(s) . And who are they? especially in terms of theological authority.

Interesting? maybe. Intuitive, unlikely. authoritative, no.

Richard

1 Like
3

There are some forums that ban AI generated content outright. Stack Overflow is one well known example that comes to mind.

3 Likes
4

That claim might be true for old-fashioned computer programs but not for AI.

AFAIK, AI is trained by providing huge amounts of text (net pages and other digital sources) to the AI. The training data is so huge that even the programmers cannot know or master what the AI learns. Thereafter, the AI uses the observed patterns within the training data to generate answers.

The training material may have a stronger influence on the answers than the programmers, as was the case in the clever strategy of Russian operators to bias the training data by producing millions of net pages with false information. The AI cannot separate facts from false claims and it can draw wrong conclusions about the patterns in the training data, which may lead to answers that are far from truth or the hopes of the programmers. The answers may be structurally excellent but contain false claims and omit crucial pieces.

4 Likes
5

I’m sympathetic. And I’m also glad you posted this publicly instead of a private appeal, because the whole community needs to hear this appraisal. Maybe the only thing I can say for [some?] participants here is that they generally seem forthcoming about labeling it AI - though of course, I guess I have no real way of knowing how much unlabeled AI may have been passed off as a person’s response. All I know is that there is a lot of explicitly labeled AI content included, so for what it’s worth, at least that portion isn’t “driven underground” for us to wonder about. Better the “enemy” that is at least kept close and tried openly as a tool than the lurking enemy that reduces us to view all content with suspicion? In overall sentiment and thought, I’m with you - I’m just not sure how it would work or how enforceable it is. But yeah - I agree the content value of a forum becomes vastly diminished to the extent that it starts to get diluted with AI content that is little more than a reflection of “internet humanity” at large rather than reflections from a single human soul (him/herself a cultural repository - but a human one.) Will be interesting to see how other (real humans) chime in here.

5 Likes
6

The point is that , by definition AI has no soul. It can logically assess, associate, and regurgitate, but theology is rarely logical.

It ends up an interactive encyclopaedia.

Richard

1 Like
7

Good thoughts, and I have to admit that I scroll past a lot of cut and paste posts, many of which are AI. AI is a part of our reality, and is likely to become more intrusive in the future, so we need to develop ways to deal with it.

3 Likes
8

Among other problems, people tend to regard AI content as an anonymous but authoritative reference akin to an encyclopedia. Take the following Google AI response:

what are the Sauk Zuni sequence of the global flood?

In the context of the global flood, the Sauk and Zuni sequences represent distinct stages of sedimentation and transgression. The Sauk megasequence is thought to be the violent beginning of the flood, creating the Great Unconformity and containing many hard-shelled marine fossils. The Zuni megasequence is believed to represent the peak height of the floodwaters, with the highest volume of sediment deposition.

Sauk Sequence:

  • Early Flood:

The Sauk megasequence is associated with the initial, rapid flooding event, potentially during the first 40 days of the flood.

  • Great Unconformity:

It is believed to have created the Great Unconformity, a significant geological boundary.

  • Marine Fossils:

It contains a diverse array of hard-shelled marine fossils, defining the Cambrian explosion.

  • Limited Sedimentation:

Some models suggest that the Sauk sequence represents a period of minimal sea-level rise and limited inundation of landmasses.

Zuni Sequence:

  • Peak Flood:

The Zuni megasequence is considered to be the time of maximum floodwater height and sediment volume.

  • Sedimentation:

It is characterized by the greatest volume of sediment deposition on most continents.

  • Progressive Flood Model:

In the progressive flood model, the Zuni sequence represents a later stage of the flood, following the initial transgression of the Sauk.

In essence, the Sauk sequence represents the initial, rapid flooding event, while the Zuni sequence signifies the peak and subsequent sedimentation during the global flood.

