The Great Debate @ARITH23: John Gustafson and William Kahan
dancefan1 HN recommenders · 2017–2023
4K views · talk · evergreen · Uploaded 2016-07-30
A 95-minute recorded debate at the ARITH23 computer arithmetic conference between John Gustafson, proponent of unums/posits, and William Kahan, architect of IEEE 754 floating point. They argue the merits of the new number formats versus conventional floats and interval arithmetic.
It is the go-to primary source whenever HN discusses posits versus IEEE floats, valued for Kahan's sharp critique despite the poor audio.
6:538-bit computer update
Ben Eater
1:21:22Light Years Ahead | The 1969 Apollo Guidance Computer
TNMoC
27:25“Hello, world” from scratch on a 6502 — Part 1
Ben Eater
47:52Finding the Missing Memristor - R. Stanley Williams
The Qualcomm Institute
1:01:07The Ultimate Game Boy Talk (33c3)
media.ccc.de
1:48:55The Thirty Million Line Problem
Molly Rocket
1:06:20USENIX ATC '21/OSDI '21 Joint Keynote Address-It's Time for Operating Systems to Rediscover Hardware
USENIX
1:34:44Jim Keller: Moore's Law, Microprocessors, and First Principles | Lex Fridman Podcast #70
Lex Fridman
4 comments from Hacker News
The Great Debate @ARITH23: John Gustafson and William Kahan [VIDEO] Audio quality is not great but there are links to slides from Gustafson and Kahan
I did some reading on unum a while back, inc. Kahan's criticisms. There's another side. There's a vid on youtube with Kahan and Gustafson talking over it. Kahan goes in like a wolf onto a lamb, much to gustafson's shock and hurt. Gustafson points out some of Kahan's criticisms are plain wrong, even citing the page in the book, then says something like "we should be working together on this, why aren't we working together?" Kahan doesn't respond. Kahan may be right or wrong but his attitude is weirdly hostile, and Gustafson is no n00b about floats. Edit: I think this is it (but it's been a while) [VIDEO]
Original posits are variable width, making them nearly useless for high performance parallel computations. Later versions don't add anything of use for low precision neural networks, and lack of hardware support anywhere make them too slow for anything other than toying around. See also [LINK] and [VIDEO]
A related discussion about number types: The Great Debate @ARITH23: John Gustafson and William Kahan [VIDEO] Kahan points out disadvantages of interval maths.