pilooch

At the moment it is much more interesting when applied to more narrowly targeted tasks. These can include video captioning but for clearly specified tasks. The (too quick of a) tentative jump from well automated targeted tasks to generic task automation is the culprit here.

In my few years of practice of deep neural networks in industry, I can say they are very nice to work with as they fit a large number of tasks and yield excellent results for many of them.

ML is the next commodity on the development stack. It is good to worry over the next few years, but after that, there should be a bunch of pretty solid tools out there for developers to work with. I am among the people that I believe are working on these tools.

You can use mairix as a search system outside of mutt. Results are a temporary mailbox.

OpenCL is slowly being supported by some of the main deep learning packages. The annoying part of Nvidia's GPUs are drivers and closed souce CUDA. Their CuDNN extension is a proprietary blob (with bugs) that boost up perfs but remains out of touch from scientists and developers. My hunch is that OpenCL will become the standard for deep learning, thus opening up to a larger set of hardware options.

No one I know of, from the last time I did check on this topic. However, it seems possible to plug ML on top of something like OpenFlow. PM me if interested to look this up further.

Yes, which is a reason why all non-linear clustering techniques can be a challenge when investigation follows outlier detection.

Well, you can try this one, it is open source and based on Caffe : http://imgdetect.alexgirard.com/# The image link: http://rocknrollnerd.github.io/assets/article_images/2015-05... It says 'studio couch, day bed' with 97% accuracy. Very likely because it is part of Imagenet.

Shameless plug: anyone interested should be able to get baseline results and above easily by using libcmaes. I am one of the authors with no time to compete, but am interested in reports on how it goes. Also if you are a researcher or a student the lib should let you experiment easily with various custom strategies.

https://github.com/beniz/libcmaes

You can compute the gradient, it just has a high budget cost.

The usual winner is a flavor of CMA-ES, though they may have picked up the functions to avoid this.

I believe this article to o be relevant here:http://m.theatlantic.com/magazine/archive/2013/11/the-man-wh...

yes, it'd need to be fully open hardware and open source software. What else ? :)

on parallelization:http://research.google.com/archive/large_deep_networks_nips2...

Fibonacci heaps are very interesting beasts, but implementation is slightly complicated. For C++ developers there's one in Boost that is not so easy to use quickly.

I've recently put an alternative one in C++11 here: https://github.com/beniz/fiboheap

leboncoin.fr is known for developing all necessary tools internally. So... they should hire this guy, that's all! losers...

reminds me of outrun :)

note that it does not really matter whether you are one of Orange's customers or not as they operate many of the inner and outbound backbones.

question is: is evolution 'target' for us to build machines or to leverage machines to increase its pace ? if pure AI were the target evolution may have 'targeted' it more directly. Singularity 'theory' forgets about evolution.

People wearing this watch do not seem very much interested in news articles ;) More seriously, my fuzzy estimate is that 80% of the time spent on my phone is for reading stuff that is definitely not the departure time of my next flight (that I can memorize five months in advance :) ).

Reinforcement learning (RL) is the future of real-world AI. Bear my words ;) RL has been around for a long time, it is the mix of optimal (in the optimization sense) decision making along with machine learning (ML). It does benefit of most recent advances in ML. As such it is likely to power the next batch of 'intelligent' applications and services.

Both my parents, and my sister have faced life threatening diseases, with lifetime treatment required. Call it bad luck. They benefit a 100% health insurance coverage for these diseases, for free. Had it not been the case, our family would not have been able to cope. This shared insurance system builds the society I want to live in.

IMO France now needs to rebuild its economy so it continues to support this gold standard system. Let's not let it degrade.

let s mine bitcoins as playing, a primitive Ender's game if you will.

solr/ES would do that almost out of the box. how does your stuff differ or make it simpler (besides cloud) ?

for the whole movie I thought it would be revealed that Sam was fro. an outsourced call center ;)

I m a search engineer among other things. New update is weird, especially ranking. try 'machine learning', gets something from a year ago then Stanford class from 2011, really ? If you re dealing with 10M entries in the index this is low enough that ranking should be better.

I know how perfect tuning these engine as is difficult, and this one requires some more work ;) good luck and congrats for YC 14 anyways!

they certainly open up to googlebots. Human generated content opened to bots but not to humans. bad SciFi...

Yes and this generates other tensions: those who fail the almost free higher education are totally left behind, with almost nochance to catch up (statistically speaking).

I love textsecure. My problems is twofold: I don't know in advance who is using it from my contacts, and I know for sure none of my non geek relations is using it, not even for business stuff, which could make sense. And so as much as I like it (thanks!!!) I am not convinced this is the solution. Am I wrong ?

don't know how others do but I find I have to recompile a program without optimization flags in order to be able to fully debug a live C++ program. Understandable but annoying.

try this https://apps.xplr.com/xplr/news/?filter=dc_source_hash:41f60... This is a demo of a professional system used by several industries to sort articles across the web.