AWS Official Blog

Amazon DevCon – George Dyson

by Jeff Barr | on |

Introduced by Larry Tesler.

Interested in beginnings of things, computing has a very clear beginning.

Institute for Advanced Study in Princeton, John von Neumann. Deserves a lot, but not all, of the credit. Picture, canisters, memory tubes. 32×32 matrix, 1024 bits, actually pixels. Dots on a CRT screen. von Neumann, theory of self-reproducing Automata.

Thomas Hobbes (1600’s), patron saint of all coders. Leibnitz, explicit algorithms, binary arithmetic, universal language. Talked about marbles, tracks, wheels. Inventing the shift register, what is now done by gates. Could have built this machine, as noted by Neil Stephenson. Reinvented by von Neumann’s gang. Built by people. Alan Turing’s first paper, “On Computable Numbers…” (1936).

Ferranti machine, with random number generator, allow machines to guess, see if it was closer to or further from objective.

Louis Richardson, during WWI looked at weather prediction, dividing Earth in to cellular grid, passing numbers around. If you had 64,000 mathematicians it could be done in real time.

Norbert Weiner, Ballistics Proving Ground (Aberdeen).

von Neumann invited to Princeton, proposed creation of a Mathematical Institute. Abe Flexner founded the institute. Wrote “The Usefuleness of Useless Knowledge.” Bambergers, philanthropists, put their money in, bulding Fuld Hall at Princeton.

von Neumann and others came out of Budapest, would be denied a visa if he applied today.

Game theory and economic behavior; coalitions are key, and this worked for Amazon. Coalition between people who wanted to buy books and people selling books.

Kurt Gdel, not given enough credit for digital revolution. He was there, next to von Neumann. Key concept, using addresses, that can be processed using arithmetic. Caught Turing’s attention. Lots of cross-fertilization among these logicians.

Gdel, was afraid and paranoid, and got himself stuck in a real-world logical paradox due to visa issues. Was essentially stuck in his room at the Institute. Became US citizen, almost drafted, clearly not a good candidate for the military. Institute basically a good home for all of these crazy people (almost like Amazon, he jokes), a home for life.

George has some great pictures, maps, letters, flowcharts, schematics, and even a room layout for Fuld Hall. Amazing letter from von Neumann and Veblen (?) expressing concern that the Germans could build an atomic bomb but the US couldn’t.

von Neumann decided computers would be more important than bombs, but used computers to make bombs. Neutron diffusion simulation, “6 months” until they would have a computer. Had access to the  ENIAC, ported the code from the unbuilt Princeton machine to ENIAC, some kind of 2-level architecture which George says is the orgin of software. Does he mean microcode and the instruction set?

IBM hired von Neumann as consultant, solving differential equations, key “von Neumann” CPU, memory, I/O architecture conceived, but von Neumann gets too much credit. Willis Ware is apparently an unsung hero.

Rockefeller foundation almost invested, backed off due to potential patent issues. First meetings held at RCA, Vladimir Zworykin. They decided there was no market, pulled out.

First meeting, set up basics. Number is an order (instruction), memory, etc.

RCA was to build a Selectron, a digital vacuum tube. 4000 bits of storage planned. Only got 256 bits working, 10K hour MTBF. RCA kept pulling resources and inventor slowly went mad. Big fuss over patents (what else is new?). All engineers gave up patent rights, each ended up with $1 for all of these patentable inventions.

Things happened quickly because everything was published ASAP. We are at 1946 right now.

Early notation, bd is binary digit. First list of order codes (instructions). Growing budget. Memos for where to put new employees, put them in room next to men’s room in basement. Almost rented rooms at a chicken farm.

Oppenheimer, Goldstine, von Neumann, Bigelow. Bigelow moved them from theory to practice, building the actual machine.

Early developers were woman, paid $275/month.

Big, funny issue, sugar was rationed and these folks used several times as much sugar as everyone else. Someone wrote a memo about this! Ironically, they now do this in [Charles] Simonyi Hall.

3400 vacuum tubes in the machine, ENIAC had over 18,000. 24 microsecond cycle time, total memory 5000 bytes, about 1 Windows icon.I/O was Teletype tape, then punch cards. CRT tube memory, Williams and Kilburn.

Kept lots of documents due to fear of patent suits. George found them in a vault, 50 years later. They are now in the rare book library, often checked out by IBM, RCA, and so forth.

Hard to get good CRT tubes, one piece of dust would ruin it. Hit with hammer to clean / fix, which often actually worked.  Bits were actually visible. Drum memory, tracked, like a hard disk. First one still exists, somewhere in Israel.

1948, first bitmap display!

“Reliable organizations of Unreliable Elements”, fundamental von Neumann paper.

Early logbook, 90 minutes to enter code, 2 minutes to run. “Code error, machine guilty.” Roof melting, tar from roof finding its way on to punch cards.  Run everything twice to confirm results.

Meteorology was a cover story for the work on bombs, many similarities. CNN weather forecasts today really do the same thing, just a lot faster. Stellar evolution, interior of a star. Ran this in the background, lowest priority. Ran planetary cycle backward to 600 B.C.  Predicted freeway traffic flow.

1952, Ivy Mike, first thermonuclear explosion.

Experiments in bionumeric evolution, Nils Barricelli, mathematical theory of evolution. Had machine from midnight until dawn, artifically created universe. Viral genetics. This was 1953, before DNA concept elucidated. Self-reproducing virus; visualizations made by punching data onto cards and taking pictures. Saw analogies to RNA, encoding and transmitting information.

Last von Neumann book, The Computer and the Brain.

By 1958, still no customers, shutting down. All involved knew that it would change the world. Many participants, but von Neumann had the connections to bring it all together. The realization of Leibniz’s dream, everything coming from 1’s and 0’s.  iPod is just a sexy way of packaging it all up.

(applause)

Q+A:

Q: Who will be the von Neumann 50 years from now?

A: Young people, perhaps biology or conjunction of several things. Moving genome into machine, that’s today’s revolution.

Q: Quantum computing?

A: Could be one of the revolutions, maybe von Neumann is just a larval stage. In 10-20 years we will know.

Q, from Jeff Bezos: What is in the box that von Neumann gave his daughter, and asked not to open for 50 years? (this is what Jeff was talking about).

A: That’s Marina von Neumann,we’ll know soon. Perhaps something from the Oppenheimer situation, von Neumann stayed quiet. Very exciting, he wouldn’t lock something up unless it was really important.

Q: Politics?

A: Meetings going to Europe, too hard to get visas to come to the US, vacant apartments at IAS for people who couldn’t get visas for self or spouse.

Q: Working in archives?

A: Had an entire year and full access to material, luxury of time, learn through recognizing handwriting.

Q: Lack of archiving of the Internet, how will this impact historians trying to cover this time period?

A: Brewster Kahle, our saint. Easier to buy a new hard drive than to delete stuff, search technology getting better. Lots of digital garbage, but good tools to sift through it. People do still have habit of printing stuff out. Stuff that should have been thrown out (but wasn’t) is the important stuff. Buy old drives, old stuff, someone will be interested!