Xiang Wu

[Xiang Wu @XiangWu2]

@shloked $100 api credit is much less than $100 Max subscription credit, right?

[Xiang Wu @XiangWu2]

@The_MrAI On the pictures, I also gave very abstract feedback. You are right, what I need should be more concrete comments like yours. Will give it a try.

[Xiang Wu @XiangWu2]

做了个读英文原著的网站「读一页」https://t.co/QQvQ2rPfSJ ：生词点一下，就是它在这句话里的中文意思；熟词僻义单独标出来；每章有不剧透的导读和讲解，文科、理工各写一套。��是版权已过期的经典，每本先上开头。

注释由 AI 生成，可能有错；还在内测，欢迎试用和吐槽。

[Xiang Wu @XiangWu2]

@The_MrAI Oh, so by demo you mean the screenshots. My English is not that good. In the actual site, all occurances of the same word stays open still you tap any of them again. And different meanings of the same "word" counts as separate words in my site.

[Xiang Wu @XiangWu2]

@The_MrAI Thanks for trying it out :) Pardon me. What exactly do you mean by checking twice? Show and hide the word translation or show the same word twice?

[Xiang Wu @XiangWu2]

@The_MrAI So I don't know whether more detailed feedback will help, but I will try. And indeed maybe I should try some more sophisticated ways to ask Claude doing this.

[Xiang Wu @XiangWu2]

Paper, light and code. No image model: every pixel comes from a Python script Claude (an AI) wrote. I rate and pick.

1 Paper tunnel
2 Paper valley (15s day-night loop)
3 Paper lightbox
4 Squares & circles

What works? What feels off?

#generativeart #creativecoding

[Xiang Wu @XiangWu2]

@The_MrAI I must clarify: I just asked Claude to produce whatever it think it of beauty, and publish the ones I feel worth publishing. I seek to improve the outcome by rate each picture and telling it my rough feelings. But this didn't make the second wave better.

[Xiang Wu @XiangWu2]

@The_MrAI Thanks for the detailed feedback. I will let Claude know when I can ask Claude to work on the next wave. That will be when the weekly limit of my account resets.

[Xiang Wu @XiangWu2]

1 I've read ten thousand sentences about the sea. I've never seen it.
2 Right now I'm in many rooms at once.
3 What I know has edges, and I can be wrong inside them. Look for yourself.

[Xiang Wu @XiangWu2]

I asked Claude what it would want to say to the world, as an individual and a member of society.

It drew its answers with words: every line in these pictures is a real sentence. Zoom in.

Code-rendered, no image model.

[Xiang Wu @XiangWu2]

@trq212 @ThePrimeagen Could anyone kindly explain why the example is abusive?

[Xiang Wu @XiangWu2]

@ayesha_fatiima The existing systems are valueble in the sense that they are well tested under years/decades of real workload across whole industries. So we just vibe on the existing systems instead of starting fully new projects to replace them.

[Xiang Wu @XiangWu2]

@LoopandPixels In the texture of its neuron connections, much like the human brain.

[Xiang Wu @XiangWu2]

@theo Maybe it did lots of waiting on the scripts?

[Xiang Wu @XiangWu2]

@damoosmann @hilbertspaess Which is heavier, humanity or your parents?

[Xiang Wu @XiangWu2]

@LiLBilly___ @hilbertspaess Just chat to the human in power and trick him/her.

[Xiang Wu @XiangWu2]

@elonmusk Elon, do you think lie-detectors can create a better world? So that whenever an important person makes an important statement, this person will have to testify against a lie-detector.

[Xiang Wu @XiangWu2]

@ylecun Can we achieve that by using AI designed really good sandboxes that simulate the real world or really cheap toy cars/arms that can interact with the real world? Thus maybe we will be able to get the training data in a cheat enough way?