标题: A Tale of Two Reviews (on Same Book) [打印本页] 作者: choi 时间: 昨天 09:25 标题: A Tale of Two Reviews (on Same Book) The book in question: Stephen Witt, The Thinking Machine; Jensen Huang, Nvidia, and the world's most coveted microchip. Viking, Apr 8, 2025
(1) Artificial intelligence | Parallel Logic. The Economist, Apr 12, 2025, at page 80.
Note:
(a) "The firm -- and the man who turned it from a pedlar of graphics chips for computer gamers into the semiconductor titan * * * other tech bosses skulk behind a PR firewall * * * His rambling disquisitions on earnings calls * * * Their preferred name, Nvision, was taken by a toilet-roll manufacturer."
(i) English dictionary:
* peddler (variants or less commonly pedlar) https://www.merriam-webster.com/dictionary/peddler
* skulk (vi; Did You Know?) https://www.merriam-webster.com/dictionary/skulk
* disquisition (n; from "Latin [noun feminine] disquisitio [investigation], from [verb] disquirere to investigate, from dis- + quaerere to seek"): "a formal inquiry into or discussion of a subject : DISCOURSE" https://www.merriam-webster.com/dictionary/disquisition
* siloed (adj): "kept in isolation in a way that hinders communication and cooperation : separated or isolated in a silo" https://www.merriam-webster.com/dictionary/siloed
(ii) toilet paper https://en.wikipedia.org/wiki/Toilet_paper
("The bundle, or roll of toilet paper, is specifically known as a toilet roll," among other names)
(b) "Its success stems from an early embrace of two ideas on the fringes of computer science. The first, in the 1990s, was parallel processing, which breaks a big task like rendering a scene in a computer game into many smaller ones. The second was neural networks -- an approach to AI which mimics in silicon how human brains work. Once relegated to simulating backgammon, today they underpin all LLMS."
(i) Mesh Flinders and Ian Smalley, What Is Parallel Computing? IBM, July 3, 2024 (dating per Google) https://www.ibm.com/think/topics/parallel-computing
("While parallel processing and parallel computing are sometimes used interchangeably, parallel processing refers to the number of cores and CPUs [hardware] running alongside a computer, while parallel computing refers to what the software does to facilitate the process")
(ii) neural network https://en.wikipedia.org/wiki/Neural_network
("In machine learning, artificial neural network * * * ")
(c) "In order to be agile, the firm does not have a siloed structure."
(i)
(A) the siloed is defined in Note (a)(i).
(B) silo https://en.wikipedia.org/wiki/Silo
(view photos)
(C) How grain elevator works:
• Export Elevator Overview. Agricultural Marketing Service, US Department of Agriculture, undated https://www.ams.usda.gov/resources/export-elevator-overview
("Unloading Grain From Barges [which is sectional heading][:] The majority of grain exported from the United States is carried by barge down the Mississippi and Columbia Rivers to export elevators, where it is unloaded and stored until it can be loaded onto ships. In addition, some grain is transferred directly from barge to ship, without being stored in an elevator. The following diagram shows commonly used equipment arranged in a typical layout, although each elevator has its own unique layout")
Imagine the grains or soybean not in a barge but in a truck from harvest directly to silo.
• "A belt tripper is incorporated onto a conveyor in order to 'trip' the material off the conveyor at specified locations between the terminal pulleys.": from the Web.
• Dan Boxter, Belt Trippers vs Plows. FEECO International, undated https://feeco.com/belt-trippers-vs-plows/
View the right panel in section 1 Belt Tripper: [illustration is atop the text in the attached pdf]
作者: choi 时间: 昨天 09:26
(2) Marc Levinson, From Denny's To Dominance; Nvidia's chief executive is not an easy boss, but he recognized that the company's chips could power innovations in artificial intelligence. Wall Street Journal, Apr 8, 2025, at page A15 https://www.msn.com/en-us/money/ ... minance/ar-AA1Ct0Co
("Mr Huang's disdain for formal organization, leaving him with dozens of direct reports. In Mr Witts telling, Mr Huang is prone to blame others for failures, abusively dressing down employees in public. Once he became CEO, 'he regularly began to blow his stack' ")
-------------------
Thanks to the arrival of large language models, Nvidia has become the darling of the stock market. If you purchased one share a decade ago and held it until today, your investment is worth roughly 200 times what you paid.
An American success story? For some investors, absolutely. On the other hand, according to Yahoo Finance, if you had purchased a share at the start of 2002, betting that the dot-com bust was over, your holding would have been worth only two-thirds as much a dozen years later—and even less after adjusting for inflation. While the compound return has been breathtaking since its 1999 initial public offering, Nvidia has been an investor’s nightmare for years at a time.
