肯德牛

@StarKnight 有点萌

完成版

幼儿园老师又一力作。 纸板箱、纸杯、卷纸芯和鸡蛋包装做的鳄鱼🐊

@ursatong 还有必须五秒不动手机，稍微旋转一下就跳转的，跳过按键颜色很淡而且判定很小。。

先祝大家龍年快樂!!很極限的關窗 把擱置已久本來想畫的大地圖版本給畫出來了 加上今年是龍年，畫一堆龍來應景一下(? 這次做成90x40的滑鼠墊，一直敲碗的大尺寸來啦!! 在FF會登場，只是量就不保證夠了(廠商咳血) 但是之後會上架網路通販，請大家不用擔心 #MHWI #魔物獵人

带猫咪去村口看了一次舞狮，回来就这样了。😂😂

我要把这个消息告诉每一个人

I’m gonna give 10 random people that repost this and follow me $25,000 for fun (the $250,000 my X video made) I’ll pick the winners in 72 hours

Woo Hoo! An Open Source Model From Google 💃💃 Finally, we have a very cool open-source model from Google DeepMind! Alpha Geometry solves Olympiad-level geometric problems and is trained fully on synthetic data!! More proof that we can solve for reasoning in LLMs using synthetic data! AlphaGeometry is cool because it combines the predictive power of a neural language model with a rule-bound deduction engine, which works in tandem to find solutions. One of the holy grails in machine learning has always been to combine predictions with some hard constraints/rules. Their blog post states, "AlphaGeometry is a neuro-symbolic system made up of a neural language model and a symbolic deduction engine, which work together to find proofs for complex geometry theorems. Akin to the idea of “thinking, fast and slow”, one system provides fast, “intuitive” ideas, and the other, more deliberate, rational decision-making. There is definitely some merit to this approach, and it's very possible that we will need LLMs combined with deduction engines in the future. Interestingly, previous approaches to encoding rules in ML models basically encoded the rules by creating enough synthetic data that conforms to that specific rule and then relying purely on the predictive model. TLDR: we can create very powerful reasoning engines with synthetic data and deduction engines. Can't wait to dig in. Also, THANK YOU GOOGLE! So good to see them going back to "not being evil" and open research. (link to blog post in alt)

周五的晚上愉快的阅读了 Neil J. A. Sloane 关于 OEIS（整数数列线上大全）前世今生 50 年的回忆文章：“A Handbook of Integer Sequences” Fifty Years Later 他 1964 年在康奈尔念研究生研究神经网络的时候，遇到了一个整数数列，他想知道通项公式以便估计增长速度。 link.springer.com/article/10.100…

网友“椰仔”在平遥电影节偶遇余华，可惜身边没带余华的书。翻了翻只有一本贾樟柯的《贾想：贾樟柯电影手记1996—2008》。 余华：“我看看你们拿的什么书？贾樟柯的呀，那我能签。” 于是流利地签上了。

我叉腰站在客厅，闺女云：像那个花瓶。 哪个？ 这个！

ティアキン地方 #ゼルダの伝説 #ポケモン

I described some of the most beautiful and famous mathematical theorems to Midjourney. Here is how it imagined them: 1. "The set of real numbers is uncountably infinite."

哇！！！有生之年系列啊！！！数学家/计算机科学家们终于找到了单块的，仅凭形状的非周期铺砌砖块！了不起了不起！

重新解构中文后，还是挺开脑洞的…

@ursatong 这人谁啊

Just a reminder to @elonmusk and others reading this tripe: The journalists that Musk suspended did not report his real-time location.

昨天晚上我在想啊，以后不能见一个 puzzle 就买一个，这经济实力不允许啊。然后就想应该做 3D 打印，然后就发现有人做了，并取得了 puzzle 发明者的许可，允许个人非商业使用。 puzzlehub.org/ppp 更巧的是我今天收到了 YouTube 推送，就是站长在 G4G14 上的演讲： youtube.com/watch?v=w5rics…

原推例子应该是2013年的Marte Dalelv案。UAE法庭在这方面臭名昭著，不胜枚举：2003年法国女子被三人强奸被判18个月；2008年澳大利亚的Alicia Gali 被三名同事轮奸，浑身淤青肋骨骨折，仍被法庭按通奸判刑一年；2016年又有英国女子被轮奸后报案被捕。这些还都是因为受害者是外国人才得到外界关注支持。