It’s actually pretty easy to distinguish AI from real text because all AI generated texts have 2.4 children. In aggregate it is indistinguishable statistically but for a given text it’s remarkably easy.
The 100k whys of AI
21–30 of 111 posts
Re: The 100k whys of AI
#22When you generate one or two blog posts with LLM they look pretty good. And you will be impressed with that one clever bit it adds that you didn't even ask for. But then you generate 50 of them and they all converge into the same pattern. It's hard to prove that an article is AI generated but they are instantly recognizable. An aside, I usually take my written blog posts through a pass on Notebooklm to generate a pod…
I suspect there are new invariants emerging. We don’t know what they are and we will probably have to reach into the liberal arts to describe them but to me what you’re seeing is akin to the subatomic world exposing itself through diffraction patterns.
You're just looking for the study of rhetoric. LLMs have clustered on certain rhetorical patterns/gestures, probably because of a combination of frequency in input and bias in training. But rhetoric also concerns the logical structures that underpin communicative techniques, and it's this logical infrastructure that's shaky or bizarre in LLM content (like the GP noticing how "pushback" always resolves without further examination).
Re: The 100k whys of AI
#23We likes this "same, complex set of mannerism" when using LLM for programming. If you ask LLM to write a certain function for you, it gives you statistically obvious implementation. But maybe for writing an original book, this feature is not so desirable
It does not. Sometimes it will spawn a mess of ad hoc python, sometimes it will do curl and sed, and very very occasionally it will use the correct tool for the job if it remembers to use the skill you developed in the previous session.
Re: The 100k whys of AI
#24It's even more worrying when you look at the contents of these "books", they are riddled with erros:
Re: The 100k whys of AI
#25Notably, in programming this is actually a desirable feature for most problems. Even human programmers are taught to produce predictable and obvious code whenever possible. I wonder is ultimately this is an artifact of optimizing the models for code, that they become less creative.
I’ve rarely experienced this. Typically what is requested is code that has unpredictable pauses, takes unbounded time, has two kinds of null, etc.
Re: The 100k whys of AI
#26When you generate one or two blog posts with LLM they look pretty good. And you will be impressed with that one clever bit it adds that you didn't even ask for. But then you generate 50 of them and they all converge into the same pattern. It's hard to prove that an article is AI generated but they are instantly recognizable. An aside, I usually take my written blog posts through a pass on Notebooklm to generate a pod…
> It's truly impressive when you just consume sporadically. But listen frequently and they converge into one blob.
And something that shows that behavior is a scammers wet dream!
Re: The 100k whys of AI
#27A nice illustration of the homogeneity of LLM responses. Another way to describe this effect would be… If you ask humans to write 1,000 books, you're asking 1,000 different humans with different experiences and different skills and different moods (etc.) to write those books. But if you ask LLMs to write 1,000 books, you're probably only talking to 3 or 5 different models, tops. And they've all trained on the same or…
Reminds of Pluribus.
Re: The 100k whys of AI
#28Earlier quoted context omitted.
> you're asking 1,000 different humans with different experiences and different skills and different moods Simply, if you ask an LLM, you're asking always to the same mind, and always for the first time.
Also since those are lazy, you are also asking always in the same manner. How homogeneous were the prompts that generated those covers? People are making cookies with cookie cutter number 5 and other people wonder how come they are all the same.
Classic self selection effect though - if you’re resorting to LLM writing you’re almost certainly skewing lazy enough to not even bother trying to add perturbations strong enough to make the response deviate from the uniformity of the slop.
Re: The 100k whys of AI
#29Earlier quoted context omitted.
I suspect there are new invariants emerging. We don’t know what they are and we will probably have to reach into the liberal arts to describe them but to me what you’re seeing is akin to the subatomic world exposing itself through diffraction patterns.
You're just looking for the study of rhetoric. LLMs have clustered on certain rhetorical patterns/gestures, probably because of a combination of frequency in input and bias in training. But rhetoric also concerns the logical structures that underpin communicative techniques, and it's this logical infrastructure that's shaky or bizarre in LLM content (like the GP noticing how "pushback" always resolves without further…
> You're just looking for the study of rhetoric.
I presume you mean, that what I and others is observing is patterns in mere rhetoric. That this is just unimportant window dressing around the actual problem solving.
Yet, generation of rhetoric seems to be one of the key usecases, and one of the key features that makes this technology seem “intelligent”.
Re: The 100k whys of AI
#30I think a majority of content consumers can already distinguish LLM content from human content. I'm looking forward to the day that they're intelligent enough to care, but I'm not holding my breath. Orwell framed it pretty well in 1984 with the machine-generated songs that were new every year, but always tugged on the heartstrings of the proles. They weren't really readers or listeners to music or appreciators of art before, and they can be caught in the trap indefinitely, since they'll never be aware or what came before or what's being done now outside their AI-driven feed.
Horselover Fat had a pretty good take on machine generated content, too.