Houshalter

8 ms·

31.
by Houshalter 8y ago
Very true but I don't know if people can use this reasoning rationally. I know a well known news site that gets posted here often. That ran a big piece on how rare terrorism is and how you shouldn't worry about it at all. But then
32.
by Houshalter 8y ago
What's special about 2018 computing power? Neural networks have been declared dead several times in the past. Because they couldn't do anything interesting with the measly computing power available at the time. Currently the bigge
33.
by Houshalter 8y ago
It's worse than just 5% because p hacking. Every analysis has many parameters, and you can choose the most significant out of thousands of possibilities. On top of this, the more interesting results are the most surprising. And therefo
34.
by Houshalter 8y ago
Those are just bad ad hoc consequentialist justifications for incarceration. If you really believe that, you can justify life sentences for every minor crime. Since it keeps criminals off the streets and increases utility. Or you can justif
35.
by Houshalter 9y ago
AI kind of sucks now though. As it gets better and more people use it, it will be more of a selling feature. That's what they are betting on anyway.
36.
by Houshalter 9y ago
As far as I know caffeine isn't associated with heart disease. AT least not in the brief googling I did.
37.
by Houshalter 9y ago
Yes, but what is the evidence it's the nicotine that's damaging the heart? The heart is part of your respiratory system. Inhaling regular wood smoke damages your heart: https://samharris.org/the-fireplace-delusion&
38.
by Houshalter 9y ago
Why is this even controversial? Vaping removes almost all of the known dangers from smoking. We know smoke inhalation itself is incredibly bad for health. It was never about the nicotine. Wood smoke and general pollution is highly associate
39.
by Houshalter 9y ago
It's also just a cherry picked example too. Surveys of experts show there isn't really a consensus, but they generally aren't that optimistic. https://nickbostrom.com/papers/survey.pdf >We thus design
40.
by Houshalter 9y ago
Note the website he founded and posted this on kind of died itself.
41.
by Houshalter 9y ago
And my comment has been flagged too. The political censorship on HN is ridiculous.
42.
by Houshalter 9y ago
Because they commit more and worse crimes. They are not more likely to be arrested and convicted than a white person. And we know this because of surveys of victims and crime reports that record the race. Like that chart I posted. The only
43.
by Houshalter 9y ago
If you tell people constantly that they are being oppressed, they will start seeing it everywhere. When a white person is treated badly by cops, which happens all the time, they will just assume the cops were general assholes. When a minori
44.
by Houshalter 9y ago
It's platform independent. Most deep learning libraries have absurd dependencies. They are difficult to install and only support specific OSes and GPUs. Javascript will run literally anywhere.
45.
by Houshalter 9y ago
It seems to have been updated quite a bit. With support for more modern NN architectures. But I don't know how fast it is relative to the others.
46.
by Houshalter 9y ago
recurrentjs https://github.com/karpathy/recurrentjs convnetjs https://cs.stanford.edu/people/karpathy/convnetjs/ I think the author of these also had one for reinforcement learning.
47.
by Houshalter 9y ago
You can copy the link to your favorites playlist and give it to youtube-dl. And it will automatically download all of them (or at least the last dozen.) I find a decent percent of my favorited videos get removed with time, so this is worth
48.
by Houshalter 9y ago
At least you are getting content from other channels. It seems like all my recommendations are to videos from the last few channels I have watched. Never anything I haven't heard of before. Even most of the channels I subscribe to neve
49.
by Houshalter 9y ago
Is advertising really necessary? Most youtubers are making more money from donations than they are from ad revenue. Many of them also do paid promotions that pay better than youtube ads. The ad rate is really bad. And you can get demonetize
50.
by Houshalter 9y ago
That's not "rule by the rich". The middle class and even much of the poor still voted. You just had to own land, which isn't exactly a high barrier. The point was voting should be one vote per household, by the head of t
51.
by Houshalter 9y ago
The first cars killed way more people than horses and they didn't get banned.
52.
by Houshalter 9y ago
The same way any other website deals with bot abuse. You give extra scrutiny to new accounts, require captchas, flag suspicious email providers and bank accounts, and block IP ranges that correlate highly with abuse. I don't think you
53.
by Houshalter 9y ago
My impression was most people found him much more charismatic than Clinton. Charisma plays a huge role in elections: From Influence: Science and Practice >a study of the 1974 Canadian federal elections found that attractive candidates
54.
by Houshalter 9y ago
Yeah what he did was definitely cool. But if you just want to test a hypothesis, it would be much easier to send a camera and not a whole person. Also a balloon could obtain a higher altitude than any amateur rocket. And people have filmed
55.
by Houshalter 9y ago
No, there is no mathematical foundation for most of that stuff. No mathematician can prove that gravity exists. You can describe the laws of physics with mathematical expressions. Just as you can describe machine learning algorithms with ma
56.
by Houshalter 9y ago
You don't verify every task. You verify some percent of them at random. And blacklist the people who cheat the system (Maybe with some leeway because random memory errors are a thing with consumer GPUs.)
57.
by Houshalter 9y ago
There are many many causes of cancer. Even if cellphones are a significant source of cancer, any effect would be almost entirely lost in the noise. Demographic change and frequency of smoking alone explain much of the variation in cancer. T
58.
by Houshalter 9y ago
Well that would work for simple hillclimbing. In a full genetic algorithm, every member of the population is the product of thousands of previous individuals.
59.
by Houshalter 9y ago
Maybe there's a reason. But I don't think you will ever be able to "prove" it. In any kind of formal way that would be satisfying to mathematicians. Solomonoff induction is the best attempt to try to formalize machine le
60.
by Houshalter 9y ago
It's not hypothetical. We haven't observed it in nature because any organism that evolved this would simply go extinct. But similar "evolving to extinction" phenomenons have been observed. Such as genes in mice that make

More ›