Platforms are finally recognizing that people don’t want to consume AI slop. A growing number of sites and apps now have tools and policies to flag, label, and ban AI-generated content.
I think ChatGPT specifically is terrible. Please, use any LLM but OpenAI. Literally any other one.
OP’s example of cooking is also a horrible use case, especially if it’s scraping the web for those SEO recipe sites. It’s literally randomizing and mixing spam.
But I found LLMs to be great tools to assist with neurodivergence struggles, for introspection and things like that. Local models helped me, significantly.
But you have to stay cognizant that you are just using a tool, a mirror of yourself basically.
It’s not mixing spam. It comes up with great recepies that are perfectly seasoned and taste great. Sounds like you never tried it and you are projecting your dislike for AI on something you don’t know.
I get that a lot of people on Lemmy don’t like Ai but come on, we shouldn’t lie about stuff it actually does well. I think it has a few very good usecases and this is one. Free recepies that work well.
Then you have me pegged wrong. I’ve been running local LLMs since 2022.
But this is what ChatGPT likely does, under the hood:
When you ask it for recipes, it goes to Google and searches for them.
It dumps the top results into its context… the same SEO spam you are using your chat to avoid, I’d presume.
It then averages that with its own amalgamated knowledge to “improvise” a recipe. It’s not thinking through the logic and physics of cooking or referencing a cookbook, it’s just taking a guess at what sounds like a plausible recipe written out.
On top of that, line by line, every measurement, every ingredient, every step is also pseudo randomized; there’s a chance it may pick 8 oz for some ingredient, or 16 oz of another. It’s a roll of the dice. That’s how LLM sampling works.
Outside the thinking block, it also has no way to go back and correct itself if one of those randomized measurements it’s obviously wrong; it just goes with it, like an improv actor.
So…That’s just a computer randomized recipe.
It may work, it’s not bad necessarily, but you have no idea if it’ll work or not until you assess it and try it yourself.
So what’s the point?
What’s the reason for doing it?
There’s no advantage over a real cookbook, where at least the actual recipe has been assembled, eaten, and enjoyed enough to be deemed worthy of inclusion by the author. ChatGPT does no such thing unless it finds real cookbooks to reference.
Now, if you feed it real cookbooks into the context (or use some built in feature that does so), and ask it “pick me a recipe I would like. Modify it.”
That’s fine!
That’s right up an LLM’s alley.
But you can’t feed it SEO spam and expect something good out. Garbage in, garbage out is the term data scientists use.
This is what I try to emphasize every time I talk about LLMs. They have to be machines you understand the inputs and outputs of, not a black box you can trust to give you reliable info when you don’t know how it gets it.
On a separate note, I really, really hate OpenAI. I think they are a pyramid scheme.
I always suggest using literally any LLM but ChatGPT. Literally any one. Ideally an open weights one, but please, just not ChatGPT for so many reasons…
Not sure why you’re getting the downvotes. Your comment is educated and makes sense. I’m still against the use of LLM’s based on the fact they are pure theft, owned by billionaires, and destroying the earth, but I do agree, at least use it correctly and use the correct tool. If all information was paid for, all data centers owned by the public, completely transparent on the info they contain, and ran completely on renewable energy, there’d be nothing really wrong with the tech (other than creating slop that destroyed the internet and eroded trust in everything, which is a massive downside).
… Though I’m personally more forgiving Apache-licensed, open weights models trained on peanuts.
Ripping out the commercial aspect and intensity of training nullifies out a whole lot of issues, just like open source code does, even if the models are still problematic (just like a lot of open source code/projects are problematic).
I’m also of the opinion that massive LLM proliferation exacerbated the cracks in internet and institutions that were already there. They were already in trouble, but all this just made it glaringly obvious.
I think you’re misunderstanding them. They specifically mentioned SEO recipe sites, which are websites dedicated to being ranked highest on search engines, which typically means they don’t correlate with recipe quality, but with clickability (or whatever the algorithms like today). There’s a lot of them, so calling them spam isn’t necessarily wrong. Lots of people search for recipes, so it’s bound to attract it.
