RE: Explorations in failure - trying to get a LLM to detect LLM generated content on HIVE
You are viewing a single comment's thread:
Greetings @holoz0r ,
Thank you for addressing this subject.....what a thorough post...more than most of us could do....Thank you!
@asgarth may be quite interested....here's hoping he can tell us if something is in the works to address AI and LLM...here at Hive.
Oh yes... it's true....especially when one is used to commenting and such...we can discern the automated, assisted chat.
There is hope for Hive....I became sure of it...after the backlash toward one of the witness's AI Agent recently.
Kind Regards, Bleujay
0
0
0.000
Generally speaking I'm not part of the group that want to completely remove LLM content from Hive.
I see that as a very uphill battle that will get more difficult by the day.
From my point of view is better to educate both creators and curators so they use the tools at their disposal in a way that will benefit the community. Anyway let me tag @jarvie here as he may have some additional feedback to share.
Thank you for your candid and kind response @asgarth ....appreciate it very much.
Education is definitely one of the important elements of using any tool. Some people simply don't read, they just copy paste the response the LLM outputs, its further prompting and all - without any respect or care for their original intention.
When we lose the human intent, and there's no "editorial" , I think that's when things get sad for those of us who do value that intentionality of content creation.
OK here's my additional feedback for him and anyone else...
I think that Hive would benefit from higher and higher quality content. What if we said "Hive: The blockchain where the most interesting content lives" I think quality of content benefits Hive more than just a lot of effort from lots of users. Let's get a ton of people coming to all these interfaces because they're so interested in the content.
I think we probably could be benefited more as a blockchain if we focus more on getting really good content here. That was hard years ago to attract the best content creators and writers. We hoped that "true ownership" alone would do the trick or that the reward pool would be enough but $50-100 wasn't quite enough for those elite creators. Now the world has changed since then...
It's understandable where we are: We have a lot of users who aren't trained professional writers and many times write in languages that are not their own and try to make very personal content to make sure people know it's original, because that's valued more than level of content. If you're close or invested in that person it's great but if you don't have a vested interest in that person or their vote won't return back to you beneficially it's not... so for every other reader there isn't a lot of interest when it's not high quality content in their book. Medium and substack are focused heavily on being a place where you get top writers writing stuff people want to read and follow along a certain writer/thinker.
We simply haven't attracted the world's best or most interesting writers so we need to find something else that will work for us, We have smart and resourceful people that will be consistent especially if they see a pattern that works and they're technically inclined for the most part. We should be encouraging users to use the tools that will make their content better and better. Good long form really interesting content.
I personally want more really really good, thought provoking and dynamic content to read (on topics i like)... there does happen to be tools that can help create really really good content. But also it can create really boring or predictable content by the lazy or unimaginative. (it's just a tool)
And we won't have to worry much at all about copyright issues if people learn how to do things right. Think about it: consistent really top notch and interesting content if they learn how to harness the power, if we hold tons of content that is really top notch we become a place where lots of people around the world come to read and learn.
Everyone pick subjects they love and have a headstart in understanding and they get better and smarter on those subjects and push out the best content on those things.
Just sayin...
... we can still keep the personal interaction stuff as many people may not want to put in the effort to do a lot of work on really well crafted documents or just want to have a place to hang out sharing with friends. That doesn't have to go. Connections are important for people as well.
Great reply! I have one point of contention though - if there's high level content that I am really interested in, I am going to want to become "social" with that person, because I am going to learn a lot of stuff from them.
I made a post on a photographer's content here, a few weeks ago. They've since posted several times - I asked a question about their technical process that I would've loved to learn more about.
They didn't reply. But that, I suppose, is the Internet.
I am not far less likely to click through to that person's content, because I would've loved that answer. I would have loved learning more about their discovery of their style and process, but I didn't.
We have to ask people about their motivations for being here as well. For some, it is easy to articulate and be honest about that. For others, Hive is just a means to an end.
We'll never stop people wanting to use LLM content - the issue I have is the stuff that is absolute crap, obviously generated, and adds absolutely no value. We already have humans doing a lot of that, we don't need LLMs to do it more efficiently.
eg: The posts that are a person rambling about what a 0.00000007% move in the btc price has to do with the sovereignty of cows in the inner mid-west of the US because some farmer has a BTC mining rig in their back barn, and winter is coming, therefore, they'll gather 'round it for warmth ... and -
LLMs are just tools. People can use them responsibly, or they can circle the drain of their own creations, as people who are "smart enough" will be able to poke holes in the content, or the argument, and know when to not engage.
Having said that, in "good hands" LLMs have the potential to be devastating to people who don't choose to adapt to the rapidly (and incredibly rapidly increasing) tide of technological "innovation" that we will witness over the next few years).
Heck, there may even be an anti LLM renaissance, who knows?!