Is good writing verifiable? I don't think it is, but LLM's have been hillclimbing writing quality. That being said this is with the help of RLHF.

However there are many other domains which have verifiable rewards in the process of learning them, despite their overall impact not being verifiable. For example, the life sciences, an LLM could be given access to data about an organism, and then make predictions about how a drug or gene therapy will affect that organism. In economics, LLM's could create models of behavior, evaluate predictions over time and see how well those predictions match reality.

>claudisms and chatgpt idioms are making them notoriously bad at prose without a considerable amount of prompting and tweaking.

I think this is because a lot of people genuinely like the claudisms, even though a small minority of technical people don't.

▲

wizee an hour ago | parent | next [-]

The quality of natural language output from today’s LLMs seems no better and often worse than LLMs from 18 months ago, at least in my experience. It’s much harder to build a good RL loop for something subjective like good writing, compared to something with an objective and verifiable right or wrong answer, such as software or mathematics.

▲

howunfortunate an hour ago | parent [-]

If you gave me natural language from a 125 IQ human and a 155 IQ human I'm not sure I could judge which human is smarter from the writing quality.

And that's a massive intelligence gap!

It may just be that writing is mostly saturated, and the remaining perceived gap is not quality but some kind of "cultural fit".

▲

habinero 25 minutes ago | parent [-]

IQ is a useless thing, but you can absolutely tell intelligence from writing. You don't need to write like [author of your choice], you need smart ideas and the desire to make them understood.

AI writing sucks because there's no intent or choice behind it. It's human thought, blended by linear algebra into pink slime and extruded into sentences.

	▲	howunfortunate 11 minutes ago \| parent [-]
		> IQ is a useless thing, but you can absolutely tell intelligence from writing. This is inherently self-contradicting.

▲

throwawayy6767 an hour ago | parent | prev [-]

It's not at all clear that LLMs have gotten significantly (much less hillclimbing) better at writing since GPT4-ish. They did get better at oulipo-like challenges (writing under silly, verifiable constraints like Fable's viral 'facetiously' poem) but that's far from being all there is to good writing.

Anecdotally, creative writing communities do report drastic differences in quality between models (it is my understanding that Gemma 4 is especially praised) so there is something to it being verifiable but it's a far cry from being something that can be distilled into objective math/code-like evals. One could even say it comes down to vibes.

>For example, the life sciences, an LLM could be given access to data about an organism, and then make predictions about how a drug or gene therapy will affect that organism.

Yes, biotech companies are doing this right now but ultimately they are just predictions and you still need cold hard biological data to ground them and iterate on. That's slow and expensive, especially for the juicy fields (human biology, food and crops, clinical trials) and you can't just plug billions of VC capital into the pipeline and hope for RSI. Physical (gotta procure all those labs and their equipment), human (gotta hire and pay specialists to run experiments), biological (gotta wait for organisms to reproduce, drugs to take effect, crops to grow), regulatory (gotta convince agencies that your fancy new drugs are legit) bottlenecks get in the way.

And biology is one of the easier fields where a path to RSI (if not accelerating) is conceivable. Good luck iterating on macroeconomics data where there are no replicates and 'experiments' take literal years if not decades.

>I think this is because a lot of people genuinely like the claudisms, even though a small minority of technical people don't.

My uncharitable take is that the abtruse jargon makes people feel smart for understanding it and gives a sense of belonging (as a closed circle of initiates who understand LLM cant), just like rationalists love to repackage old or unsavoury ideas under new nerdy smart sounding names.

▲

atleastoptimal an hour ago | parent [-]

By evaluations against human writers, LLM's have improved at writing.

https://vulsar.ai/benchmarks/creative-writing-v1/#cohort-ove...

>you still need cold hard biological data to ground them and iterate on... you can't just plug billions of VC capital into the pipeline and hope for RSI

Capital can get you a lot: access to wetlabs, historical medical data, ability to run simulations, creation of autonomous bio labs which run thousands of experiments in parallel. An LLM, in training on this flywheel of data can take in more information than a human biologist would be able to read in multiple lifetimes.

>Good luck iterating on macroeconomics data where there are no replicates and 'experiments' take literal years if not decades.

The better an LLM gets at simulating human behavior, the more accurately you can simulate economic policies and their effects via running simulations of agents in virtual societies. There are many routes that models can take to create a broader sense of the world without needing the world to provide data.

	▲	throwawayy6767 27 minutes ago \| parent [-]
		>By evaluations against human writers, LLM's have improved at writing. I would really take those benchmarks with a pinch of salt. We already know benchmaxxing is a thing for evals that are objectively verifiable like code, and coding models that seem to perform well on benchmarks don't pass the vibes test, imagine all the caveats that come with such a subjective test like creative writing, not to mention control issues (are they the same human writers across benchmarks? Are they the same testers?) >Capital can get you a lot: access to wetlabs, historical medical data, ability to run simulations, creation of autonomous bio labs which run thousands of experiments in parallel. An LLM, in training on this flywheel of data can take in more information than a human biologist would be able to read in multiple lifetimes. I don't disageee and am in fact excited to see the advances, what I am arguing is that it is going to be slow because the hypothetic feedback loop that would improve the model is bottlenecked. There are actual physical and biological limits in our universe. Unless VCs find a way to make mice embryos and crops grow 10x faster (among the many, many other things that would be required for acceleration) the trickle of data is not going to be anything like the accelerating returns imagined for math and coding. And that's assuming biological RSI is possible which has yet to be proven - current foundation models are interesting but they are still specialized and there is no clear, obviously uniform way to encode the many, many modalities that exist in biological data. >The better an LLM gets at simulating human behavior, the more accurately you can simulate economic policies and their effects via running simulations of agents in virtual societies. There are many routes that models can take to create a broader sense of the world without needing the world to provide data. You wouldn't need to 'just' simulate human behavior (good luck with that btw) but the entire political and economic conditions in which they live. Even assuming this is possible the amount of compute you'd need to run would be astronomical. Given how all RAM capacity for '27 and '28 is near sold out, hyperscalers are running out of cash flow to finance buildout and Sam Altman complains when users say 'please' to LLMs I don't think this is in the near horizon.