hacker42

14 ms·

91.
by hacker42 10y ago
Check out these generated images: http://arxiv.org/pdf/1605.09304v1.pdf However, the technique does not seem to have a generative interpretation.
92.
by hacker42 10y ago
That depends on the nature of the data, I think. If the data has a lot of sequential, sparse, hierarchical statistical dependencies (like source code, text or data streams), they might be better modeled by an LSTM. If you have high-dimensio
93.
The Toxins That Threaten Our Brains(theatlantic.com)
4 points by hacker42 10y ago|1 comments
94.
The cognitive science of trolling(quora.com)
1 points by hacker42 10y ago|0 comments
95.
by hacker42 10y ago
It leaves the impression of a quick and weakly informed decision (especially so in the context of executing people). I personally would never respond to an important inquiry on mobile because the relatively limited screen and input interfac
96.
by hacker42 10y ago
On a closely related note, there is a collaboration of DeepMind and the FHI going on and they are about to present a paper on safely interruptible agents at UIA 2016: http://lesswrong.com/r/discussion/lw/noj&#
97.
by hacker42 10y ago
As far as I know, Gabor filters show up in all kinds of multi-layered models, even in unsupervised ones. Gabor filters are quite canonical.
98.
by hacker42 10y ago
One point that I've not seen mentioned yet, is that the neural revolution somewhat aggravates economic inequality. What recent progress basically has shown is that deep learning works better with more layers and more resources. Geoffre
99.
by hacker42 10y ago
I am not seeing the chicken and egg problem. Isn't it always the case that when we consider the optimality of something it depends on some definitions?
100.
by hacker42 10y ago
I have a pretty old machine without memory upgrade options. Of course, it would not be an issue on a machine with 16 GB. I've been following the bugs with tab-related memory and perf issues for quite a while, but maybe you are right th
101.
by hacker42 10y ago
This is a related bug report: https://bugzilla.mozilla.org/show_bug.cgi?id=1125557
102.
by hacker42 10y ago
I am still patiently waiting until unloaded tabs have a memory footprint of 0KB instead of 0.7MB.
103.
by hacker42 10y ago
I think none of the things I've mentioned rule dualism completely out, but they likely restrict the possibilities of where to find it.
104.
by hacker42 10y ago
Actually the neural net approaches are older than SIFT. Neural nets learn the distribution and even causal factors in the data. To me it seems that this distribution is often just too complex for it to be robustly captured by something that
105.
by hacker42 10y ago
Typos fixed. "All" was hyperbole of course, but I think it definitely does not look good for the majority of the work done on features. SIFT was recently outperformed PN-Net for example.
106.
by hacker42 10y ago
The answer to the question which technique will lead to AI depends on who you ask. Geoffrey Hinton and Jürgen Schmidhuber are convinced that neural nets trained with backprop will likely lead to AI with minimal additional fixed-function str
107.
by hacker42 10y ago
Tens of thousands of engineers (audio, vision, linguists etc.) spent millions of hours for billions of dollars in the past 30 years to invent algorithms that reliably tell us something about a bunch of data. For example, an corner feature a
108.
by hacker42 10y ago
Source?
109.
Proof of “√2 is irrational” using only logical formulas(us.metamath.org)
2 points by hacker42 10y ago|0 comments
110.
by hacker42 10y ago
Yes, this is roughly how it is believed to work. In fact, one can make out neurons that are reliably involved in representing concepts like location and orientation [1], places [2] and persons [3]. Usually, multiple neurons are involved in
111.
by hacker42 10y ago
Do you really expect anyone to sensibly respond to your passive-aggressive tone?
112.
by hacker42 10y ago
I've forgotten where, but I've recently read a blog post/opinion that we might be encountering the dawn of an applied mathematics winter. Instead of meticulously crafting algorithms for particular problems, we just apply stoc
113.
Deep Learning Without Poor Local Minima(arxiv.org)
123 points by hacker42 10y ago|32 comments
114.
To Help Kids Thrive, Coach Their Parents(nytimes.com)
2 points by hacker42 10y ago|0 comments
115.
by hacker42 10y ago
Statistics is about inferring probabilities from data such that we can make predictions (where data are discernible differences of some quantities). Inference means finding out what the world is about using some sort of representation (a mo
116.
by hacker42 10y ago
There is a Spacemacs layer with nearly the same functionality called evil-snipe which binds s and S.
117.
by hacker42 10y ago
That won't happen because the industry controls our governments too much. There is a lot of value to be produced before the algorithms become really dangerous. Moreover, controlling AI research is ever harder than nuclear research. Cre
118.
by hacker42 11y ago
Why does he handle the bare wires of the test lamp near the mains? Is there no risk the way he hooks it up?
119.
by hacker42 11y ago
I think the only way out is political top-down decisions rather than mental shifts. Habits are just too strong and the competitive disadvantage of frugality just too inconvenient for most of our primate brains. Environmentalism and passive-
120.
by hacker42 11y ago
I'm also getting the feeling that they are being dishonest about the techniques they've used.

More ›