basch

I didn’t really specify whether the regression to the mean was from data or tuning. It’s just happening. They are creating neutral average language instead of varied expressive.

It’s always been that the voters determine reality.

Chef Watson was 2012-2018, and a really fun product while it lasted. There’s multiple books published from its output and insights available on Amazon.

I suspect it’s partially an average / regression to mean.

If you built a model only based on long form journalism and The New Yorker voice, including both the template structure of the article and the grammar and syntax, the model would replicate the patterns it is built upon. You could do the same for high vocabulary rap.

When you stuff all of English or all of human thought into it with somewhat equal weight, including very technical writing the sum is a lack of style and elegance and personality.

The second problem is that writing often doesn’t happen linearly. You write bookends and fill in, edit and revise. You sort of know where you’re going and can then make decisions how to get there. Language models may benefit from being redesigned to perform passes. Select a template of some sort, fill in the mad lib, then run a detail pass, then run a persuasion / emotion pass, then run a style pass. Current model writing very much sounds like I would expect autocomplete to.

Maybe generative AI. Chef Watson was chemical and flavor pattern matching used to discover unexplored combinations.

If you were building Cooking AI from the ground up, the chemistry and physics would be more like formal logic than plausible word vomit.

staggering has to be a better use of resources, and I would think most workflows can accommodate using a 6.0 of one model and a 6.1 of a completely different one without balking.

It's interesting that you took my ambiguity and lack of conviction regarding a right and wrong side as me picking one.

>Private industry should be able to profit off the sun and it's disturbing that you think asserting this is a dunk.

It wasn't a "dunk", you added that interpretation yourself.

Im sure theres an element of "private industry should profit off the sun because I'm a temporarily embarrassed millionaire and the government cant be trusted to run efficiently" vs "the government should build and maintain communal infrastructure so everyone can benefit and profit isnt extracted" with both sides thinking their version keeps things cheaper in the end for the customer.

But they are monetizing the export button you press before the api call is made.

That Banking website is a good example of GPT-Image-2.5 having absolutely no self control, to a point I am just shocked they havent been working to fix it. It's like the first thing the model does is "generate 100 slogans" and then try and figure out how to cram them all into the image.

You see it constantly, and its worse with GPT than Gemini models, but they both do it. They seem to think they need to tell not show. They cant not annotate their thoughts into the image. A lot of the time you see either random signs on posts added, tshirts with text describing the story, or hand written annotation over the image.

There's really no reason this should be a problem anymore on default settings, or that everybody should still be getting the exact same style of edgy poster out of GPT with icons for bullets, and lists that add nothing jammed into the image. They need to have some kind of seed library of templates they start from, and higher temperature on selecting different styles.

Like crypto, the "I dont like AI" is more "I dont like HOW people are using the tools." It's the people the author doesnt like.

at some point you can simulate decompiling a game purely from enough video footage of the game being played.

why? Microsoft used Vista to test Windows 7.

seems completely normal to release free beta software and ask for feedback.

did you test it? the photo application opened a photo, pinch to zoom worked, the menus are full of buttons. it seems like the features i tested worked. they are mostly in similar spots as photoshop/photopea.

its a little impressive that models have a compressed approximation of photoshop inside that works with minimal tinkering.

the icon for windows doesnt work.

They also develop the OS and development platforms that the apps are built upon. The apps don’t exist on islands.