Hacker News new | past | comments | ask | show | jobs | submit login
Claude Shannon Turns 1100100 (newyorker.com)
141 points by anthotny on Apr 30, 2016 | hide | past | web | favorite | 44 comments

Try to imagine where we would be today if Bell Labs and/or MIT had fired him for his eccentricities, "trivial" distractions and prolonged absences. Imagine if he hadn't been present when Bardeen & Brattain(and subsequently Shockley) produced the modern transistor and took off on one of his tangential side projects. Just "wow".

edit: giving Bardeen & Brattain their due.

When I was at MIT, we spent a History of CS class on him. The fact that you can calculate te theoretical limit on info transmitted given a channel's parameters is kinda mind-blowing. Like Einstein figuring our theoretical limit on velocity.

He also wrote one of the most important master theses ever - from his NYT obit - "In what has been described as one of the most important master's theses ever written, he showed how Boolean logic, in which problems can be solved by manipulating just two symbols, 1 and 0, could be carried out automatically with electrical switching circuits. The symbol 1 could be represented by a switch that was turned on; 0 would be a switch that was turned off."

The other important Master Thesis is probably Gödel's.

True. Wonder what other Master's theses that were underrated at publication but went on to be seminal....

I hate non-standard length binary numbers! Just write 01100100 for God's sake...

On topic, Shannon is my favorite scientist of the last century, after von Neumann of course. If his only contribution was connecting Boolean algebra to digital logic, he would have been a pioneer. But no, he goes ahead and defines information theory out of nowhere, even though the majority of applications that can utilize it were not yet practical! I mean, come on Claude!

Funny, I saw it and thought 'hey, seven bit ASCII---how appropriate'.

Do you also prefer "standard" lengths of decimal numbers, and if so, what are those?

On a related note, I once had the weirdest bug in a C program that I couldn't figure out. It turns out that I had stylized an array of integers with preceding zeros, like: {025, 057, 112} to make it pretty. Little did I know that those zeros made the numbers octal!

Yes, I do prefer standard lengths of decimal numbers. With decimal numbers, we break every group of 3 up with a comma. It's much easier to read 1,000,000 than 10000000, after all. If you disagree, I hope you noticed that I wrote two different numbers (1 and 10 million).

So I prefer if you break up binary numbers on either 4 or 8 bit boundaries with a space and leading zeroes, as is very common in hex editors. So I'd have written either 01100100 or 0110 0100, either of which would be easier to interpret.

Using an initial 0 to indicate octal is a terrible convention. I’m very glad that Python 3 makes 0123 into a syntax error, and uses the form 0o123 instead for octal literals, to match the style of 0x53 (also adds 0b1010011 for binary).

I wonder if there's a reason for keeping the leading 0 there past tradition/convention. 0x as a prefix makes sense to differentiate between 0x12 and 012, but if it's 0x 0b and 0o, imo it'd make more sense to drop it and just leave it as o12, b12, x12, etc.

o12, b12, and x12 are all valid variable names

Yeah... not sure how that slipped my mind, seems quite obvious now. Thanks

I prefer the Scheme way: #x53, #o123, #b1010011

It feels more natural to my taste. We already use #1, #2, ...

With binary especially I feel it's easier to read if the number is written out in nibbles. Besides, many of us were introduced to binary through ASCII, so the 8 bit format feels even more natural imo.

Oh damn, that must have been tough to find!

Ah, the oblivious tyranny of the victorious 8-bitters.

The days of the 6-bit register or the 18 or 36 bit words are gone, dust beneath the heels of the conquering hordes of the 8008 microprocessor heirs.

Yeah, especially since 01100100 can't be misinterpreted as -28

01100100 has been misinterpreted as 294976 though.

I suppose we can rely on context to choose the correct interpretation, as Shannon is unlikely to have been born 294,976 years ago or 28 years in the future.

your displeasure only illustrates the limits of your interest

there are a number of areas of inquiry where using 1100100 is more appropriate than 01100100

and considering here it is being used to represent an age, one would have to show a spec on ages represented in binary to determine what 'standard' is

At least we can assume that age is unsigned...

for now :p

on Shannon Laboratory (from https://around.com/where-are-they-now-bell-labs/):

[quote]AT&T was always oblivious to Shannon’s role as the father of modern communications, and (with its usual tin ear), couldn’t figure out how to benefit from its own history.