This is AI channeling Timothy Clarey, who appropriated the language of sequence stratigraphy as a guise for flood geology, and whose work has been rechurned by YEC organizations, pastors, and evangelized by enthusiastic YEC posters. YEC does little research, but has an outsized, volumous, presence on the internet, and it is evident that AI has incorporated this content into its training.

Then you have to deal with assertions like "I posted directly from Google AI…so the claim was not my own."

I agree that wholesale AI content drags the forum down, and as such, flagging would be an improvement.

6 Likes
9
  • Outlaw AI, and you can be sure that I’ll never tell you. :laughing:
10

I do not think it is a case of outlaw, but it is acknowledging the status as being on a level with Wiki in terms of authority, It has its place.

Richard

11

To be clear, I’m not suggesting that the use of AI by forum participants be outlawed or that AI is completely useless, but only that the “raw screenshots” of search results not be posted. If participants wish to discuss a result of an AI search that they have done, that’s fine. But I’d just condense the AI info into one’s own words and type the information in the context of the specific point being addressed in the thread. Rather than just lazily posting a large screen of (potentially misleading) text that many members will just scroll past because they don’t wish to sort through everything that the AI has spewed.

8 Likes
12

Hmm,

That seems to e a problem as it means that that the person has to understand what they are quoting,

Some people seem not to trust their own words or understanding.

Richard

1 Like
13

That’s the same way I feel about prolific video links too (especially to long movies or documentaries and without any summary statements to indicate that the poster themselves have waded through it to distill out some questions, and they apparently hope others will do that for them.). Note … This is not a reaction to your video post which I am still very much looking forward to viewing, but just haven’t had opportunity to do so yet.

9 Likes
14

Yeah, I agree there too. At least a video link is a little easier (visually) to scroll past and ignore if one is not interested. I actually tried to “shrink” the rectangle of the video link when I posted, but it seemed to automatically jump to fill the whole width of the screen in a most annoying way :grimacing:

4 Likes
15

How about asking AI to provide a brief summary of the main points of the videos and posting that as text with an acknowledgement or its source!?

16

The problem is that AI is known to have biases (sometimes) and it may not summarize a video appropriately. Or if it does, it may not highlight the specific idea in the video that the poster wishes to focus on. In the case of the video I posted, there are graphs and charts and diagrams that make the logic easier to follow, and so a “summary” of the video would be pretty incoherent without seeing the diagrams. But including the visuals in a “summary” would just lead to a lot of visual clutter on the thread which is what one is trying to avoid…?

4 Likes
17
  • So if I think AI knows more about a topic than I do, it’s very possible both AI and I are clueless and I should not post anything about the subject? :laughing:
3 Likes
18

At least if its you that is clueless I can have a conversation with a flesh and blood human and try to rectify that :wink: Just Joking. Actually, I’ve appreciated your posts in the past where you have said something like "Bard told me…: and you present a series of points. That at least gives a specific focus to engage with, if one wishes.

3 Likes
19

And it admits that it does so.

I did a fun thing one day: I asked a really out-there question of Chat GPT, then gave the answer to Grok – without identifying the source – and asked for an assessment of accuracy. Grok assessed that a couple of items were misrepresented and a couple of others were just made up!
Google AI is even worse.

The best use of AI is to find a starting point, or get some clarity for something that isn’t easily found (like arcane rules to strategy games, or population statistics).

A bit scary: I asked Grok something a few days ago about a certain discussion, specifying that it should give references, and in its answer was a reference to a post here on Biologos that wasn’t even a week old yet. That tells me that Grok is watching the web. Then I was trying to find something posted on Quora, and it not only found the question and an answer, it referenced an answer written less than an hour earlier. That tells me that Grok can access the Web in real time.

3 Likes
20

To quote ChatGPT, “I can be misled”.

Which is a cue to skip what follows. Even ChatGPt knows that Google AI makes stuff up.

Not even close – Wikipedia is reviewed regularly and has been shown to have overtaken Britannica in trustworthiness. AI doesn’t even grasp what “trustworthy” means.

3 Likes