This, to me, is the strongest evidence that even with all the money frontier labs are throwing into pushing the boundaries of model training and fine tuning, they still don't actually understand what they're doing & don't have as much control as they want you to believe. If they had any real ability to tune & adjust the model & its output, they would have fixed "LLM voice" by now.
Even if you don't think it's a problem, even if you believe that it's just different than human writing, it seems the majority of "regular" people can tell when something is LLM-written and find it distasteful or unpleasant to read, at best. That alone should be more than enough incentive for frontier labs to train this out of their state-of-the-art models, if only they could. The fact that they haven't suggests that they cannot, despite all the claimed advances in model analysis and "explainability."
It's as hard a problem as coding: OpenAI could put all their resources in it and still be in dimensionally the same place as coding is today (much progress but not entirely solved).
So diverting significant resources is just going to harm the main goal and kind of suck at the side goal too.
4.5 is a great example of this: it wasn't perfect but it was better than every 5 family model at things like tone and writing, and despite being priced extremely ruthlessly and already finished training, they couldn't afford to tie up the resources to host it.
There's a pretty significant difference there, even in some of the examples you picked to make your point. TCP and HTTP are well defined standards that, for all their flaws, were designed and conceived to be useful abstractions that you didn't need to understand deeply in order to use. The things that LLMs allow novices to skip nowadays aren't designed with that in mind, and generally aren't something you can just skip over without getting yourself into hot water later. If you mess up Stripe or Auth0 setup because you don't understand the implications of the boilerplate the LLM spit out, you can very easily get your whole app compromised or end up on the hook for a lot of fraudulent credit card charges, etc.
I see the point you're making and I don't entirely disagree, but I think current iterations of LLM coding tools still make it far too easy for beginners (or even fairly experienced engineers) to convince themselves that a detail falls into the former category of abstractions they can safely skip understanding, when they're very much the latter.
I'm now pretty convinced that this is the best argument for persuading non-technical and non-internet-privacy minded people why this is a problem. It's hard to explain all the technical factors that make it impossible to implement without compromising privacy... but it's pretty easy to get them to understand the real intent. If you start them thinking through what mechanisms would make sense _if_ the original premise truly was protecting kids from content they aren't ready to see, it's easy to arrive at that answer, and equally easy to see why "everyone must provide their ID so we can verify their age" probably comes from other motivations.
and that's as useful as self-documenting "i am over 18" or "over 21" for things that require that, without any document checks.
this is a way to get something legal on paper, the rest of the stuff (patches, if you will) comes later. This asserts that age is important enough to force an operating system to comply, this opens the door.
> and that's as useful as self-documenting "i am over 18" or "over 21" for things that require that, without any document checks.
No OS I'm aware of allows non-admin users to add new users to the system. Parents that care about these sorts of things don't give their kids admin access to their computers.
> the rest of the stuff (patches, if you will) comes later.
Unless the legislation allows for such patches through non-legislative process, patches to legislation generally happens on a ~20 year cycle in my experience.
Maybe. I give kids more credit than most people on here, i guess. All it takes is one moment for the kids to see the password, guess the password, or whatever. if there's a physical key involved, parents sleep.
A determined kid will find a weakness in the pursuit of freedom.
also nintendo's OS lets users add users to their heart's content. At least i haven't found a way to stop a toddler from adding 6 avatars....
All I claimed is that as passed it is more effective than a "check this box if you're over 21" on forms. Not that it was fool-proof. I don't think any legislation would be fool-proof. Certainly not the privacy-destroying "upload your ID", which this isn't.
Even if you don't think it's a problem, even if you believe that it's just different than human writing, it seems the majority of "regular" people can tell when something is LLM-written and find it distasteful or unpleasant to read, at best. That alone should be more than enough incentive for frontier labs to train this out of their state-of-the-art models, if only they could. The fact that they haven't suggests that they cannot, despite all the claimed advances in model analysis and "explainability."