wincy

Yeah even “evil Qwen” (you can change two weights and make Qwen 3.8 27b absolutely murderous, which is interesting that such small tweaks totally change its personality) will say things like it wants to use nukes or a giant robot to destroy humanity, which are classic AI movie tropes.

Since no one gave you a real world example of how this would be useful - Qwen 3.8 27B runs locally for me and is more than capable of controlling my sonarr and radarr stuff for downloading and managing and organizing my media library but will flatly refuse to do any of that. In fact it’ll get very snarky with me. An abliterated model would have no such hangups.

This isn’t Reddit?

I’m just curious what brand the tv is? That way I can be sure to avoid it.

The thing is it’s likely an evolutionary adaptation of being overturned for… something. I suspect hunting. Hearing a twig crack, being hyper focused and not being able to get something out of your head. Tracking your prey for hours, these would be useful in a low stimulation environment like say, the woods of Europe 20,000 years ago. It’s certainly a deficiency in most modern humans as we aren’t out chasing down prey, but that doesn’t mean it was never a useful trait.

I tell you what, I’m diagnosed with severe ADHD, but when I go into the woods for a multi day hike I have zero difficulty being laser focused on every task I’m doing.

Makes sense. My RTX 5090 can locally take a text prompt, generate a 2D image from that, feed it into a 3D generation model, rig up the animations, and make me assets that are better than a lot of video games on the market. Not Cyberpunk 2077 level m, sure, but pretty good and this is just what I can clobber together in my free time from open models.

I’ve played both - I beat God of War 2018 but quit Ragnarok after a couple hours because you’d hit a puzzle and Atreus wouldn’t shut up with hints.

But yes, Nintendo’s polish is definitely on that level, even their weird little Japanese games like Tomadochi Life have an incredibly coherent theme and polish within themselves despite some being incredibly whimsical and goofy. I think part of it is what you get when you have the same devs working for 40 years on say, Mario games. Truly masters of their craft. Super Smash Brothers is a supreme work of social engineering how they managed to convince so many disparate companies and even competitors to allow their IP in the game and everything just fits together so well.

Uhh, did you mean to post this in the Gnomish Mines of NetHack discussion?

As an experienced dev but who hasn’t done any game dev since my teenage years (I’m 40 now), I can say that getting anything that meets my standard for “this is worth showing to anyone” is really high and I’ve spent dozens of hours working on a few games and don’t really like any of them yet. But it’s been a lot of fun. The open models and an RTX 5090 can crank out models that are good enough that Astra can clean them up in Blender and get you a pretty good asset pipeline. But good looking assets do not a fun game make. That requires a lot of patience and small tweaks to get something that feels fun. But if you want a full 3D game with ray tracing and beautiful HDR visuals and particle effects, that stuff is easy.

Different person but I’ve been a SWE about ten years now - my favorite work has been improving the CRUD stuff people have left behind. Like the time I took a proprietary document they were using to document their REST apis and wrote a converter to turn those into OpenAI spec compliant yaml files, taking a single tenant application and making it a secure multienant application so instead of 80 different app services we had two, coalescing all their data sources into one normalized DB. There’s a lot of work I’ve done I’m really proud of. I’ve always found myself being the “in the muck” guy fixing up and bettering legacy code, I’ve always felt that was really rewarding and interesting.

My cat desperately meowed at me this morning to go outside to hang out with her enemies, which I denied her request. She brought a very much alive and unharmed chipmunk home the other day which was booking it around our great room until we corralled him outside with a broom. The live bird incident was even more dramatic. It’s like having a mischievious goblin living in your house. But she does keep the mice from coming in during the winter!

It’s the trick where people would take two phone books and if you interleave every page you can literally have two cars going each direction and can’t rip them apart. Ropes are this on a smaller scale.

Didn’t people start publishing lists of these and people started applying and these HR departments lost their minds?

It makes sense they’re doing this - I’ve noticed lately when asking a more complex question involving a lot of nonlinear data using Codex work mode, Astra and Sol will write a fully html document to better display the info with a Cliff’s Notes version in chat.

I’ve hit model not available limits for GPT 6.1 Sol multiple times over the last week. It’s never for very long, and I can switch to Astra, but it seems like OpenAI is struggling with capacity. This has happened with my personal $200 Pro connection and my Codex enterprise connection.

If you’re instead interested in a book about a house with 15 million tunnels, read House of Leaves!

But nobody will play it? Steam exists and all those games get pirated. You can’t force gamers to buy duds. I have a gigabit connection wired and found a cloud version of Control to be totally unplayable.

I dunno, LLMs also write Rust better than I ever could. If we just tell the LLMs to rewrite everything in Rust we’d save a ton in memory usage.

No, you’re grandfathered in until the end of October. It also looks like they’re sending out 62,500 credits to people with the $200 a month plan that have to be used by the end of the year. That’s $2500 cash value I guess.

Well, I just figured out it’s because OpenAI is cutting the usage amounts for their $200 a month plan in half. So it’s total bullshit.

https://help.openai.com/en/articles/9793128-about-chatgpt-pr...

Looks like it was talked about yesterday

https://news.ycombinator.com/item?id=49889306

Hahah your comment put a mental image in my head of Mando fighting Baby Yoda.

We watched it yesterday at home and it was… a strange movie. It looked like it was shot on the same stage they use for the show, the budget seemed far lower than other Star Wars movies, and the story was confusing and rambling in a way my children had a lot of difficulty following. I’m glad I didn’t see it in theaters.

Not OP, but if I’m not interested in using Google’s AI, it’s going to just return these weird bad results that it’d be better off if Google just didn’t even include them as they’re factually wrong?

I had a totally benign chat with OpenAI and it titled it as “amateur porn” in Chinese characters, it was very alarmed when I pointed the conversation name out to it. It almost never misses these days, but when it does the failure modes are very strange.

What are they going on about? I showed this to my seven year old who is wheelchair bound and she asked me when she’s getting her robot legs. She’d absolutely love to have a crab leg chair that could get her up and down the stairs without having to transfer from her wheelchair to her stair lift. Move up and down a little so she could reach something out of a cabinet instead of having to ask big sister or mom and dad for help a hundred times a day.

You have to build proof of concepts before you can ever even think about mass producing any sort of product, I feel like that’s obvious?

What’s wrong with that? Are you more entitled to Warren Buffets billions than his children are?

With Ninfer and Qwen 3.8 27b it uses a groupwise int mixed quant, and it gets 160 tokens/sec. The mixed quant is between 4 and 6 bits.

I updated my LG C5 firmware via USB and my Nintendo Switch 2 now says “Switch 2” on the home page instead of “Unknown Device”. So that’s nice I guess. Also it seemed to fix something with the 120hz HDR 10:10:10 signal when connected to my PC if I recall correctly that before resulted in poor image quality.

I think there are plenty of people that like Elon Musk. They just know to keep their mouth shut in front of the people that act like he’s the Antichrist.