Should we have some kind of policy on the use of AI in posts?
This is increasingly becoming an issue, where people copy and paste giant blocks of AI generated slop that’s low value noise. With no citations.
I understand that easily accessible AI tools like ChatGPT are like new toys people are fascinated by, but the value provided is extremely low, while taking up oodles of space. I click on a new thread only to realize that it’s another AI slop serving - I would not have bothered had I known ahead of time. Perhaps at least there should be some kind of header in square brackets [AI] so we have the option of skipping the garbage. And better yet, an option to automatically prevent any [AI] tagged post from showing up - something I would utilize immediately.
While AI of yester year (literally last year) would tend to hallucinate a lot. The new foundational models do a great job in deep research mode of doing phd level research in a matter of minutes that would take weeks in the past and it provides great reference documentation.
I do agree the a simple chat ai response is likely not going to be of great value in this group. I tell others that this group has to be the best source of science backed longevity information on the internet.
I agree, and enough to bother posting about it, with clumsy human-written words.
However, it’s also a shifting landscape, so my opinions are for current tech. Next week, who knows?
I use AIs to suss out issues that I can’t get anyone here to talk about , or when I’ve run into a wall, or exceeded my doctors’ expertise. But (as you doubtless know, but it bears repeating) AI is good at sounding authoritative yet is frequently in error, using made-up references. This happens often enough that I have to check everything anyway. Especially when my health is dependent on the results.
The tech is changing rapidly. But right now, today, I’m with @CronosTempi in that if I know a post is AI-generated, I’d rather skip it. I have simply seen too many material errors in my own AI-based discussions to rely on AI-generated posts, unless I know the material well enough that it’s a restatement of what I would have said anyway, if I’d had the time to make nice and pretty.
I have changed my view on AI given the quality of the more recent searches.
I think it provides useful information, but I am unwilling to argue against AI. If a human being is not willing to invest the time to produce a usable argument then that’s life.
I think it is important, however, to give the source of some AI text.
Is a post based upon something that I had been discussing with the hacking aging group. I think it is useful and does a reasonably good analysis on an important issue
I have also used AI to post summaries of youtube videos and to tidy up the transcripts. I think that is better than just posting the video.
I posted an AI summary of the process of stem cell division.
I think all of those were good use of AI. Now if people think any one of those was a waste of time then please tell me. It is a good discussion to have where an answer would be useful and would guide me as to what to do in the future.
Everything will be AI very soon regardless of your approval. IMHO AI is a wonderful (fast) source of information. I use it every day and for everything. If you don’t like it block whatever irritates you. .
I’d like to get more people’s thoughts on this. Without citations I view the AI responses of significantly less value than if a full list of sources are included. So I think we should strive to have most AI responses include citations; I’m not sure what LLMs do this as a standard matter of course, and if in some you need to be using the pay version and not just the free version. Perhaps someone here can comment on this if they’ve looked into it.
We are all capable of posting a query into a chatbot and getting an answer, so just posting an AI response may not be that much value-add. At the same time, we all have limited time and cognitive bandwidth, so sometimes I find the AI postings people have made to be interesting because I’m interested in the topic, and had not yet considered a query for the AI on the topic, so its helpful and saves me time and effort.
Anyway - I’d like to get more input on this topic from our community members to help me scope out the issue, then I’ll post a “poll” to get people’s thoughts on how we may want to manage this type of thing. So please - add your comments if you feel strongly about any of these issues or can see other issues related to how we use AI in the forums.
I gave four examples where I had used AI. I am not precious about this and would like to know if any were thought to be of little use as it will guide me as what to do
The use of AI to summarise a video and tidy up the transcript, I think is entirely positive when posted with the video. I think it is actually rude to just post the video when you can also add a summary and transcript.
The post about regeneration is something I obtained whilst discussing issues elsewhere. This is a category of posting I would happily not repost to RapaNews, but I do it to be helpful. Hence if people don’t actually positively say they want this sort of thing I won’t post it.
Just me but I don’t care what people do here. It’s easy to skip over. Also AI generated stuff, though limited, is at least understandable. So usually doesn’t take long to skim.
I use the silly thing all the time. My son tells me they spend millions of dollars generating polite replies. He says stop being polite, lol. Costing too much and it’s just a dumb machine.
Mind you, one man opinion and all that. Honestly, I couldn’t understand why you posted that, and therefore I felt you needed to justify it by a follow up post explaining the relevance. But the connection was so tenuous and nonspecific I found no value. However, that’s me.
I’m not discounting the value of AI for research (I do it myself!), but simply throwing out walls of text with no citations is an issue - low effort low value, actually often negative value as you have to wade through it all with extremely high chance of misinformation.
All just my opinion of course. I think having some way to mark such posts so they can be optionally hidden by users would be a big user experience improvement. YMMV.
The problem with ChatGPT (not sure about others) is that sources are often entirely fictional/AI generated/hallucinated, especially when it comes to niche areas of research (as much of longevity science is).
This is not uncommon either, so unless people are going to go individually verify that each citation is a real paper, listing sources doesn’t get around this issue.
I think the link between regeneration and rejuvenation is really interesting. Also how stem cells link to hypoxia. However, I am a democrat. I think looking at likes and your response I have 2 votes for these things and one against. I don’t think anyone is suggesting not posting transcripts and summaries with videos. That takes perhaps 10 mins elapsed time I would prefer not to invest. It is important that people express their views because I don’t want to waste my time on things people don’t find useful.
I think that is no longer a valid criticism most of the time. I find chatGPT has some subtle analyses that other LLMs dont have. I pay the cheap subscription but it did help me solve PD.
