jmmcd

26 ms·

271.
by jmmcd 8y ago
> what would you do with that mapping Well, I could write another tool to analyse it, eg extract some statistics; I could write another tool to generate it, eg based on data.
272.
by jmmcd 8y ago
I really liked this article. I feel about the same about this as when I first understood state machines. Sure, we could write a loop with an if-else-if-else block inside to implement a state machine, but once we understand the formalism pro
273.
by jmmcd 8y ago
Yes, but it would be nicer if the table was stored as a table -- as data -- not as code.
274.
by jmmcd 8y ago
MIT has absolutely no shortage of GOFAI-ers.
275.
by jmmcd 8y ago
I think the NYT headline is at fault here. College of Computing makes more sense than of AI. You can't put CSAIL inside a College of AI, for a start.
276.
by jmmcd 8y ago
> find another way to measure developer productivity And solve the halting problem while you're at it
277.
by jmmcd 8y ago
> in the end you have to start developing that pattern against pattern type of thinking in favor of composing note against note & note after note. I think there is a typo, or this is ambiguous. Are you saying you should use "p
278.
by jmmcd 8y ago
Ok but the present article is by Allen Downey, a well-known figure here and an academic. I don't think we could suggest he is part of astro-turfing.
279.
by jmmcd 8y ago
Newton wasn't so humble! He said that, but only as part of a deliberate insult to one of his rivals, who was short.
280.
by jmmcd 8y ago
Not really. There are other journals that I do read.
281.
by jmmcd 8y ago
I agree. In my field there are entire journals which I regard as rubbish and which are 95%+ Chinese/East Asian. Part of the problem is integrity, but a larger portion of the bad papers are just meaningless cargo-cult science.
282.
by jmmcd 8y ago
My main problem with the paper is that it's about groups, not individuals. Groyp selection has been debunked. There are better ways to think about the wider variance that don't involve groups. Think instead about each individual &
283.
by jmmcd 8y ago
The ground ends up purified (somewhat) either way, I guess. If we want to use the nickel, will it make a big difference?
284.
by jmmcd 8y ago
It's not good to reveal your scheme, or for your scheme to be so simple. If a HN database ends up in the open, your accounts on other sites become vulnerable.
285.
by jmmcd 8y ago
If I understand right, that figure is measuring the accuracy of a prediction of electricity production, not the constancy of the production. We can't smelt aluminium with no power even if we have predicted that there'll be no powe
286.
by jmmcd 8y ago
It's a ceremony and party that is paid for. If you see it as a process of transferring money you are missing something quite important.
287.
by jmmcd 8y ago
Ireland here. Never heard of a fake cake. Had an amazing Bailey's and chocolate biscuit cake for my own wedding.
288.
by jmmcd 8y ago
Bostrom's orthogonality thesis!
289.
by jmmcd 8y ago
I wrote to Ken Thompson when I was an undergrad doing a history of computing essay. And he wrote back with a potted autobiography! It was probably a bit before email became the scourge that it is now. The sad part is: I lost the email in a
290.
by jmmcd 8y ago
The same people who like this kind of thing usually also like portmanteaux (the lowest form of wordplay) and then offer it as evidence that they love language.
291.
by jmmcd 8y ago
I agree -- though I'm not sure how to make the right name for this technique. Maybe "relative-probability GAN" is the best. I also wondered before clicking what a relativistic GAN would be: maybe as the activity of a neuron b
292.
by jmmcd 8y ago
"Possible" v "Permissible" is not the problem. What seems to be needed (based on reading the post but no other knowledge) is the phrase "including but not limited to".
293.
by jmmcd 8y ago
You understand the timeline, but the virtue-signalling mob on Twitter probably doesn't.
294.
by jmmcd 8y ago
How about "obsolete"? "Deprecated"? "No longer supported on this hardware"?
295.
by jmmcd 8y ago
Bad title (in the original) because the examples in the paper are mostly drawn evolutionary computation and artificial life, with only a few relating to neural networks being used in those fields.
296.
by jmmcd 8y ago
Some readers might enjoy looking at code -- I came across this repo: https://github.com/akelleh/causality
297.
by jmmcd 9y ago
Well it's in the article.
298.
by jmmcd 9y ago
> Sometimes conferences pay a publishing company to do the review process for submissions, by the way. I'm not sure that your submissions are actually getting processed for free by conference volunteers. Well, I'm sure, because
299.
by jmmcd 9y ago
EAs optimise black-box objective functions. NFL prevents them from doing well on all possible objective functions. But rather weak assumptions are enough to evade NFL, for example the assumption that maximal steepness in the objective space
300.
by jmmcd 9y ago
I agree, but I intended that point as part of my final point -- an EA can draw a solution from any search space, including a space of trees, for example, as in GP.

More ›