People complain about AI writing, and I sympathize.
Yet I often find it a net improvement to ask an AI to explain something I’m reading more clearly.
Dominant models prioritize broad appeal, making current outputs generic.
People complain about AI writing, and I sympathize.
Yet I often find it a net improvement to ask an AI to explain something I’m reading more clearly.
Positive users praise AI for clarifying complex texts and serving as a thinking companion, while negative users object to its writing as inauthentic slop that enables deception.
No Digg Deeper questions have been answered for this story yet.

I like my documents to follow the 3-act storytelling format - even technical docs. Because we humans are deeply wired to be storytelling animals; we consume information presented in the story format best.
So I find myself editing aggressively. And the model becomes a co-intelligence in the art of crafting a “story”.
I personally love the process. And the results. But that’s me.

@dwarkesh_sp Both can be true

@dwarkesh_sp If you asked asked a knowledgeable human to explain it more clearly, would that not be as good? If you randomly chose between the original and AI versions, would the AI consistently be better?

@dwarkesh_sp Key difference is intention right? In the first, reader is subconsciously expecting an energy transfer and feeling through words. In the second, you’re directly asking AI for writing

@dwarkesh_sp Is that based on the quality of the writing, or that it's response is specifically tailored to the areas in which you have questions?

@dwarkesh_sp it's great at explication, just not so great at actually writing. Anyone who disagrees is free to try to take my money!

@dwarkesh_sp The issue seems to be telling it to "write", rather than to just do the thing and copy the output.
The notion of "writing" invokes a sense of verbosity and forced narrative.

@dwarkesh_sp You're absolutely right!

@dwarkesh_sp
The bad mouthfeel of AI-assisted writing is downstream of everyone using 1 or 2 two models to write, and those models thus having to appeal to the lowest common denominator.
This is not the future. I've seen no convincing argument that LLMs are fundamentally incapable of imitating one's personal style of writing, only that such models are inefficient to serve. As smaller models get better and better + models get cheaper to serve (both locally and in hybrid setups), we'll reach a point where it's almost impossible to distinguish between human and AI writing. It will still be important to write things yourself, for the sake of your own thinking, but the signaling benefits will disappear.
What we need is a more robust sense of what is and isn't slop, whether it's written by humans or machines.
People complain about AI writing, and I sympathize.
Yet I often find it a net improvement to ask an AI to explain something I’m reading more clearly.

@dwarkesh_sp AI as a reading/thinking companion is underrated.

@dwarkesh_sp The problem is the amount of people passing it off as their own. The pretending started with the internet, social media glorified it, AI amplifies it uncontrollably. We are being lied to at scale.

@dwarkesh_sp non fictional / informational i find it can rearrange nicely fiction is just really hard to wrangle it to good qualty

@dwarkesh_sp Ai will improve our speaking and writting by a huge margin, just watch..were been speaking boring and vague that time is over

@dwarkesh_sp Text based AI will go away, I guess. Like animals, we will start communicating only through voices. Writing may become extinct. What do you say?

@dwarkesh_sp neuralese is lingua franca for better or worse

@dwarkesh_sp Companion for one's self is different from creator for other's consumption. Poisoning the well vs well, not.

@dwarkesh_sp example? I agree when I'm studying math sometimes, but curious if that's what you're talking about

@dwarkesh_sp Here is a problem I have with AI explaining math: It talks too much. Never the right amount to enforce engagement. You cant truly understand math from info dump. There needs to sufficient play.

@dwarkesh_sp Ironic that AI had to get good at generation for us to realise it’s better at comprehension