If you rate-limit on server side, it would mean that you would allow only a few legitimate users to use the site at any given time.
521 karma · joined September 5, 2014
Language is weird.
Personally, I think of the past being on the left, and the future being on the right. (Thai is written from left to right, like English.)
Also, there's a link to a page about laziness on the left.
Therefore, it is not obvious whether quantum computing will affect AI at all. Perhaps someone will figure out a new quantum algorithm for AI, but I don't think that's very likely to happen soon.
Misleading headlines and clickbait make it difficult to efficiently browse news feeds. Natural language processing and recurrent neural networks could soon be leveraged to generate unbiased, accurate summaries of articles and may replace traditional hand-tailored headlines.
A 2015 Stanford paper[1] describes a process for generating headlines from the text of news articles using recurrent neural networks. The process described in the paper utilizes the first 50 words of an article, and concludes that most of the time, the summary is valid and grammatically correct.
In the years since the Stanford paper was published, several open source projects have emerged that simplify the process of getting started with automatic headline generation[2][3]. The software is readily available - the next step is for developers to include these open source tools in their own consumer-facing applications.
It’s unlikely that major news outlets are going to adopt NLP-generated headlines any time soon, and even if they did, it’s unlikely they’d be employed to benefit the reader. Even so, there’s a place for these programatically generated headlines: in the hands of consumers and software developers working in the realm of news aggregation.
NLP-generated headlines could still be optimized for clicks without compromising their accuracy. Several versions of these headlines could be AB tested without any manual effort required on the author’s part. It’s also worth noting that journalists at major media outlets generally don’t write their own headlines anyway4, it’s usually done by editors.
Headlines could even be tailored to each reader: a technical reader may appreciate a more sophisticated headline that contains jargon, but a layperson might appreciate something more simplified. News reader applications and content aggregators seem like natural fits for this type of technology.
With the amount of time people spend scanning through headlines every day, a more effective way to summarize content could drastically improve the reader’s experience. Google became the best search engine in part because it trusted the PageRank algorithm more than webmaster-supplied summaries. In a very similar sense, it’s become difficult to rely on the headlines provided content creators, so it's quite possible that the concept of traditional headlines is ready for disruption just like to concept of search was 20 years ago.
Remote: Yes
Willing to relocate: Yes
Technologies (proficient in): Java, C/C++, Python, Rust, Haskell, Linux/Unix Command Line (Bash, Git, etc.), Advanced Data Structures and Algorithms, Mathematics
Technologies (experience with): Scala, JavaScript, HTML, CSS, PHP, SQL, Racket, Ruby, Standard ML, Arduino, Coq, TensorFlow (deep learning library), Deep Learning, Neural Networks, Programming Language Theory
Résumé/CV: https://drive.google.com/open?id=1NmnLugVSFOweRHHaFiCEzHqOAM...
Email: theemathas@gmail.com
I am a computer science student looking for a summer internship.
My abilities are guaranteed by my awards from programming competitions, including winning a gold medal from the International Olympiad in Informatics, with rank 22nd out of 311 contestants from 81 countries.
By the way, the Getchu and illustration2vec links on the news page are broken.
Edit: This part from the Tips page might be why it initially didn't generate great images:
The input of the model consists of two parts, the random noise part and the condition part. If you generate a good image, you could try to fix the noise part and use random conditions to get more good images. We have observed that a good random noise is important for the better generation.
Edit 2: Actually, no. According to the news page, if the noise is fixed, the generated pictures would be all similar.
In order to make our model more accessible, we build this website interface with React.js for open access. We also make the generation completed done on the browser side, by imposing WebDNN and converting the trained Chainer model to the WebAssembly based Javascript model. For a better user experience, we would like to keep the size of generator model small since users need to download the model before generating, so we replace the DCGAN generator by SRResNet generator can make the model 4 times smaller. Speed-wise, even all computations are done on the client side, on average it takes only about 6 seconds to generate a single image.
Everything is scheduled so it's asymptotically optimal. However, there's an added cost to search for algorithms. This cost is exponential in the proof length, but constant for each problem since it is independent from the inputs. Of course, this constant is ridiculously huge.
Abstract: An algorithm M is described that solves any well-defined problem p as quickly as the fastest algorithm computing a solution to p, save for a factor of 5 and low-order additive terms. M optimally distributes resources between the execution of provably correct p-solving programs and an enumeration of all proofs, including relevant proofs of program correctness and of time bounds on program runtimes. M avoids Blum's speed-up theorem by ignoring programs without correctness proof. M has broader applicability and can be faster than Levin's universal search, the fastest method for inverting functions save for a large multiplicative constant. An extension of Kolmogorov complexity and two novel natural measures of function complexity are used to show that the most efficient program computing some function f is also among the shortest programs provably computing f.
The catch is that the constants in the big-O are enormous.
Most important part of that section:
> If our community wants to remove implicit nulls from industry, where appropriate, our community should make greater effort to appreciate the complex compromises that have to be made in practice so that our work can successful transfer out of academia.
You can still get some of the code from the linked paper, though.
Relevant quote from the paper:
> When executed, a ClassCastException is thrown inside the main method with the message “java.lang.Integer cannot be cast to java.lang.String”. The reason is that the JVM does not directly support generics, implementing them through type erasure and type casts. This is fortunate since it causes the JVM to catch the error before it can directly cause severe problems, such as accessing memory associated with a field that is not actually present (though it might be possible that it can indirectly cause problems).