Review: The Grand Design

 The Grand Design came recommended by the Amazon Kindle direct mail ad, and it was available at the library so I checked it out and read it. It's an easy to read Physics primer that discusses the difficulty reconciling quantum mechanics and general relativity, and discusses what the evidence for the accuracy of these theories are.

What I enjoyed was how clearly written the descriptions of the weak Anthropic principle and the stronger versions of the Anthropic principles are. In particular, the authors go over how finely tuned the fundamental constants of nature are, and even go so far as to explain why the Universe had to have 3 dimensions instead of 4 or 6 or 10:

If one assumes that a few hundred million years in stable orbit are necessary for planetary life to evolve, the number of space dimensions is also fixed by our existence. That is because, according to the laws of gravity, it is only in three dimensions that stable elliptical orbits are possible. Circular orbits are possible in other dimensions, but those, as Newton feared, are unstable. In any but three dimensions even a small disturbance, such as that produced by the pull of the other planets, would send a planet off its circular orbit and cause it to spiral either into or away from the sun, so we would either burn up or freeze. Also, in more than three dimensions the gravitational force between two bodies would decrease more rapidly than it does in three dimensions. In three dimensions the gravitational force drops to ¼ of its value if one doubles the distance. In four dimensions it would drop to ⅛, in five dimensions it would drop to 1/16 and so on. As a result, in more than three dimensions the sun would not be able to exist in a stable state with its internal pressure balancing the pull of gravity. (kindle loc 1530)

Where the book fails is that it promotes M-theory as the one theory that would unify quantum theory with relativity, but doesn't go into why it's superior to all the other theories. One issue is that it's not a single theory, but a class of overlapping theories that can effectively have constants plugged in to satisfy the constraints of the universe we find ourselves in. The authors pretty much state that scientists have to give up on the idea of our laws of physics all deriving from one fundamental theory that fixes all the constants. That's quite disappointing for those who value elegance in their theories, but of couse, nothing says that our messy universe has to correspond to a model of a simple and elegant fundamental theory.

Review: The Woman Behind the New Deal

 The Woman Behind the New Deal is the biography of Frances Perkins, FDR's Secretary of Labor. I've read biographies of FDR before, and they usually gloss over Frances Perkins, but over time as I've read other historical reports (such as Robert A Caro's biography of Lyndon B Johnson and Robert Moses), I saw references to her over and over again.

It turns out that I'd been missing an important part of the origin of the New Deal, the first woman to serve in the President's Cabinet, and the person to whom he delegated all of the various aspects of the New Deal to. A key scene in the story is where Perkins named all the important aspects of what she wanted to accomplish before she would accept the position to FDR. In fact, she noted that Roosevelt was not a deligent person nor was he attentive to detail, but was indeed a C student:

She disapproved of the tendency at Telluride to select for admission only the boys with the highest academic grades, noting that it is often the C student with strong leadership skills and a good personality who makes a greater mark on the world. “Franklin Roosevelt would never be admitted to a first-class college today,” she said. (kindle loc 6978)

Perkins was extremely religious and also quite unlucky with her personal life, marrying a husband who was promising but who turned out to have a family history of mental illness, and after an initial setback in his career never came out of his depression and ended up in one institution after another, leaving her to be a single mom to take care of their only daughter. In that time, mental illness was considered shameful, but society also disapproved of working mothers, so she had to hide a lot of what she was doing. In the ultimate repudiation of her, her own daughter never gave her the respect she deserved, despite Perkins sacrificing so much of her own well being (Perkins herself worked until the day she died, and her family was not present when she passed away):

None of the children were ever told about Frances’s benevolence while Susanna was alive. In fact, they recalled, Susanna generally ridiculed her mother, as she did many people who worked for a living. Frances had spent her life laboring on behalf of America’s workers and had produced a child of the leisure class, more ornamental than useful, who felt contempt for people who worked for a living. (kindle loc 7106)

