f4
dlmair
f4b3f3c86fe3d909e2ed6949d8df26848cb839f343ae6fbbeca61f89d09aac9e

可能帮助不大。NN 只能处理 “巨大空间里一小簇有意义的点” 的例子如人脸之于图像空间, 字符串空间里有意义的子串“句子”也是极其稀少。 所以nn 不能用来解密破解md5等。 (这是我入行前思考的第一个问题, 为何nn能区分猫狗不能破解md5)

Replying to dlmair

https://www.lesswrong.com/posts/8viQEp8KBg2QSW4Yc/solidgoldmagikarp-iii-glitch-token-archaeology

比特币(和人工智能)世界真是太奇妙了, petertodd 和 gmaxwell 的名字竟然列在了GPT 的5万个基础token里面( 这5万个是整个互联网词汇按照频率排序出来来的, 一个有3个名字, 他俩位于其中, 不可思议, 虽然他俩有回复此事)

从文章第一个截图看, petertodd 是被AI宝宝👶严重保护的名字哈, 是不是他就是中本聪啊

简直是太有意思的故事了, 希望有流量有粉丝的各位打V写成故事, 我要追星🤩哈

#[0]

#[1]

我去查了下, 里面不少人名字如川普等

https://www.lesswrong.com/posts/8viQEp8KBg2QSW4Yc/solidgoldmagikarp-iii-glitch-token-archaeology

比特币(和人工智能)世界真是太奇妙了, petertodd 和 gmaxwell 的名字竟然列在了GPT 的5万个基础token里面( 这5万个是整个互联网词汇按照频率排序出来来的, 一个有3个名字, 他俩位于其中, 不可思议, 虽然他俩有回复此事)

从文章第一个截图看, petertodd 是被AI宝宝👶严重保护的名字哈, 是不是他就是中本聪啊

简直是太有意思的故事了, 希望有流量有粉丝的各位打V写成故事, 我要追星🤩哈

#[0]

#[1]

基于这两篇文章(对 GPT tokenizer 的 “奇怪/不可言说” 字符的调查), 我已经相信 PeterTodd 就是我们的 Satoshi Nakamoto :)

https://www.lesswrong.com/posts/aPeJE8bSo6rAFoLqg/solidgoldmagikarp-plus-prompt-generation

https://www.lesswrong.com/posts/8viQEp8KBg2QSW4Yc/solidgoldmagikarp-iii-glitch-token-archaeology

马斯克买下twitter之后把 npr 这个标准主流媒体“治”的不轻。 前几周因为给npr打上标签从此恩怨加深, npr在这篇播客里面提到自己的故事, 他们给twitter发邮件,对方只回了一个 “💩” 的 emoji。 马斯克太秀了

https://www.npr.org/transcripts/1175656697

一个功能尤其是新功能的“滥用”漏洞是风险很大, 潜伏很久的。 如这里的一个比特币消息签名都会被交易所滥用来证明地址的归属:

#[0]

一个功能尤其是新功能的“滥用”漏洞是风险很大, 潜伏很久的。 如这里的一个比特币消息签名都会被交易所滥用来证明地址的归属:

note1cw6l46qcxvtq75j483zrmfld7vy0fav6g8wkaaq4pw0sh2kaacfs5y7x8x

今天被广泛讨论的投资 app 有知有行,在他们 appstore 介绍里显示:

‘’’

【使命】 我们憧憬这样一个世界:人们不因理财而焦虑,更好地专注当下、安心工作、享受生活。

‘’’

可,这不正是比特币的能力吗

The original reddit post meant that OP_VAULT could be misused as a new type of multi-signature. The correct way to use it is to have both paths belong to one party. But there is a possibility of misuse, one paths belongs to the exchange and one path belongs to the user. Exchanges will promote such addresses on their apps and wallets, and this promotion is easy

1. bip345 introduces 'covenant' a new utxo, non-deterministic utxo, doesn't it?

2. from GitHub and Twitter comments, bip345 behaves too hurry, but 1 month ago the author made a lot of changes with the help of others didn't he?

3. bip345 introduces new features, in terms of security comparisons, any progress over timelock-multi-signature address? And bip345 repeatedly disparages the time-lock protocol(bip65), and claims itself to be the 'perfect ultimate wallet solution', which is too suspicious, isn't it?

4. bip345 needs a watch-tower all the time to look at the on-chain txs, how can an individual do that? Technically(mobile phone) not allowed, time (need work, sleep, vocation) not allowed for an individual, isn't it?

Some thoughts

1. bip345 introduced a new utxo which is new, unprecedented, and not in a minor scope. Change is Too big.

2. bip345 seeks fast activation, and inevitable activation, but after only one month, it seems that the author has made significant changes to it, including detail implementation. Too hurry.

3. only discuss the benefits without reciprocity, and avoid discussing the disadvantages of its misuse. Compares are Not even for other time lock method

4. It is very beneficial for institutional users, but not friendly enough for individual users, which exacerbates inequality. A simple one like watch tower is not practically implementable for most of the individual users

说的是 ios 版本吗, 第一个版本我测试使用时候, 发现用一会儿后UI 很卡, 同时测试右滑回退是可以一直回退的。所以猜测(类似 cordova hybrid 的app) 一直会 stack push 每一个新的 webview, 导致性能变差