Houshalter

12 ms·

61.
by Houshalter 9y ago
Why should we expect there to be any mathematical foundation to this stuff? It's quite possible to imagine an alternate universe where GAs and neural nets don't work. Because they have different datasets that don't fit the st
62.
by Houshalter 9y ago
Evolvability can also evolve away too. For instance, a gene that decreases the mutation rate to 0. Most mutations are harmful, so any organism with the gene will be more likely to have successful children. And eventually the gene will becom
63.
by Houshalter 9y ago
This is common wisdom I think is false. You absolutely do get stuck in local optima frequently with reinforcement learning. OpenAI has a good example somewhere of a robot trying to put a peg through a hole. Trained with regular gradient des
64.
by Houshalter 9y ago
SGD is embarrassingly parallel as well. You can train a net on several different examples simultaneously and combine the gradients or learned weights. The reason it's not done so much is because the bandwidth of moving huge numbers of
65.
by Houshalter 9y ago
Youtube has several levels of monetization. Some videos can still run ads, but only with advertisers that opt into it. They are really low paying ads and the creators make a fraction of the regular rate on it. Also what in particular is wro
66.
by Houshalter 9y ago
They did that months ago. The whole 'adpocalypse'. You can not get ad money on anything even remotely controversial on youtube. Recently they've taken down tons of gun channels among other things.
67.
by Houshalter 9y ago
No model is perfectly correct. That's the problem. Any model that approaches perfection would be unimaginably complicated.
68.
by Houshalter 9y ago
Aren't the inflation adjustments entirely based on the cost of common goods like flour?
69.
by Houshalter 9y ago
Here's a story of a guy who used the same method to come up with the optimal diet for himself: https://www.gwern.net/docs/statistics/decision/1990-dantzig.... The first try he ended up with a diet that r
70.
by Houshalter 9y ago
I just love the armchair experts who come out of the woodwork whenever this topic comes up. I'm sure you can accurately psychoanalyze the hidden motives. Of a subject you haven't even read the wikipedia page on.
71.
by Houshalter 9y ago
This are quite unspecific criticisms. IQ is very well researched and widely supported, as are twin studies.
72.
by Houshalter 9y ago
That's interesting. I don't think half siblings should be considered a representative random sample though. It may simply be that babies adopted away were more likely to be from low IQ fathers. The ncbi link seems to support my hy
73.
by Houshalter 9y ago
Science may be blocked in your country.
74.
by Houshalter 9y ago
Well adoption studies show that environment doesn't matter much to IQ. Kids raised in the poorest families didn't have higher IQs than kids raised in the richest families ( http://sci-hub.la/10.1007/s10519-007-
75.
by Houshalter 9y ago
On reddit you can not downvote comments from the user page. The downvote button is there but it doesn't do anything. People who know this can get around it possibly, but it's stopped a lot of that.
76.
by Houshalter 9y ago
I don't see the difference. The process of designing a thing is the same whether it's an algorithm or a pile of gears. Sure real world inventions can involve new scientific discoveries. But most don't, and scientific discover
77.
by Houshalter 9y ago
C# was made by microsoft for windows specifically. The majority of programming languages were not. When I started programming I got very frustrated getting stuff to work right. I remember it used to be near impossible to get pip on windows
78.
by Houshalter 9y ago
I think it's just a matter that most people use Windows, and programming software is optimized for linux. I once wrote up all the steps required to get youtube-dl working on windows. https://news.ycombinator.com/item?id
79.
by Houshalter 9y ago
Bandwidth and storage have been slowing down and stagnating for some time now. My internet isn't much faster than 10 years ago. The difference between video codecs makes a huge difference in the quality of video I get to see.
80.
by Houshalter 9y ago
>Make it clear that math, in any form, is not patentable. Ok but CRISPR isn't math. And the distinction between designing an algorithm and designing a machine is kind of arbitrary. If you are going to get rid of one you should get r
81.
by Houshalter 9y ago
HNs algorithm could be improved if new storied were briefly shown on the front page so a few people could see them and have a chance of voting. Like how new comments start a the top of the thread and fall if they don't get votes.
82.
by Houshalter 9y ago
Google did do this. It took ungodly amounts of computing power and only did slightly better than random search. They didn't even compare to old fashioned hill climbing or Bayesian optimization.
83.
by Houshalter 9y ago
Youtube's algorithm really favors long videos now which hurt the market for short comedy skit stuff. And their ad rate is terrible. The onion's channel makes like $150 a video. Which is nothing for professionally made content. Mos
84.
by Houshalter 9y ago
These NNs have fewer neurons than an insect brain. They have to use so little computing power they can be provided for free to everyone in the world. Of course they have limitations. But there's been exponential progress in the last fe
85.
by Houshalter 9y ago
The average person never gets the opportunity to be an early investor in startups. You could buy 10 bitcoins when they were $1. You couldn't buy Google stock until after they were successful.
86.
by Houshalter 9y ago
>Hormones, reflexes, instincts, empathy, etc. Again, human inference and learning are (as of yet) insurmountably more complex than statistical regression. Complexity doesn't mean better. Humans are absolutely garbage at statistical
87.
by Houshalter 9y ago
>When you make judgements about individuals based on their race No one is doing that. The algorithm does not take race into account. It's just an assertion that it uses other factors as a proxy for race. I've not seen this prov
88.
by Houshalter 9y ago
This is a different problem than what people are complaining about. In fact it may be the exact opposite problem in some ways. The problem you state is that, if you base an outcome on some measure, people will manipulate that measure. E.g.
89.
by Houshalter 9y ago
>Also from the abstract: "In some cases, black defendants are substantially more likely than white defendants to be incorrectly classified as high risk." But that's not really true: https://www.chrisstucchio.com
90.
by Houshalter 9y ago
>Policing in America is racist. Citation needed. Cops visit black neighborhoods more, sure. But they do that because they commit vastly more crime. Which is the cause and which the effect? It's also a lot more complicated than that.

More ›