 Nevertheless, Perkins achieved a great deal. She introduced worker's comp, the 40 hour work week with mandatory overtime, unemployment insurance, and social security. Her next target would have been universal healthcare, but the war got in the way. What a different America we would have lived in if her agenda had been completely achieved. She turned the labor department's statistics department into what it was before Trump decided to start firing bearers of bad news:

Lubin turned the bureau into a respected source of economic statistics. He quickly improved the country’s system for gathering employment and wage statistics, and put systems in place to minimize political pressure to show positive results. For the first time, the federal government tracked hourly earnings and average weekly hours by industry. Lubin also modernized the cost-of-living index. One component in the outdated index was the price of high-button shoes, still included long after people had stopped wearing them. The BLS also tracked productivity, an important gauge of business modernization trends, and it made separate forecasts of growth in different occupations in order to steer workers to new industries. (kindle loc 2627)

Not only was she progressive about worker's rights, she was the first to recognize the threat of Hitler, and did everything in her power to help German jews escape the regieme.  This was despite the prevailing sentiment against immigration:

Frances knew restrictive policies were politically popular. “It is generally recognized that the United States can no longer absorb annually hundreds of thousands of immigrants without serious economic and social dislocations,” she wrote. “Certainly the present restrictions can not be relaxed while millions of workers are unemployed and maintained at public expense.”12 In fact, Frances had to spend much time reassuring disbelieving citizens that immigration had indeed been curtailed. Many refused to believe government statistics, and they circulated reports alleging that 1 million foreign sailors jumped ship in the United States each year, or that five hundred thousand Mexicans strolled across the border in the previous decade. In her annual report in 1935, Frances blasted these accounts as “fantastic exaggerations.” (kindle loc 3488)

Reading this book, you really see the parallels between the current political climate and the what happened in the 1930s and 40s. There was never a time when America's population was positive disposed towards more immigration, and even back then misinformation was still very much prevelant. I'm continually surprised when progressives refuse to admit that immigration is very much in tension with progressive objectives despite that long history. (Think about it: the rest of the world is generally much more conservative than American progressives --- importing a lot of immigrants will not result in a population that's willing to support progressive objectives!)

In any case, this was a book very much worth reading. If you're a feminist, progressive, or student of history you owe it to yourself to read this.

Review: Broken Stars

 Broken Stars is a collection of contemporary science fiction stories, selected and edited by Ken Liu, who's translated many science fiction authors including "The Three Body Problem." I picked it up because it's far faster for me to read English than for me to read Chinese, and obviously I'm not very in touch with Chinese science fiction (or any other contemporary Chinese fiction for that matter)

A lot of the stories in this volume aren't basically science fiction. They're effectively fantasy exercises (including one where a time traveler travels back in time and reinvents the internet), without the rigor of science fiction that I normally expect from my preferred science fiction authors like Alastair Reynolds. Even someone like Iain M Banks (whose Culture novels are set at such a high tech level as to seem magical) do pay attention to the plausibility of many details.

What you do get out of many of these stories is an appreciation of Chinese history. Lots of events in contemporary and past Chinese history is placed in context and you can see how the authors felt about those events. There's just not any "hard" science fiction or even cyberpunk in this collection.

On reflection maybe I shouldn't be surprised. China's demand for tech talent and work environment that probably a heavy STEM engineer or scientist wouldn't have time to write, so the folks writing science fiction in China come from the literary sector and won't be deeply immersed in the science.

At the back of the book is a couple of essays about the state of science fiction in China. This would be informative, but obviously the book was edited prior to the major scandal about WorldCon in China in 2024, where the voting was rigged. It seems like science fiction in China cannot be separated from the governing environment the country is in.

Review: Genius Makers

 Genius Makers is a biography/chronicle of Geoff Hinton and a number of his students, starting from Hinton's entry into the AI field until the night him and two other students won the Turing award.

