4 ms·

> Agent can build one for you. Even today. No… No they cannot. This is why people say LLMs are rotting your brain. It’s not currently possible to distinguish

> Agent can build one for you. Even today. No… No they cannot. This is why people say LLMs are rotting your brain. It’s not currently possible to distinguish LLM content from human content automatically and asking the slop machine to do it won’t achieve anything. You would know that if you understood the problem domain. That is exactly why it is important to build real knowledge as a human, instead of relying on LLMs to do all your thinking for you. Give it a shot sometime.

scotty79 4mo ago
> No… No they cannot. Have you tried? Give it your feed. Ask it to pass it through antislop + SLOP_Detector + slop-forensics and whatever else your agent can find for you. Will it filter out every last bit of AI content? No, but I'm sure it will spare you from more than half. Will it leave in every last bit of human produced content? No. But the ones it's going to filter out are gonna read like slop to you anyways. We are living in the real world. There are no perfect solutions for every wish. But a lot can be done towards fullfilling some of their parts. Thanks to agentic coding more than ever. The difference between telling AI what to make and seeing how it goes, instead of sitting on your hands proclaiming it's impossible to make is really small. Hey, the best thing that could happen, it could confirm what you are already sure of.
- solid_fuel 4mo ago
  > Have you tried? Give it your feed. Ask it to pass it through [...] I've reworded this a few times but I still can't figure out how to say it gently, so I'm just going to say it. Your lack of knowledge about something - in this case the detection of LLM generated content - is not on the same footing as actually having knowledge about it. In this case I am already familiar enough with the state of the art for content classification to know that LLM detection systems are not reliable enough to apply to a feed. > The difference between telling AI what to make and seeing how it goes, instead of sitting on your hands proclaiming it's impossible to make is really small. And both achieve the same result: no usable tool. It's not worth spending the time or money it would take to slop together something that isn't going to do the job.
  - scotty79 4mo ago
    > Your lack of knowledge about something - in this case the detection of LLM generated content - is not on the same footing as actually having knowledge about it. I'm not sure why are you saying that. Do I need to be on equal footing to suggest something to you? I'm obviously not on equal footing as a complete ignorant. I never touched the subject because my brain slop filter serves me perfectly well even though it's not 100% accurate. > And both achieve the same result: no usable tool. The goal is not to build revolutionary tool that achieves state-of-the-art in slop filtering. The goal is to make your life a bit better by cobbling few things together at near zero cost. > It's not worth spending the time or money it would take to slop together something that isn't going to do the job. What money? Have you maxed out your sub for the week? What time? I do such things while I'm cooking my dinner. Processing this topic in your head (which you obviously did for quite a while to earn your footing) already costed you more time and at least as much money.
    - solid_fuel 4mo ago
      > I'm not sure why are you saying that. To explain to you that you are wasting your time and that your suggestions on this are not going to move the needle, nor are they particularly valuable. > I never touched the subject because my brain slop filter serves me perfectly well even though it's not 100% accurate. The whole idea is to avoid wasting brain power on filtering slop. > The goal is not to build revolutionary tool. The goal is to make your life a bit better. And since it isn't possible to build the required tool, wasting time and money trying to get an LLM to do something that it can't will in no way make my life better. > What money? Have you maxed out your sub for the week? Are you somehow under the impression that all or even most people pay for LLM subscriptions? I concluded it wasn't worth the cost at all. > What time? I do such things while I'm cooking my dinner. You either spend time writing the prompt and checking the result, or you don't. I'm not interested in jiggling an LLM while I'm cooking. > Processing this topic in your head (which you obviously did for quite a while to earn your footing) already costed you more time and about as much money. Actually if you have knowledge about the field it doesn't take much time at all to sort things by difficulty and understand what approaches are possible and what are worthless. And I have the advantage of carrying that knowledge forward, which means I don't need to spend time prompting an LLM to learn what I already know.

scotty79 4mo ago

> No… No they cannot. Have you tried? Give it your feed. Ask it to pass it through antislop + SLOP_Detector + slop-forensics and whatever else your agent can find for you. Will it filter out every last bit of AI content? No, but I'm sure it will spare you from more than half. Will it leave in every last bit of human produced content? No. But the ones it's going to filter out are gonna read like slop to you anyways. We are living in the real world. There are no perfect solutions for every wish. But a lot can be done towards fullfilling some of their parts. Thanks to agentic coding more than ever. The difference between telling AI what to make and seeing how it goes, instead of sitting on your hands proclaiming it's impossible to make is really small. Hey, the best thing that could happen, it could confirm what you are already sure of.

solid_fuel 4mo ago
> Have you tried? Give it your feed. Ask it to pass it through [...] I've reworded this a few times but I still can't figure out how to say it gently, so I'm just going to say it. Your lack of knowledge about something - in this case the detection of LLM generated content - is not on the same footing as actually having knowledge about it. In this case I am already familiar enough with the state of the art for content classification to know that LLM detection systems are not reliable enough to apply to a feed. > The difference between telling AI what to make and seeing how it goes, instead of sitting on your hands proclaiming it's impossible to make is really small. And both achieve the same result: no usable tool. It's not worth spending the time or money it would take to slop together something that isn't going to do the job.
- scotty79 4mo ago
  > Your lack of knowledge about something - in this case the detection of LLM generated content - is not on the same footing as actually having knowledge about it. I'm not sure why are you saying that. Do I need to be on equal footing to suggest something to you? I'm obviously not on equal footing as a complete ignorant. I never touched the subject because my brain slop filter serves me perfectly well even though it's not 100% accurate. > And both achieve the same result: no usable tool. The goal is not to build revolutionary tool that achieves state-of-the-art in slop filtering. The goal is to make your life a bit better by cobbling few things together at near zero cost. > It's not worth spending the time or money it would take to slop together something that isn't going to do the job. What money? Have you maxed out your sub for the week? What time? I do such things while I'm cooking my dinner. Processing this topic in your head (which you obviously did for quite a while to earn your footing) already costed you more time and at least as much money.
  - solid_fuel 4mo ago
    > I'm not sure why are you saying that. To explain to you that you are wasting your time and that your suggestions on this are not going to move the needle, nor are they particularly valuable. > I never touched the subject because my brain slop filter serves me perfectly well even though it's not 100% accurate. The whole idea is to avoid wasting brain power on filtering slop. > The goal is not to build revolutionary tool. The goal is to make your life a bit better. And since it isn't possible to build the required tool, wasting time and money trying to get an LLM to do something that it can't will in no way make my life better. > What money? Have you maxed out your sub for the week? Are you somehow under the impression that all or even most people pay for LLM subscriptions? I concluded it wasn't worth the cost at all. > What time? I do such things while I'm cooking my dinner. You either spend time writing the prompt and checking the result, or you don't. I'm not interested in jiggling an LLM while I'm cooking. > Processing this topic in your head (which you obviously did for quite a while to earn your footing) already costed you more time and about as much money. Actually if you have knowledge about the field it doesn't take much time at all to sort things by difficulty and understand what approaches are possible and what are worthless. And I have the advantage of carrying that knowledge forward, which means I don't need to spend time prompting an LLM to learn what I already know.