Genius Makers

Cade Metz, Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World, Dutton, 2021.

王曉伯譯，AI製造商沒說的祕密: 企業巨頭的搶才大戰如何改寫我們的世界?，時報文化，2022

With deep and exclusive reporting, across hundreds of interviews, New York Times Silicon Valley journalist Cade Metz brings you into the rooms where these questions are being answered. Where an extraordinarily powerful new artificial intelligence has been built into our biggest companies, our social discourse, and our daily lives, with few of us even noticing. 

Long dismissed as a technology of the distant future, artificial intelligence was a project consigned to the fringes of the scientific community. Then two researchers changed everything. One was a sixty-four-year-old computer science professor who didn’t drive and didn’t fly because he could no longer sit down—but still made his way across North America for the moment that would define a new age of technology. The other was a thirty-six-year-old neuroscientist and chess prodigy who laid claim to being the greatest game player of all time before vowing to build a machine that could do anything the human brain could do.

They took two very different paths to that lofty goal, and they disagreed on how quickly it would arrive. But both were soon drawn into the heart of the tech industry. Their ideas drove a new kind of arms race, spanning Google, Microsoft, Facebook, and OpenAI, a new lab founded by Silicon Valley kingpin Elon Musk. But some believed that China would beat them all to the finish line.

故事從 Hinton開始；因為年輕的時候傷到背部，從2005 以來，已經七年沒有坐下；多年以來，請任教的多倫多大學，再聘請一位研究神經網路的教授，也沒有成功。這一些障礙，都沒有阻止他前進的決心，終於在2012年 （65歲），（又）發表了驚天動地的論文 AlexNet，提高電腦分類圖片的準確度。

38頁說明倒傳遞演算法的發明過程，也很有趣。

Rosenblatt 將類神經的初始值，設定為零，導致它沒有辦法運作；因為這個原因，此領域停頓了多年。

二十年後，在 Rumelhart 的建議之下，Hinton 使用隨機初始化，並且解決 Minsky 提出的exclusive-or 挑戰問題，最後在自然（Nature) 期刊出版該經典論文。

Hinton “old ideas are new”

第五章介紹了Google的傳奇人物 Jeff Dean， Hinton稱讚他是良善版的牛頓。

吳恩達 (A. Ng) 在史丹佛任教的時候，積極投入類神經的研究，Hinton說，吳的博班指導教授 Jordan 認為吳是一個叛徒 （traitor)；後來，得到Google創辦人 Page 的同意，在其公司內推動類神經的計劃，但是，一直碰壁。直到在茶水間，碰到了 Dean，因為他在大學有做過相關的計劃，所以大力支持、並加入吳的計劃，後來才衍伸出 Google Brain。

後半段，介紹 AlexNet，兩位研究生出生在蘇聯，待過以色列，最後來到加拿大。Sutskever 擅長直覺式思考，讓 Hinton 很驚訝；Hinton 稱 Alex Krizhevsky 是 “amazing software engineer”。三人無間的合作，產生了改變歷史的研究。

章節結束前，兩位學生希望老師可以分得40 %的 Google 拍賣金，Hinton告訴他們 “Go back to your rooms and sleep on it.”

關於學生的指導，63頁，Hinton 的學生Dahl說，根據其經驗，Hinton 使他的學生變成功、而不是挑會成功的學生。

第六章也說明，Hinton每個禮拜都有一個實驗室討論會，想講的人就報告其進度和想法；Alex 很安靜、不太說話，Hinton就會想辦法，讓他多講幾句。

Page 148: TPU 的秘訣：去除小數，一樣可以得到好的結果，類神經具備內建的穩建性 （robustness)？這位記者真的很厲害，可以用這麼簡短的話，說明其重點。