The book is well written but spotty, with too many characters towards the end of the book showing up and being given barely enough time for you to fully comprehend what's happening. I thoroughly enjoyed to story about how Geoff Hinton, driven by his wife's dislike of living in the USA when Ronald Reagan was president, ended up at the University of Toronto:

At the height of the revival in neural network research, Hinton left Carnegie Mellon for a professorship at the University of Toronto. A few years after this move, as he struggled to find new funding for his research, he wondered if he had made the right decision. “I should have gone to Berkeley,” he told his wife. “Berkeley?” his wife said. “I would have gone to Berkeley.” “But you said you wouldn’t live in the U.S.” “That’s not the U.S. It’s California.” (kindle loc 751)

I left Google at 2010, which was when Google started gobbling up AI talent (including Hinton and his students) at a rapid pace and integrating it into many areas of the company. Google at that time also left China because the Chinese government hacked Google (true story --- I know the people behind the detection). Despite that, Google still thought it could come back into China after Deep Mind became the #1 go player in the world, with Eric Schmidt giving a patronizing talk about Chinese AI:

 China’s tech giants had already embraced deep learning. Andrew Ng had been building labs at Baidu for years, and, like Google, he was erecting a vast network of specialized machines to feed new experiments. Similar work was brewing at Tencent. In any case, even if it did need Google’s help, China was unwilling to take it. The government, after all, had blacked out the match in Wuzhen. It did not take long for Schmidt to realize just how naïve his message had been. “I knew when I gave the speech that the Chinese were coming. I did not understand at the time how totally effective some of their programs would be,” he says. “I honestly just didn’t understand. I think most Americans wouldn’t understand. I’m not going to misunderstand in the future.” (kindle loc 3109)

What was real obvious to outside observers (i.e., anyone with an ounce of common sense) was that the Chinese government would never let a Western company have any significant market share in the country. That Western companies one after another fell into the trap and did a ton of technology transfers to China for free at their own expense just tells you how short-sighted and badly run most Western companies are.

Another thing that comes through in this book was how one top executive (Andrew Ng, Qin Lu) after another with ties to China actually moved back to China or took top jobs at Chinese companies and helped with that technology transfer. Of course, much of this happened before the current cold war between the USA and China, but it's still pretty amazing to watch.

A lot of people I knew from my Google (or even pre-Google) days show up in this book (e.g., Jeff Dean). The depictions are sort of accurate so I do find most of the book believable. The book was written in 2021, so pre-dated the era when ChatGPT took the world by storm. Nevertheless, for an understanding of the history of neural networks and deep learning, this is a book well worth your time.

Review: Human Compatible

 I took my one and only AI class from Stuart Russell, who wrote Human Compatible. Written in 2019, this book predated OpenAI's ChatGPT and the LLM revolution, but nevertheless anticipated many modern concerns about the rise of AI. It addresses concerns such as the paperclip apocalypse with a critique of current AI approaches to problem solving.

Fundamentally, Russell's critique of the current AI approach is that the systems that are designed have an explicit goal and 100% certainty about their goals. This is appropriate if the AI system is incompetent and sucky, but will lead to bad outcomes if the system is superior in intelligence to humans and can prevent humans from interfering with its goals by turning it off.

The solution, Russell claims, is for the AI system's goals to be to be to assist the human's goals and to infer those goals from the human's statements and behavior. The inherent uncertainty about human goals will force the AI system to ask questions, and allow itself to be turned off if necessary. There's excellent analysis as why this is and why this is rational even in the case of super-intelligence.

The solution is elegant, interesting, and obviously unenforceable --- all it takes is one bad actor with super-intelligence to deviate from this principle and we'll be back at the paperclip apocalypse again. On the other hand, we're probably very far from being able to encode this sort of thinking into an AI system, so obviously this has no direct impact on current research.

Nevertheless, it's a great book that's well written and has an intelligent solution to what's widely perceived as a common problem. Recommended.