They just seem to be cautioning that the LLM might pull from bogus sources since there’s not really a straightforward benchmark for a good recipe that it might reference, and that you have to remember you’re using a tool. Your own intelligence is the real test. Which isn’t bad advice when dealing with AI output, but I think you already got that. I would say, don’t be baited into the idea that everyone is some hardcore AI hater because they are skeptical, there are certainly some, but assuming it alienates the many people that have surprisingly nuanced takes.
Yeah good points. I dont know where it gets its raw data but it’s capable of modifying and thinking about ingredients, so I think its behavior is superior to just recepie sites.
LLMs are text models. They’re like weather models; they live in the context of the input, it’s literally their entire “world view.” The inputs are the most important part of the quality of the answer.
You need to be able to audit the inputs. If ChatGPT doesn’t even show them, that is a tremendous issue.
And I don’t mean to criticize recipes it’s given that have turned out well, but that’s, literally, mathematically, objectively, a roll of the dice if it’s not grounded in real recipes.
As an alternative, I would highly recommend:
Asking the LLM service to find the most accredited cookbooks in the niche you want.
Get them locally, copy and paste the categories you want into chat.
Ask it to pick or (or modify/synthesize) a good recipe from that.
This is what text models were designed to do before OpenAI commercialized them.
See, this is where I think you’re wrong. There’s no thinking going on. It’s a probability analysis based on word adjacency in a text dataset, with a little bit of $RND thrown in. If you ask the same question multiple times, you’ll get different answers because of the presence of the random seed. If you ask it why chocolate coated shrimp is a bad idea, it’s not going to answer based on culinary training saying that those two flavour profiles don’t belong together, it’s going to say no because some other crazy person actually tried it and that data made it into the dataset. (And yes, someone did, I checked…)
This is why people who don’t know how computing works should be banned from touching anything LLM adjacent. The amount of misinformation makes me pull my hair out. Even them calling it “AI” is idiotic and I correct anyone using the term. It’s a word generator. Or if you will, a bullshit generator built on stolen knowledge then sold back to you as slop.
+1 for this.
Well, qualified:
I think ChatGPT specifically is terrible. Please, use any LLM but OpenAI. Literally any other one.
OP’s example of cooking is also a horrible use case, especially if it’s scraping the web for those SEO recipe sites. It’s literally randomizing and mixing spam.
But I found LLMs to be great tools to assist with neurodivergence struggles, for introspection and things like that. Local models helped me, significantly.
But you have to stay cognizant that you are just using a tool, a mirror of yourself basically.
It’s not mixing spam. It comes up with great recepies that are perfectly seasoned and taste great. Sounds like you never tried it and you are projecting your dislike for AI on something you don’t know.
I get that a lot of people on Lemmy don’t like Ai but come on, we shouldn’t lie about stuff it actually does well. I think it has a few very good usecases and this is one. Free recepies that work well.
Then you have me pegged wrong. I’ve been running local LLMs since 2022.
But this is what ChatGPT likely does, under the hood:
When you ask it for recipes, it goes to Google and searches for them.
It dumps the top results into its context… the same SEO spam you are using your chat to avoid, I’d presume.
It then averages that with its own amalgamated knowledge to “improvise” a recipe. It’s not thinking through the logic and physics of cooking or referencing a cookbook, it’s just taking a guess at what sounds like a plausible recipe written out.
On top of that, line by line, every measurement, every ingredient, every step is also pseudo randomized; there’s a chance it may pick 8 oz for some ingredient, or 16 oz of another. It’s a roll of the dice. That’s how LLM sampling works.
Outside the thinking block, it also has no way to go back and correct itself if one of those randomized measurements it’s obviously wrong; it just goes with it, like an improv actor.
So…That’s just a computer randomized recipe.
It may work, it’s not bad necessarily, but you have no idea if it’ll work or not until you assess it and try it yourself.
So what’s the point?
What’s the reason for doing it?
There’s no advantage over a real cookbook, where at least the actual recipe has been assembled, eaten, and enjoyed enough to be deemed worthy of inclusion by the author. ChatGPT does no such thing unless it finds real cookbooks to reference.
