Discovering errors in Donald Knuth's TAOCP(glthr.com) |
Discovering errors in Donald Knuth's TAOCP(glthr.com) |
Previously, I had the impression that Knuth was some magical figure of perfection. What I realized was that the truth was more surprising: he makes the same kinds and the same frequency of errors as anyone else (in fact likely more, because every page is packed with a lot of detail); what sets him apart is that he cares so deeply about getting everything right — he has basically invited a DDoS on his time and attention, where every person in the world is strongly encouraged to write to him with errors on every page that he has ever written over several decades — and he does go back and look at all of them carefully (example: https://retrocomputing.stackexchange.com/questions/18117/whi...); and despite all this he somehow continues to function, producing new pages at a nonzero rate. After some experience with his responses (handwritten with pencil on a printout of the emails sent to him), I came away even more impressed.
Putting this here because I don’t know where else to tell this story and I simply have to share:
I flew into Munich a few months ago, and since I have an EU passport I was able to use the shorter self-scan line for EU citizens at passport control. In front of me was an older couple who was having trouble with getting their passports to scan. I noticed they had US passports, so I politely informed the gentleman that he was in the wrong line. He turned around and thanked me, and that’s when I noticed the guy looked exactly like Donald Knuth! Before I could compose myself enough to say anything though, he and his wife had left to find the correct line. Obviously I couldn’t be sure it was him, but I checked Knuth’s public schedule and he was apparently speaking at some conference in Venice about two weeks later, so I’m guessing he decided to come a bit early and enjoy a vacation before the conference.
So yeah, I’m like 99% sure I technically corrected an error by Donald Knuth, but unfortunately I missed out on my reward check. Didn’t even get a selfie with him.
Somewhat recently we spotted him at a Hitchcock movie festival at Palo Alto, which my wife and I were attending.
Random run-ins are surreal :-)
Speaking of errors, in the lecture he once forgot to save a file in Emacs and was confused about the compilation result briefly. Made him very relatable.
1. You are unlikely to find errors in the algorithms themselves, especially if they've been officially published. You might find some infelicities, but these are not counted as full errors. For example, the author here found some confusing-but-not-wrong comments about local variables and unused registers. These are counted as "suggestions" (worth 0x20¢) rather than "errors" (worth 0x$1.00).
2. Knuth is pretty generous with credit -- if your suggestion leads him to find an error, you get credit for the error. The author here said that some defined variables went unused. Knuth pointed out that those variables were in fact used in an exercise. However, in looking this up he noticed a variable-related error in that exercise. Author is credited with 0x$1.00!
3. Exercises are more likely to contain errors and infelicities than the main text. And there are an awful lot of exercises.
4. Knuth includes a whole bunch of stuff in his books that is not related to CS. Lots of weird trivia and references. This stuff is more likely to be wrong than the main text. For example, Knuth mentions "icosahedral objects inscribed with Greek letters" and includes a reference to an article in the Bulletin de l’Institut français du Caire. But the author points out that the article is actually in the Bulletin de l’Institut français d’archéologie orientale. Whoops! 0x$1.00 for you!
Bletchley Park now has a prominent bust of Rejewski that credits his accomplishment. The Polish wisely passed on their knowledge to the British to keep the intel safe, because they expected a German invasion.
On another note, I hope Professor Knuth has a continuity plan in place that ensures that his book series gets completed despite his advanced age (I'm worried about that, but tact prohibited me from asking, of course).
And that's perfectly fine.
He should make them $5.12 so it’s a check for 1,000,000,000 cents, instead of 100,000,000.
- don't read them
- get all the books, put them in a bin bag and shake vigorously with some lumps of coal, to give them that "used" look
- go through the books, underline things at random and make notes (also at random) in the margin such as "how true", or even better "wrong!"
- put books on shelf in office - never look at them again
This has worked for me, though I must admit that Searching & Sorting and stuff about random numbers are pretty good.
I got my copies free from Addison Wesley for doing some book reviews for them - not reviewing Knuth, needless to say!