Review: Katabasis

 I have to applaud R. F. Kuang's PR team. Within the same week, I got 3 magazines in my mailbox with a profile of her: The New Yorker, The Atlantic, and Outside Magazine (!!). This was all done in coordination with the launch of her new book, Katabasis.

R. F. Kuang's superpower is skating to where public attention will be. Yellowface, for instance was about cultural misappropriation, and Katabasis is about graduate school (and University as well) as hell, timed to coincide with the decline of public support of elite Universities. Set in Cambridge (keep in mind that Kuang despite having 4 degrees is going for yet another PhD, so she's spent essentially all her life in one school or another), the universe of Katabasis is an alternate world in which magic and the study of it is a defined rigorous subject in academia, and the protagonist, Alice Law, is a graduate student under a demanding supervising professor. When that supervisor dies in a research accident, she decides the only thing to do is to go to hell and retrieve him. The other student Professor Grimes supervises also decides to join her.

This is by far the weakest link in the story. The cost of visiting hell is high (half your remaining life), and while Alice Law might have had sufficient motiviation to recover Grimes (her hatred for him is revealed later in the book), the after-recovery plan stinks and makes no sense. In fact, this aspect of the book smells a bit like an autobiography, complete with R. F. Kuang's in-real-life husband (who does suffer from a similarly dehibilitating condition) playing the role of Peter Murdoch.

The depiction of hell is kinda bland and boring. That's probably because no matter how Kuang tries to depict it, academia isn't actually hellish. Think about it: people willingly give up additional 4-6 years of their lives at low pay for a chance to get a tenure track position. If it was truly hell, you would have to pay more to get less talented people, but many of the brightest people in the world sacrifice so much to live their lives at University, and as one of my academic friends once said, "Being a tenured professor means you get to work on whatever you want which is like being retired already."

As a fantasy novel, I'm not sure Katabasis works. The problem with writers who come from out of genre to write fantasy and science fiction, is that they tend to write magic as a form of deus ex machina. There are no rules, anything goes, and so whatever happens in the novel is whatever the author can think up. This is fine if you're telling a bedtime story to a 3 or 5 year old. In a full length novel, what happens is that the reader feels that the author is unfair and there was no way for you to have seen the ending coming (especially since in this case the ending comes in the form of a gift from a character Alice Law betrayed!).

The way to predict the novel's ending, of course, is through meta-cognition --- you knew that R. F. Kuang married her husband, so the ending must involve Alice Law rescuing Peter Murdoch and them getting to live happily ever after. I guess that's why Kuang's PR team landed all those profiles of her in various magazines --- so you might not come away after reading the book feeling cheated.

Capitola Overnight

• Video from Xiaoqin's GoPro

This year the kids declined to do any biking for Labor Day, and we ended up doing a hike or two instead. But for my birthday, I wanted to revisit Jamison Creek, a road I hadn't climbed for at least 15 years, so I booked a room at the Monarch Cove Inn, laid out a route and then invited Mark Brody to join Xiaoqin and I --- we'd taken the last room at the Inn but he found suitable accomodations in Santa Cruz despite their being a triathlon on that weekend. I'd always wanted to know how the Roadini handled as a credit card load touring bike, so rather than bring my custom Strong frame (which had the Ortlieb mount on it making it a pain to remove for just an overnight trip), I installed the Carradice Bagman on the Roadini and the Carradice Nelson Longflap, which unlike many modern bikepacking bags actually have a light attachment loop mounted. The Roadini had Conti Terraspeed tires on it, but I wasn't going on any challenging off road riding on this strip, so I installed a Michelin Power Cup 28 on the front Ritchey Zeta Wheel, and a leftover nearly worn out Vittoria Corsa NEXT 34mm tire on the rear.

Loading up with myself and Xiaoqin's minimal overnight setup, the ride felt heavier than I expected but on the other hand my overnight setup included a CPAP machine so that was to be expected. Mark Brody missed his train to Mountain View the night before so we couldn't get started until he arrived on the first train the next day, well after 9:00am. Eva would also join us for the climb up highway 9.

Xiaoqin headed up expecting us to catch her, but had forgotten that she was unloaded while both Mark and I were carrying a load. The climb up Highway 9 was easy though with the weekend traffic it could hardly be called enjoyable. At the top, we waited for Eva and then Mark. Mark had suffered from insomnia and did not have a good time up highway 9, so he wouldn't be able to join us on Jamison Creek.

Descending Highway 9 was fine. We were passed by 2-3 groups of cars but they occured at places where we could pull over. Once onto Highway 236 into Big Basin the traffic petered away and we arrived at the intersection with China Grade road having marvelled at how many trees had survived the fire with just charring at the base instead of burning. Mark was lagging, though, which meant that descending China Grade despite its washed out and bumpy ride was better than taking the extra 10 mile loop through Park HQ.

Arriving at the bottom of China grade, I waited for the others. The descent was scary, but if you pretended it was a mountain bike trail it's way easier than most singletrack at any beginner's mountain bike park. You had to be willing to stay off the saddle, however. From there it was only 3 miles to the bottom of Jamison Creek Road but we were out of water and stopped at the Golf Course just before the intersection to refill our water bottles. Once we saw the menu we ordered some food as well --- a salad split with Xiaoqin and myself to avoid overloading stomachs prior to a hard climb, and some Tacos for Mark to get him to Santa Cruz via Branciforte.

Climbing Jamison Creek was a bear and I wasn't looking forward to it given how warm it was. I was pleasantly surprised therefore, to find that it was almost completely shaded. It was still warm and my shirt and shorts were completely soaked by the time I was half way up, but it wasn't the scorcher it could have been. The steep parts near the intersection with Empire Grade road required getting up and standing on the saddle while breathing hard, and I questioned bringing the heavier bike for a few moments but once we got onto Empire Grade road the grade lightened up dramatically.

Empire Grade road is much like Skyline Boulevard, with swoopy curves, rolling hills, but with much fewer vistas than the latter. The traffic, however, was much lighter which made it a lot of fun. Once we descended past UCSC we were in Santa Cruz proper, where the route took us past the boardwalk and over the bridge. From there a road closure forced us off the route into some unplanned unpaved excursions but all was well. The hitch was the final block to the hotel where I'd gotten confused as to the street the hotel was on but two phone calls with the hotel manager solved the problem, and we found the place and were checked in.

After a shower and stretching we got dressed and went to dinner at Mijo's Tacos, followed by a short walk exploring the little cute hotels that looked very European on the beach. Then we went to Gayle's bakery to eat chocolate cake and buy secondary breakfast. (Monarch Cove Inn served breakfast but it would be anemic by cycling standards)

The next morning we got up early enough to see the sunrise over the water from the Inn grounds! On the West Coast you don't expect that but the position of the Inn,, the time of year, and the Monterey Bay all conspired to give us a beautiful view of the sun over the ocean fog. Mark arrived just after we'd had breakfast and were packing. We checked out of the hotel and rode the traditional "apple pie ride" route over to Trout Gulch Road and to Corralitos before starting the climb up Eureka Canyon.

Eureka Canyon is a beautiful and lightly traffic'd climb through the redwoods to Summit Road. While the surface is bump and in many places falling apart, at climbing speeds that doesn't bother the cyclist much. The shade helped a lot and even at the top the sun wasn't bothersome. We had a quick snack at Buzzard Lagoon road and therefore skipped the summit store in favor of beelining to Los Gatos via the Los Gatos Creek trail. We even skipped Aldercroft Heights and the long way around the lake in favor of a dirt trail Xiaoqin had found earlier this. year.

After the ice cream Mark headed off to the San Jose train station while Xiaoqin and I made our way home. It was a great trip at over 100 miles in 2 days with almost 10000' of elevation gain. The Roadini more than proved itself a capable credit card touring machine.