sushirain

8 ms·

31.
by sushirain 10y ago
If we trained a net on many classes except for cats, then it would require a few examples of cat to recognize cats.
32.
by sushirain 10y ago
The difference from the 60's mis-predictions is that today we already have speech and visual object recognition systems working with 95% accuracy and we only need to get them to 99%. This gap will be closed by data collection and faste
33.
by sushirain 10y ago
Local minimum? In the next 10 years many deep learning applications will materialize: speech recognition will reach human level in production, autonomous trucks on highways, autonomous cars for consumers, AR, personal robot cleaners. The f
34.
Chat Smarter with Allo(googleresearch.blogspot.com)
4 points by sushirain 10y ago|0 comments
35.
by sushirain 10y ago
You do need deep-learning to get state-of-the-art OCR when the image is not scanned from a book, but photographed text (or text-in-the-wild) in unconstrained conditions. It is a good example of the superiority of deep learning. One recent e
36.
Facebook Users Are Sharing Fewer Personal Updates and It's a Big Problem(fortune.com)
2 points by sushirain 11y ago|1 comments
37.
Facebook Wants You to Post More About Yourself(bloomberg.com)
23 points by sushirain 11y ago|23 comments
38.
by sushirain 11y ago
I don't think that AlphaGo was trained more on Lee Sedol's games, than on others' games. The team said that they can't find computer weaknesses until AlphaGo plays against top caliber.
39.
Ask HN: Android browser Ad blocker?
9 points by sushirain 11y ago|9 comments
40.
Robert Scoble: Augmented Reality Decompression [video](facebook.com)
2 points by sushirain 11y ago|0 comments
41.
by sushirain 11y ago
Deep learning experts understand how neural nets work. But: 1. They don't usually try to explain the specific activations in the hidden layers of the network. This is hard and depends on the specific trained net. 2. They can't gua
42.
by sushirain 11y ago
I think he meant recurrence when he wrote recursion, because recurrence is popular in deep learning (RNN) and recursion is not.
43.
by sushirain 11y ago
Perhaps landing into a tank of non-corrosive liquid would prevent such collapses? Or on some kind of "glue"/alloy that quickly freezes. You can also have spearguns firing cords from the top of the rocket to the ground after l
44.
by sushirain 11y ago
Larry and Sergey started Google because they wanted to be rich? Even if they had known for sure that it wasn't the way to the 1%, but say to the 10%, would they not have started Google?
45.
Why video games are essential for inventing artificial intelligence(togelius.blogspot.com)
1 points by sushirain 11y ago|0 comments
46.
by sushirain 11y ago
The article mentioned brain drain, but more specifically, Canada let Hinton, Lecun, and Sutskever go... :)
47.
AMA: the OpenAI research team(reddit.com)
130 points by sushirain 11y ago|41 comments
48.
by sushirain 11y ago
These are very interesting points: "the average elapsed time between key algorithm proposals and corresponding advances was about eighteen years, whereas the average elapsed time between key dataset availabilities and corresponding adv
49.
Datasets Over Algorithms(edge.org)
4 points by sushirain 11y ago|1 comments
50.
by sushirain 11y ago
What do you mean by hand engineered? Is a part-of-speech tagger trained on a manually annotated corpus hand engineered? If not, then which "stacks" are hand-engineered and still popular today?
51.
by sushirain 11y ago
You are right. If deep learning takes more and more of the processing, we will need GPUs or more CPU cores, not necessarily clock speed.
52.
Learning to Reason: PhD Thesis Proposal, Wojciec Zaremba [pdf][2014](github.com)
4 points by sushirain 11y ago|0 comments
53.
by sushirain 11y ago
Some of my least favorite quotes: "A single neural network is never going to be able to write news articles with actual sense or meaning." - why not? "they're never going to be providing you with an honest opinion of a p
54.
I taught a computer to write like Engadget(engadget.com)
1 points by sushirain 11y ago|1 comments
55.
by sushirain 11y ago
Highlights •The Blue Brain Project digitally reconstructs and simulates a part of neocortex •Interdependencies allow dense in silico reconstruction from sparse experimental data •Simulations reproduce in vitro and in vivo experiments withou
56.
Reconstruction and Simulation of Neocortical Microcircuitry(cell.com)
1 points by sushirain 11y ago|1 comments
57.
Neural technologies could become the next battleground(foreignpolicy.com)
30 points by sushirain 11y ago|4 comments
58.
by sushirain 11y ago
From the article: "The list of dimensionless physical constants increases in length whenever experiments measure new relationships between physical phenomena. The list of fundamental dimensionless constants, however, decreases when adv
59.
Dimensionless physical constant(en.wikipedia.org)
1 points by sushirain 11y ago|1 comments
60.
by sushirain 11y ago
They are effective because: - They use more parameters (and fewer computations per parameter.) - They are hierarchical (convolutions are apparently useful at different levels of abstraction of data). - They are distributed (word2vec, though

More ›