Back when Lucent split from AT&T and was granted most of Bell Labs (along with bragging rights for the invention of the transistor), I was busy trying to convince AT&T that a services company required an R&D capability. The original divestiture plan did not anticipate a basic research component—only a development lab. Long story, but eventually we managed to slide over around 10% of the scientists into the new AT&T Labs.

However, we couldn’t slide over its long history, mostly rooted in the physical sciences. We had the people, but lacked a well defined “brand.” By odd coincidence, I proposed if Bell Labs was granted computing in the divorce, we should own communications—and claim Shannon as our direct forebear.

AT&T HQ was ambivalent—they were unaware of Shannon’s seminal discoveries, and couldn’t imagine how claiming paternity of the communications revolution could benefit a long distance phone company. But we were given permission to name a lab building after Shannon. Fortunately, one of the lab’s directors (Ron Graham) was a mathematician and juggler and knew Shannon’s widow, and received permission.

But they still barely leverage this connection. Momentum is sometimes more powerful than entropy.[/quote]

The genius featured in almost all the CS textbooks who never won a Turing Award. I don't even know exactly where that ranks on my list of reasons for strongly disliking the ACM, but it's on there.

"Claude Shannon" is actually "Amazing Genius" encrypted with a one time pad.

"Claude Shannon" is also an anagram for "Chaos Laden Nun".

Unfortunately, also "Nachos Unladen"

The information regarding the unladen state of the nachos can be compressed into fewer bits by just saying "chips".

You'd probably understand if it was "chps" too.

Claude Shannon is my favorite scientist too. He made break throughs in the most literal sense. Everything was different after The Mathematical Theory of Communication.

I feel lucky and proud to have studied his work, through the lens of signals and systems, at his Alma Mater, University of Michigan.

IEEE also had a nice retrospective:

Claude Shannon: Tinkerer, Prankster, and Father of Information Theory


His insights are so fundamental and easy to state that it would have been nice for them to give an example: "Suppose you send ones and zeroes through a wire and noise causes 10% of the digits to randomly flip the wrong way. How much data, if any, can you send through the wire?"

We're only 28 years away from a nice, round number!

Many people seem to. I've actually been looking for negative feedback, since I thought some of the earlier drafts were better. I am just some guy with an opinion though. Overall it was a really difficult piece, given how abstract Shannon's work was -- what sort of visual metaphor do you use for the man whose work was to split the concept of "communication" from any concrete implementation? But, I am happy that people like his work, and I will certainly pass along your compliments.

You are looking for negative feedback for another Googler's work? Well that's not very googly.

I am not employed by Google. If I were, I would still have a right to disagree with the art direction. The execution was fine; I preferred other concepts which did more to display Shannon's work. Shannon's juggling was about as relevant as the flamethrowing trumpet he designed. Juggling was supposed to be more 'accessible'. Should doodles be representative, or accessible?

My bad. You are literally just a guy with opinions. I thought you were soliciting for feedback on a public forum for a coworker. I got confused.

Seeing a seven-bit value with three ones, my brain autocorrected 1100100 to 1101000 in order to turn this into a codeword in a perfect cyclic code...

I feel like the article ends very abruptly. It would have been nice with a few words on what he did in his late years and at least have his death mentioned towards the end, so the reader gets a feel of actually going through his life.

It seems odd to end at a conference he showed up to once.

As an aside, Dr. McEliece was one of the best teachers I had while at Caltech, with an amazing ability to present dry material an a way that was both informative and interesting.

i love a story where studying a flaw in a system creates an entirely new area of study

investigating line noise was also a part of mandelbrot's revelation on self similarity

his original 1937 thesis(o) is extremely approachable, as is the case with many papers where a new concept is developed therein

i recommend reading it

(o) http://dspace.mit.edu/bitstream/handle/1721.1/11173/34541425...

fun quote:

" This design requires that the primes less than 10,000 be known. If desired, the machine could be made to automatically connect in new counters as the primes were found, but there are many accurate tables of primes up to 10,000 so that this would not be necessary.

As to the practicability of such a device, it might be said tnat J.P. Kulik spent 20 years in constructing a table of primes up to 100,000,000 and when finished it was found to contain so many errors that it was not worth publishing. The macnine described here could probably be made to handle 5 numbers per second so that the table would require only about 2 months to construct. "

i am always amazed at the idea of tables of useful figures being a limited resource, think tycho brahe or log tables

Ironic to write in binary a number which is only significant in decimal.

"The thesis melded George Boole’s nineteenth-century Boolean algebra (based on the variables true and false, denoted by the binary one and zero)"

I just have to complain about this clause referring to the values true and false as "variables"

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact