ghosted @ghost@fairground.moe

@niconiconi I love it when you "just want to draw some boxes", and end up inadvertently unearthing some algorithmic eldtritch horror. Usually in these cases, the problem has been solved a grand total of like, once, fourty years ago, in Fortran and everybody just keeps using that one solution.

It's a fallacy to assign personhood to a mathematical model that happens to be able to produce convincing human language. But it is also a fallacy to deny personhood just because a system has a purely mathematical foundation -- ultimately, we, our brains, also have to emerge from the very much mathematical physics substrate of this universe.

A Python programmer learned Lisp and Haskell. This is what happened to their code.

I'm working on a blog post

僕、日本語中国語両方下手、只偽中国語本当上手思。何時本物上手成良毎日思。

@leo_song@mastodon.social 有研究者认为语言模型局限性太大，并无法实现 AGI。但是这种暴力堆砌的方法（将人类的每一种能力，例如语言、逻辑、情绪，都训练为一个巨型模型，然后用胶水语言将它们连接起来，并实现人类开发者给定的目标），有可能可以实现一种“封闭式准人类”。这种“封闭式准人类”将会非常擅长将人类曾经产生的东西重新组合，也可能替代专业工作者，甚至可能会产生某种简化版本的“意识”，但无法作出真正有创新的科研或艺术成果。https://bengoertzel.substack.com/p/the-rise-of-the-closed-ended-quasi

看完这篇文章，我感觉我就是“封闭式准人类”……