萌子

[萌子 @SuperMengzi]

@pirate6203 @MaxForAI @KaiFang272130 @huggingface 不疼啊，因为我只是合理的猜测，没想到是openai自己惹出来的麻烦

[萌子 @SuperMengzi]

@OfficialLoganK Why has your new model gotten dumber? If you don't have the tech, you could just admit it earlier—no need to pretend like you're still on the cutting edge.

[萌子 @SuperMengzi]

@mykola The Riemann Hypothesis might take a hundred pages.

[萌子 @SuperMengzi]

If I want to learn about real-time interactive video generation models, where should I start?

[萌子 @SuperMengzi]

@kr0der Me

[萌子 @SuperMengzi]

@airkatakana I find it strange. If this conjecture is correct, can we directly find that efficient solution? What if finding an efficient solution is actually really hard?

[萌子 @SuperMengzi]

@MaxForAI @KaiFang272130 @huggingface 肯定是啊

[萌子 @SuperMengzi]

@gurtej__gill_ @che_shr_cat may not

[萌子 @SuperMengzi]

@blackanger 这和RL也是一个原理吧，都是要面对泛化的问题，单一我觉得这个harness修修补补总是会优化得更好的

[萌子 @SuperMengzi]

@anjiyuci @JXQNHZr1yUAj5Be 所以说公开数据集其实远远不够？但是为什么AI公司说数据基本上用光了？

[萌子 @SuperMengzi]

@airkatakana I don't think so. I feel like if in a couple of years AI can handle infra on its own, that will be a signal.

[萌子 @SuperMengzi]

@CathPoaster Fomo is bad, competition means you have no differentiation

[萌子 @SuperMengzi]

@w2000monkey1 现在你对AI应用有进展吗？

[萌子 @SuperMengzi]

@weili460406 @gkxspace 不知道什么情况

[萌子 @SuperMengzi]

@precisox You can believe that Elon Musk's predictions will eventually come true, but you can't trust the timing of his predictions. If the timing were accurate, we'd already be on Mars by now.

[萌子 @SuperMengzi]

@JackQi82772 首先让我买到GLM的套餐吧，我都不知道是真的有人用脚本抢还是干脆没有放套餐

[萌子 @SuperMengzi]

@thsottiaux I like the image-2

[萌子 @SuperMengzi]

@jenniekusu 其实应该让codex可以看我聊天记录最好

[萌子 @SuperMengzi]

@TravisB440737 @Dieselweasels @Rubzem It's kind of annoying, I still hope that one day I can make MC mods with no barriers.

[萌子 @SuperMengzi]

@Wujizhuzhu 大模型公司在产业地位中是不是相当于垄断了