Couldn't agree more. To add on I think you can learn so much by doing something incredibly stupid you're really not supposed to in an intentionally constrained environment (in a good way, like the grandparent comment). The sort of absurd things that get posted here occasionally and get a few commenters genuinely annoyed that they're doing that with their time instead of curing cancer. Plus it tends to be easier than it looks.
(I think some people have caught on to this and found out that Claude is pretty good at it.. not sure what the point of that is but whatever)
As long as AI is only able to create crap that people are embarrassed to admit they watch (which I think is likely IMO just given how it works) I think we'll be mostly fine. Trash TV didn't ruin the television industry and mobile shovelware didn't ruin the gaming industry.
I felt the television industry and gaming industries were in fact ruined by those trends. Or at least ruined in the same sense that global warming ruins our environment.
A lot of the replies are insisting AI will get better at writing with more development but I don't see it. Even if you have a mathematically perfect writing AI you still run into the same problems you would have if you handed off your writing task to someone on fiverr or something. It can't magically know what you want to say, it only has the information you gave it. A prompt complex enough where it won't get any wrong ideas has to contain as much information as the output would have.. so just write it.
> A prompt complex enough where it won't get any wrong ideas has to contain as much information as the output would have.. so just write it.
This is only if the output is fully compressed. Writing is not just about encoding the writer's ideas but also about how the reader will ingest those ideas. The writer needs to consider when to put in rests in between complex ideas to help the reader flow through the text. This suggests the LLM could be prompted by a dense complex idea to be presented with the boilerplate needed for the human mind read smoothly and without unnecessary effort.
I also believe this. Post-training LLMs with vague metrics can only be achieved with RLHF, which is not impossible, but extremely costly and difficult. Instead, companies will opt for RLVR, focusing on math and programming tasks. This pushes objectives away from writing quality; often far away. That is why older models, in my view, actually read better than newer ones. It's by design.
You can brute force it by making it try random stuff then judge itself on it. You don't have to always use an LLM's output. Sometimes you can use that plus other things to add flavor. An LLM is actually really good at judging if something is good or bad. It just has a really hard time coming up with new things. But if you had unlimited compute you can throw in some rng and whimsy and get something resembling what humans do.
You can prove that doing this will spiral training into a fixed point. There was a lot of research into getting this to work in the past, but it never truly worked well. The hope was that if RLVR was used quite a bit, and the general performance crossed some threshold, that it would then be possible. However, since it has been shown that RLVR only concentrates the distribution of outputs rather than truly shift it, I doubt this will ever be a viable strategy.
I think for this argument to be true, the axiom that supports it is that the models have just as much context as they will ever have, and you cannot see being able to give them more / enough to be able to understand your perspective. That feels unlikely to be a position that doesn't change. As a society we're giving more and more context each day to this, and that makes this a valid opinion now, but one that erodes over time.
its not about dumping more and more info into the context, its about the intention. whats not in the context is just as important as what is. and i dont see how that can be automated.
also were seeing models become worse at writing as they get smarter.
Size of context is not the entire story here, it's ability to properly feed and index the context that's needed on this sort of thing. E.g. your entire slack/discord/email/github/jira/zoom meeting/coffee chat ... history is the context that you bring to the table on this sort of thing. Most of this is unindexed. Much of this will not be in the future.
> The latest models got even worse.
Which models? This is one of those things that likely has both model and domain specific aspects that impact your experience. In my experience with OpenaAI models predominantly (I previously worked there), they've improved significantly over the last 6-12 months. My experience with Claude is worse, but I haven't spent as much time getting into a mechanical sympathy there. They're still not perfect though and I have many steering docs that help avoid the biggest problems in the models I use when generating docs.
Claude writing quality got unbelievably bad with Opus 4.7, with no improvement in Fable. Opus 4.6 was fine. Im starting to see it as a security risk - my brain just can't process its word vomit, so just tell it to go on, implement whatever
You’re using a non-deterministic algorithm to generate output. If you want deterministic rules applied to it, you have to use deterministic systems to do it.
none of the rust-based minecraft server have enough functionality implemented to use them for survival, maybe for some minigames but that's about it. Pumpkin has gotten far out of the many[1] but there's still lots to be done.
These game servers are really the perfect use case for agents. Would anyone really care that much about code quality if the server works fast and reliably?
The number of implementations and the lack of success suggest that it is a genuinely hard problem for humans, and "agents" are not going to magically make it simpler. Unlikely that humans abandoned their projects (or failed to band together) solely because of their desire for code quality.
If you have 90 minutes with nothing else to do I highly recommend giving the launch event with Jay Leno a watch.. it's really something unique. https://youtu.be/_JzfROUDsK0
This might just be because it's a demo, but man are things ever fast... they open up a document and it just instantly comes up. They load a video and bam, instant. There's a little segment where Bill Gates is showing off opening up multiple instances of Excel and they open up so quickly. I tried opening up Excel on my desktop PC and the speed is nowhere close to what this video is showing. I open up a document nowadays and there's a splash screen, there's some lag as the application checks if there's an update, the application appears and the layout starts to jitter and flicker... none of that is in this demo.
Obviously some of this is for show, but even still it indicates people cared about having a very smooth and snappy user experience back then and that it was absolutely important to showcase performance. Somewhere along the way developers just quietly agreed that latency is the user's problem.
It's not fast. Look at 19:40 when he's starting new apps. You can see each newly launched app slowly paint individual regions of their windows. At 50:56 you see Photoshop slowly load in tiles of an image. That's a strong memory of that era for me; watching apps slowly repaint bits of their windows as the HDD thrashes. And that would have been running on a machine with probably 16MB of RAM, so far beyond what most users would have.
It wasn't really Windows' fault. It was well optimized. Machines of that era were just really slow.
Windows sucks but Linux is not a very good operating system either. Everything is a bodge to cover up the assumption that it thinks its a multi-user PDP11 system from the 70s, nobody can agree on which set of bodges to use for the same thing, and it's a miracle it even works at all. But it's better than the alternatives and nobody wants to try again so we got used to it. (That also applies to why Windows is used in corporate environments too)
Well for example yours - if you don’t pay for these rare books and don’t store them in good condition, then you have decided that they are not worth saving. Many of these rare books would run you like I don’t know, 1 buck?
I don't think I have enough money or room to buy all the books I think are worth saving, and my list no doubt has some overlap with someone else's. If I did buy them, what if someone else wants to read them? If I lose it or it gets destroyed in some way, the chance it will be gone forever increases (assuming there are multiple copies). Entrusting the availability of knowledge to individuals like that sounds like a bad idea. There could be some kind of publicly funded organisation that can take care of a big collection of books, afford to keep them safe, and make them accessible to anyone.
(I think some people have caught on to this and found out that Claude is pretty good at it.. not sure what the point of that is but whatever)
reply