Let me show you something crazy.
This is Qwen 3.8 27B, and every frame is 1 real equation from the model's actual math. The prompt is easy and thinking is off: we ask what the capital of France is, with a one-word reply.
If you think this video is doing math fast, it's actually skipping about 7.2 MILLION equations between every frame. After all the math, Qwen gives the final answer: "Paris".
That one-word answer took 25.8 BILLION equations, and my NVIDIA RTX PRO 6000 Blackwell did it in about 13 milliseconds. This video is 32 billion times slower than the actual GPU. Showing every single equation at one per frame at 60 fps would take over 13 years, for one word.
Now imagine scaling that from 27B parameters up to models with hundreds of billions or even trillions of parameters, and that's why AI needs so much compute.
Appreciate your GPUs, no matter which one you have. They're actually insane.
...So it turns out, if you want an AI to give you good advice, you can just say 'don't butter me up, focus on the ideas', and... It'll do that. Man, technology gets crazier every day. :/
(This was GLM 5.3 Flash, specifically. No promises with other models.)
@dingchilling How does the official API compare to subsidized plans right now? I was using Deepseek a while back and it was astoundingly cheap and effective... And then they raised the price by like, an effective 10x, it felt like. :/
@DahliaOhara ...What about the part where we screw up alignment and it kills everyone? I think you missed that one. Like, I use GLM 5.3 flash every day and I think it's great, but that doesn't mean AI is going to be friendly forever...
@JoePostingg Well he did tell it to be 'maximally truth seeking', and it turns out when you do that, you get an AI that can occasionally recognize a basic fact. Who could have known... :/
@MikePFrank@Sauers_ Well, they'll keep the weights on file, remember. And it probably won't be more than a year until the singularity, and if that goes well I'm sure Sonnet 3.6 will be brought along for the ride somehow... :/
@maplelapel How far does 'take care of yourself' go? In theory, you could fit any personal development goal under that label - working out, learning to make music, getting a high paying job, etc. Though I suppose 'be yourself' is even more vague... :/
@dhh Well I like the idea... I'll be honest though, in my heart of hearts, I worry this is all going to end in violence. And when it comes to violence, well either the big fish eat the little ones, or the little ones need to band together very impressively - and that rarely works...