Now, if you feed it real cookbooks into the context (or use some built in feature that does so), and ask it “pick me a recipe I would like. Modify it.”
That’s fine!
That’s right up an LLM’s alley.
But you can’t feed it SEO spam and expect something good out. Garbage in, garbage out is the term data scientists use.
This is what I try to emphasize every time I talk about LLMs. They have to be machines you understand the inputs and outputs of, not a black box you can trust to give you reliable info when you don’t know how it gets it.
On a separate note, I really, really hate OpenAI. I think they are a pyramid scheme.
I always suggest using literally any LLM but ChatGPT. Literally any one. Ideally an open weights one, but please, just not ChatGPT for so many reasons…
Not sure why you’re getting the downvotes. Your comment is educated and makes sense. I’m still against the use of LLM’s based on the fact they are pure theft, owned by billionaires, and destroying the earth, but I do agree, at least use it correctly and use the correct tool. If all information was paid for, all data centers owned by the public, completely transparent on the info they contain, and ran completely on renewable energy, there’d be nothing really wrong with the tech (other than creating slop that destroyed the internet and eroded trust in everything, which is a massive downside).
All true.
… Though I’m personally more forgiving Apache-licensed, open weights models trained on peanuts.
Ripping out the commercial aspect and intensity of training nullifies out a whole lot of issues, just like open source code does, even if the models are still problematic (just like a lot of open source code/projects are problematic).
I’m also of the opinion that massive LLM proliferation exacerbated the cracks in internet and institutions that were already there. They were already in trouble, but all this just made it glaringly obvious.
I think you’re misunderstanding them. They specifically mentioned SEO recipe sites, which are websites dedicated to being ranked highest on search engines, which typically means they don’t correlate with recipe quality, but with clickability (or whatever the algorithms like today). There’s a lot of them, so calling them spam isn’t necessarily wrong. Lots of people search for recipes, so it’s bound to attract it.
They just seem to be cautioning that the LLM might pull from bogus sources since there’s not really a straightforward benchmark for a good recipe that it might reference, and that you have to remember you’re using a tool. Your own intelligence is the real test. Which isn’t bad advice when dealing with AI output, but I think you already got that. I would say, don’t be baited into the idea that everyone is some hardcore AI hater because they are skeptical, there are certainly some, but assuming it alienates the many people that have surprisingly nuanced takes.
Yeah good points. I dont know where it gets its raw data but it’s capable of modifying and thinking about ingredients, so I think its behavior is superior to just recepie sites.
That’s not how it works.
LLMs are text models. They’re like weather models; they live in the context of the input, it’s literally their entire “world view.” The inputs are the most important part of the quality of the answer.
You need to be able to audit the inputs. If ChatGPT doesn’t even show them, that is a tremendous issue.
And I don’t mean to criticize recipes it’s given that have turned out well, but that’s, literally, mathematically, objectively, a roll of the dice if it’s not grounded in real recipes.
As an alternative, I would highly recommend:
Asking the LLM service to find the most accredited cookbooks in the niche you want.
Get them locally, copy and paste the categories you want into chat.
Ask it to pick or (or modify/synthesize) a good recipe from that.
This is what text models were designed to do before OpenAI commercialized them.
See, this is where I think you’re wrong. There’s no thinking going on. It’s a probability analysis based on word adjacency in a text dataset, with a little bit of $RND thrown in. If you ask the same question multiple times, you’ll get different answers because of the presence of the random seed. If you ask it why chocolate coated shrimp is a bad idea, it’s not going to answer based on culinary training saying that those two flavour profiles don’t belong together, it’s going to say no because some other crazy person actually tried it and that data made it into the dataset. (And yes, someone did, I checked…)
This is why people who don’t know how computing works should be banned from touching anything LLM adjacent. The amount of misinformation makes me pull my hair out. Even them calling it “AI” is idiotic and I correct anyone using the term. It’s a word generator. Or if you will, a bullshit generator built on stolen knowledge then sold back to you as slop.