Forgot your password?

Comment: Re:Wish I could say I was surprised (Score 1) 107

by TheRaven64 (#47430255) Attached to: Peer Review Ring Broken - 60 Articles Retracted
In the UK, university research departments are assessed base on the Research Excellence Framework (REF, formerly the Research Assessment Exercise [RAE]). Each faculty member is required to submit 4 things demonstrating impact. These are typically top-tier conference or journal papers, but can also be artefacts or examples of successful technology transfer. The exercise happens every four years, so to get the top ranking you need to write one good paper a year. The only incentive for publishing in second-tier venues is meeting other people who might lead to interesting collaborations.

Comment: Re:Wish I could say I was surprised (Score 1) 107

by TheRaven64 (#47430227) Attached to: Peer Review Ring Broken - 60 Articles Retracted
Reproducing work is often a good thing to set for first-year PhD students to do. If they reproduce something successfully, then they've learned about the state of the art and are in a good position to start original research. If they can't reproduce it, then they've got a paper for one of the debunking workshops that are increasingly attached to major conferences and that's their first publication done...

Comment: Re:Tannenbaum's predictions... (Score 1) 124

by TheRaven64 (#47425531) Attached to: Prof. Andy Tanenbaum Retires From Vrije University
Predicting that x86 would go away was more wishful thinking than anything else. At the time, Intel had just switched from pushing the i960 to pushing the i860 and would later push Itanium as x86 replacements (their first attempt at producing a CPU that it was impossible to efficiently compile code for, the iAPX432, had already died). Given that Intel was on its second attempt to kill x86 (the 432 largely predated anyone caring seriously about x86), it wasn't hard to imagine that it would go away soon...

Comment: Re:A great writer (Score 2) 124

by TheRaven64 (#47425431) Attached to: Prof. Andy Tanenbaum Retires From Vrije University
I found Modern Operating Systems better than the Minix book. The Minix book tells you exactly how a toy OS works in detail. Kirk McKusick's Design and Implementation of the FreeBSD OS (new version due out in a month or two) tells you how a real modern OS works in detail. Modern Operating Systems gives you a high-level overview of how modern operating systems work and how they should work. If you want to learn about operating systems, I'd recommend reading the FreeBSD D&I book and Tanenbaum's Modern Operating Systems and skipping the Minix book (which was also a bit too heavy on code listings for my tastes).

Comment: Re:Does this mean the death of Minix3? (Score 1) 124

by TheRaven64 (#47425395) Attached to: Prof. Andy Tanenbaum Retires From Vrije University

I feel it necessary to point out, though, that OS X is not a microkernel system comparable to Minix

While this is true, it's worth noting that a lot of the compartmentalisation and sandboxing ideas that most of the userland programs on OS X employ (either directly or via standard APIs) have roots in microkernel research. OS X is in the somewhat odd situation of having userspace processes that are a lot more like multiserver microkernels than its kernel...

Comment: Re:Charge what it costs to certify (Score 1) 123

by TheRaven64 (#47398031) Attached to: FDA: We Can't Scale To Regulate Mobile Health Apps
This is exactly how they do it now for software in medical devices. Except that that 'private firms' are the manufacturers. It's ludicrous for them to claim that a certification process that basically amounts to asking the manufacturers to promise that they've tested it really well 'won't scale'.

Comment: Re: Marty! (Score 2) 551

Apparently the early script drafts had a more plausible explanation: that the spare brain capacity of humans in a dream-like state was used as processing power to run the AIs. One of the editors thought this was too complicated for a movie-going audience to understand and so replaced it with a magic perpetual motion machine.

Comment: Re:AI is always "right around the corner". (Score 3, Interesting) 551

Translation is like predicting the weather. If you want to do an okay job of predicting the weather, predict either the same as this day last year or the same as yesterday. That will get you something like 60-70% success. Modelling local pressure systems will get you another 5-10% fairly easily. Getting from 80% correct to 90% is insanely hard.

For machine translation, building a database of 3-grams or 4-grams and just doing simple pattern matching (which is what Google Translate does) gets you 70% accuracy quite easily (between romance languages, anyway. It really sucks for Japanese or Russian, for example). Extending the n-gram size; however, quickly hits diminishing returns. Your increases in accuracy depend on a corpus and when you get to the size of n-gram where you're really accurate, you're effectively needing a human to have already translated each sentence.

Machine-aided translation can give huge increases in productivity. Completely computerised translation has already got most of the low-hanging fruit and will have a very difficult job of getting to the level of a moderately competent bilingual human.

Chemist who falls in acid is absorbed in work.