Profile cover photo
Profile photo
Wayne Radinsky
17,977 followers -
Software Design Engineer
Software Design Engineer

17,977 followers
About
Wayne's posts

Post has attachment
The US Air Force did a test where unmanned F-16s flew alongside piloted F-16s in formation. It's part of a Air Force Research Laboratories program started in 2016 called Loyal Wingman, tasked with enabling autonomous planes to accompany human pilots into combat.

Post has attachment
A genetic programming system to "automatically architect a deep neural network with optimal hyperparameters for a given dataset" has been created. It's called DEvol (DeepEvolution) and uses Keras.

Post has attachment
Developers in Ukraine and Turkey get stuck in vim more. And, jQuery and CSS developers get stuck in vim more. But not C and C++ developers. Ruby even less than C and C++, oddly enough.

I use vim every day and it never even occurred to me anyone would have trouble getting out of it. But I guess if nobody's ever explained to you that vim is modal, which is the most fundamental thing about it, then you could have trouble.

Post has attachment
Rather than train your own machine learning models, you can just call APIs from Google, demoed here by Sara Robinson at Google I/O. The image API will tell you what objects are in an image. The natural language API gives you the transcription of a sentence, it's entities, sentiment, and syntax. Entities refers to the proper nouns that Google looks up in its knowledge graph. For sentiment, it can give you the overall sentiment, or the sentiment towards a specific entity. Syntax tells you the part of speech of every word and sentence diagram for the sentence. The translation API can translate to another language. The Video Intelligence API lets you understand your videos at shot, frame, and video level.

Post has attachment
Françoise Beaufays talks about using deep learning for keyboards and "deep internationalization" at Google I/O. She says the idea is to never have a line of code that says, if language == German, do this, else do that, etc. Everything about German should be learned from data. There are 1,342 languages with over 100,000 speakers.

Post has attachment
"Patreon doubles in a year to 1M paying patrons and 50K creators. Patreon's novel idea of fans just directly paying the artists they love is having its hockey stick moment."

Post has attachment
The WannaCry ransomeware has netted its creators 49.22604414 BTC, worth USD $110,807.83 at the current price of $2,251.00 per BTC (total from adding up the 3 bitcoin addresses hardcoded into the ransomware listed on this page).

Post has attachment
AlphaGo vs Ke Jie, the world's #1 ranked Go player, starts in under 2 hours. It's part of a "Future of Go Summit" taking place over the next 5 days in Wuzhen, China. Subsequent days will feature players playing against each other with AlphaGo on their team (1 human + AlphaGo vs 1 human + AlphaGo), and a 5-on-1 match pitting 5 humans against AlphaGo.

Post has attachment
A new deep learning algorithm can translate a traditional recipe for one culture into a different culture. So you could translate "lasagne" into Japanese culture and get a sushi lasagne. Sounds like "style transfer", but for recipes. Except the description of the algorithm makes it sound more like how the word2vec system can be used to make SAT-style analogies.

"The third part of the system takes ingredients and clusters them via what's known as a word2vec model. Commonly used in textual analysis, this is a way of quantifying associations between words, or, in this case, recipe ingredients. Similar words/ingredients are represented by similar associations of words/ingredients. The result is a rather deep way of saying how similar ingredients are to one another, and, as such, what ingredients can be substituted for other ingredients. This is how the system comes up with its alternative recipe variations."

Post has attachment
"An algorithm summarizes lengthy text surprisingly well." "An algorithm developed by researchers at Salesforce uses several machine-learning tricks to produce surprisingly coherent and accurate snippets of text from longer pieces. And while it isn't yet as good as a person, it hints at how condensing text could eventually become automated."

"The software is still a long way from matching a human's ability to capture the essence of document text, and other summaries it produces are sloppier and less coherent. Indeed, summarizing text perfectly would require genuine intelligence, including commonsense knowledge and a mastery of language."
Wait while more posts are being loaded