Every company has its ups and downs, but such lengthy stretches of poor returns might lead one to think there were a few management mistakes along the way. One virtue of Stephen Witt’s “The Thinking Machine” is that it is not entirely admiring of Jensen Huang, Nvidia’s first and only chief executive.
Mr. Witt, a business-and-technology journalist, appropriately credits Mr. Huang for his abrupt decision in 2013 to transform Nvidia from a designer of graphics chips for videogames into a pioneer in semiconductors that can process billions of computations for artificial intelligence. But as “The Thinking Machine” makes clear, Mr. Huang may not be the best exemplar for aspiring CEOs.
Mr. Huang’s biography, at this point, is well known. Born in Taiwan in 1963 and raised partly in Thailand, he came to the United States around the age of 10. As a result of a relative’s misunderstanding he landed at an elementary school in tiny Oneida, Ky. When his father finally found work in Oregon, young Jensen moved there, attending public school, excelling in table tennis and choosing Oregon State University rather than a more prestigious and costly institution.
He took a job sketching semiconductor designs on paper for Advanced Micro Devices. Two years later, in 1985, he joined LSI Logic to create software for chip designers, and rose to head a division with $250 million of revenue, even as he was taking night classes to earn a master’s degree in electrical engineering from Stanford.
Perhaps intensity is the norm in Silicon Valley, but Mr. Huang comes across as a brilliant grind. He and his wife, Lori, also an electrical engineer, “worked constantly, traveled rarely, and barely socialized outside the semiconductor industry,” Mr. Witt reports. As a young manager, he was known for his sharp criticism of others, often to the point of insult: “He didn’t have much patience for people who disagreed with him.”
Sun Microsystems was an LSI customer, and Mr. Huang became friendly with two Sun chip designers, Chris Malachowsky and Curtis Priem. At a now-famous 1993 meeting at a Denny’s restaurant in San Jose, Calif., the three decided to start a company to design chips for videogame consoles. Venture capitalists provided funding. Two bedrooms in Mr. Priem’s condo served as the initial office.
Its first chip, released in 1995, was a dud. As Mr. Huang confesses, “every single decision we made was wrong.” Most of the staff was laid off. Only in 1997 did Nvidia release a profitable product, just in time to avert a financial crisis. “Our company is thirty days from going out of business” became Mr. Huang’s mantra.
Other bet-the-company decisions followed. Some paid off, others did not. GeForce, a semiconductor launched in 1999 for gamers who wanted to speed up their computers, was an instant hit, but a revised version, introduced in 2003, failed badly. Around the same time, Mr. Huang directed Nvidia to develop an ill-fated communications chip. An ambitious product repurposing videogame technology for scientific users proved to be a fiasco when it was introduced in 2006.
How much responsibility Mr. Huang bears for such disasters is not clear from these pages. Mr. Witt does not blame Mr. Huang for a 2002 Securities and Exchange Commission investigation that forced Nvidia to restate three years of earnings and led to the ouster of the company’s chief financial officer. One wonders whether these problems had something to do with Mr. Huang’s disdain for formal organization, leaving him with dozens of direct reports. In Mr. Witt’s telling, Mr. Huang is prone to blame others for failures, abusively dressing down employees in public. Once he became CEO, “he regularly began to blow his stack,” Mr. Witt writes.
In 2013 Mr. Huang met with an Nvidia engineer named Bryan Catanzaro, whose research on neural networks had put him in danger of a “Requires Improvement” rating. Mr. Catanzaro described how AI was advancing quickly. After intensive study, Mr. Huang became a convert. He declared Mr. Catanzaro’s work to be the single most important project in the company’s history, offering Nvidia a once-in-a-lifetime opportunity. It proved to be so, helping make Nvidia one of the most valuable companies on the planet.
Mr. Witt is adept at explaining the hardware and software behind AI. Lay readers mystified by parallel processing and large language models will find “The Thinking Machine” worth reading. As a history, however, the book is incomplete. Mr. Witt draws extensively on interviews with current and former employees, who seem to both worship and to fear Mr. Huang, but the author doesn’t appear to have had access to company archives or sensitive documents. This may reflect the proclivities of Mr. Huang, who is quoted in another recently published book, “The Nvidia Way,” as saying, “I don’t love talking about our past.”
Mr. Huang also does not like talking about the potential risks of AI, currently a matter of extensive public discussion. When Mr. Witt raised the subject in a final interview, Mr. Huang’s “anger seemed uncontained, omnidirectional, and wildly inappropriate.” When Mr. Witt asked other Nvidia executives about the topic, the author laments, “the executives were more afraid of Jensen yelling at them than they were of wiping out the human race.”
Mr. Levinson is the author of “The Box: How the Shipping Container Made the World Smaller and the World Economy Bigger” and other books.