Maybe an LLM winning a Knuth bug bounty is a milestone on the path towards “AGI”?
https://en.wikipedia.org/wiki/The_Art_of_Computer_Programmin...
Kudos to everyone who found bugs, and got the bounty. Congrats to everyone who used these to get their PhD-CS
They put out a Unicomp clicky keyboard a few days later, picked up that too.
Need to find another error so I can get an account....
Like I know it is capable. I know how to set up the hand feed for it (and I successfully do that for thicker paper). And ostensibly I know the right dimensions and how to line it up in a PDF. But then it's just failure after failure, and I go back to cut and paste or handwrite.
He should have used TeX ;) I've heard handles such layout problems automatically.
An AI application might make a good search interface for Knuth's opus.
I later got a check for identifying a minor issue with the early history of superimposed coding. I happen to have copies of the relevant patent case containing examples predating Mooers' randomized superimposed coding.
("Happened to" because I had visited the Mooers archive at the Charles Babbage Institute in Minnesota to research some of the early history of chemical information management. Mooers is one of the "fathers" of information retrieval, and in fact coined the term "information retrieval" at a chemistry conference.)
I found one and got my cheque on page Arabic one in one of his books. Paragraph one. First sentence. The very first word.
Picasso used to dine and dash as it were by drawing a doodle on the back of his check when the bill came due, and often enough the owner would choose to frame the check instead of cash it.
For a long time most of the cost of writing checks to Knuth is the writing of the checks, not the cashing of them. He’s paying for X00 checks at a time and the energy to fill them out. And anyone who had gotten their first check from him would not cash it.
Though these days I can cash a check via a phone app and so I don’t need to forfeit the check to get the money.
Its incredible that both of these technologies is in active use at the same time.
I do wonder whether this is more likely to be a case where the journal actually changed its name over time (perhaps because the Institut itself did) and then made the older papers available under the new name - which would mean both references are ultimately correct.
I don't completely know what is going on here, but I guess it is something like this: the institute has since 1898 officially been called Institut français d'archéologie orientale, and its journal has always officially been called Bulletin de l'Institut français d'archéologie orientale. However, historically, people would sometimes add du Caire (in Cairo) to the institute's name (to specify its location) – this habit was supported by the history that, prior to 1898, the institute (or its predecessor) was called École française du Caire (French School of Cairo) – and then unofficially abbreviate it from Institut français d'archéologie orientale du Caire to Institut français d'archéologie du Caire or even Institut français du Caire. And since the journal is named for the institution, once people got in the habit of unofficially abbreviating the name of the institution, they applied the same habit to the journal.
So Bulletin de l’Institut français d’archéologie orientale has always been the official name of the journal, but Bulletin de l’Institut français du Caire is a historical unofficial alternative name.
As I said, I'm just speculating, I don't really know. But this seems more plausible to me than the journal or institute changing its name, because I can't find any evidence of any name change since 1898, which was long before the publication of the 1930 article.
However, that is yet to come to pass.
However, when we at Mark Williams Company writing the floating point routines for the C compiler (https://winworldpc.com/product/mark-williams-letsc/3x), we were heavily reading Chapter 4, particularly on long division.
Also did spend some time in the Random Number generation chapter.
An interesting difference between my 1965 version and the ones published today is that long multi-page foldout in Sorting and Searching showing tape-based merge sort is no longer part of the book.
Edited to add: Amazon often has fake or at least printed with lower quality/cost versions. It was hard to find a real copy of his concrete math book when I was searching. The 'i18n' edition is printed on the cheapest of cheap paper for that book. I suspect that is related. But the fan-out pages are in the sorting volume.
#2 Vol 3 is on the second edition, with the tape merge fanout between pages 338 and 339. So it would be a particular printing that didn't have it.
Even in the old paper days banks where a little hesitant to let me come into your bank with a check from you and cash it without me also being a customer of that same bank. In theory you could do it, but sometimes if you tried you got the runaround.