今天开始我尝试下中英文双语版。写中文版主要是觉得还是用母语更能够清晰表达自己的一些想法；写英文版的主要目的是练习下英文写作输出。
做了什么很酷的事？#
- 终于在 App Store 上架了 灵瓜背单词 第一个版本
- 这个历时一个月开发的产品的上线让我异常的兴奋。这次 Apple 也跟给力，提交版本之后很快就给我过审了，甚至还搞的我有点措手不及，因为我后端服务没部署好。
- 这个 App 目前肯定是存在大量的问题，但这对我的长期计划来说是一个全新的开始。冰冻三尺非一日之寒，一个好的产品也需要足够时间的打磨才能成型。我坚持长期主义，之后会花很多时间来慢慢打磨这个产品，目标就是为学习语言的人提供足够的价值，使得这个灵瓜能够为语言学习者提供极大的帮助。
- 这里我也同时非常感谢现在试用这个 App 的朋友们，特别是感谢很多给我提供反馈和宝贵建议的朋友。
写了哪些文章？#
- 在知乎上水了几篇回答，顺便为 App 做做营销 [🐶]
看过哪些视频？#
- RoboPianist: A Benchmark for High-Dimensional Robot Control (Teaser)
- 这是论文 RoboPianist: A Benchmark for High-Dimensional Robot Control 的一个很有意思的动画演示。是一项关于多任务学习、零镜头泛化、多模态 (声音、视觉、触觉) 学习和模仿的研究。目前能够为 150 多首乐谱生成指法。
- https://www.youtube.com/watch?v=VBFn_Gg0yD8
发现了什么很酷的玩意儿？#
- BookAi
- 这个 App 很有意思，让你可以通过跟一本书进行对话的方式来阅读了解一本书。不过这个其实就是做了个书本封面的目录，其他的应该全部都是直接调用 GPT3.5 或者 GPT4 的 API，因为 GPT 模型训练的过程中其实相当于已经把这些书的内容学习并内置到模型中去了，所以这种完全不需要提供什么额外的信息和提供额外的处理了。
- Segment Anything
- Meta 终于也发布了他的 Ai 产品，这是一个能够识别并分割任意物体的 Ai 模型。我用很多图片尝试了下，效果非常赞。
- 下面这张图中，模型很准确的识别出了两只狗叼着的棍子，而且这个棍子中间一段还被狗嘴给遮住了。
- 下面这张图中，模型也是很容易就识别出来图片中的各种蔬菜，比如一根小扁豆
- 最后这张图里面的几只棕熊其实很难分辨，我点了下最后面那只的头和屁股（中间很大一部分身体都被另一只给遮住了），模型也是很准确的识别出来了
- Bartosz Ciechanowski
- 这是一片篇很长的博客文章，但通篇都是干货。这篇文章用非常漂亮的可交互式动画从最基础的受力分析开始讲起，最后把自行车的动力学原理、结构分析设计等等都讲的一清二楚。文章是英文，如果英语不是很好或者懒得花时间看的同学，直接看下动画就能看个差不多。
English Edition
Cool things I do#
- Released the first version of LinguaTale on App Store
- I was so excited that the first version was approved by Apple in a so short period of time. So I released it to the public as soon as possible.
- This app may not be so perfect, but it is a new start of my long-term plan. Rome wasn't built in a day and so is a fine product. It needs time to be polished again and again. I will keep developing this product and make it valuable for every one who learns a new language.
- Thanks to all the users for trying this app and giving me feedback.
Blogs I wrote#
- I wrote a few articles on Zhihu.
Videos I Watch#
- RoboPianist: A Benchmark for High-Dimensional Robot Control (Teaser)
- This is an interesting demo for paper RoboPianist: A Benchmark for High-Dimensional Robot Control. This is a research on multi-task learning, zero-shot generalization, multimodal (sound, vision, touch) learning, and imitation.
- {{video(https://www.youtube.com/watch?v=VBFn_Gg0yD8)}}
Cool stuff I discovered#
- BookAi
- This is an app based GPT-3.5. Users can chat with one 'book' you want to read, and the 'book' will answer all of your questions. You can use up to 35 different languages to chat with the bot.
- Meta just released its AI product Segment Anything
- Segment Anything Model (SAM) is a new AI model from Meta AI that can "cut out" any object, in any image, with a single click.
- I tried it on many pictures and it works perfect. As shown in picture below, it can easily recognize the stick carried by two dogs even though it was partly covered by dog's mouth.
- Another try as shown below, it can segment a single hyacinth bean.
- Picture below shows that it can easily segment a brown bear's head and butt by two clicks.
- Bartosz Ciechanowski
- A blog that give a very clear animation to illustrate some basic physics phenomenon. This is a very good resource for high school students who may be currently learning physics.
- By reading this article and watch all the animations in it, you will probably be familiar with the basic principle of bicycle and you will also know a lot of whys on bicycle designing.