Profile cover photo
Profile photo
Janet Wagner
415 followers -
Technical Writer and Journalist
Technical Writer and Journalist

415 followers
About
Posts

Post has shared content

Post has shared content

Post has shared content
Nice walk-through paragraph-by-paragraph by Aman Agarwal and Qiang Lu, in a seminal paper «Playing Atari with Deep Reinforcement Learning» from Google's DeepMind (2013).
Add a comment...

Post has shared content
I guess I have written enough code & text to have the confidence to Share now.

Trying to predict earthquakes. My suspicion is that current approaches (usually predicting on past seismic events) won't get very far, the events being part of such a big multivariable system. But throw in some more data, like radio anomalies, that have been noted as precursors, might have a chance.

https://elfquake.wordpress.com/
Add a comment...

Post has shared content
A newspaper article about Geoffrey Hinton , called the 'Godfather of AI', in the popular style of The London Telegraph, which mixes intimate interview with Terminator's sci-fi images.
Add a comment...

Post has shared content
Semantic segmentation of images is getting really good now. The inverse problem of generating "photographs" from semantically described images is also starting to show dramatic progress (Chen and Koltun):
Abstract: https://arxiv.org/abs/1707.09405
Paper: https://arxiv.org/pdf/1707.09405.pdf
Video: https://www.youtube.com/watch?v=0fhUJT21-bs&feature=youtu.be
Add a comment...

Post has shared content
AI pioneer Andrew Ng says his new online course will help build ‘an AI-powered society’

Lots of people will tell you they’re nervous about the changes artificial intelligence will bring to the world, but Andrew Ng is confident it’s all for the best. The former AI chief of Baidu and founder of Google Brain is on a mission to build what he calls an “AI-powered society” — one where smart computers are as integral to businesses as electricity. And to bring about that future, Ng, now an adjunct professor at Stanford, will share what he knows best by teaching. Today, Ng is launching a new course on deep learning on Coursera, the online education site he co-founded. The syllabus will follow his popular machine learning course, which has attracted some 2 million enrollments since its launch in 2011....Not everyone will be able to leap into the course, says Ng, they’ll need “some basic machine learning or coding knowledge.” He also recognizes that there are plenty of introductions to deep learning already online, but says these are spread about on blogs and YouTube channels, making it difficult for individuals to get the whole picture. “I wanted to create a very clear path for anyone,” says Ng. “After someone finishes this course, they’ll be able to put deep learning on their resume with confidence and apply modern AI to a large variety of problems.” The course has video elements and programming exercises, and it will teach users basic concepts before moving on to how to implement them. “In the first few weeks you code up deep learning algorithms yourself, pretty much from scratch,” says Ng.
Add a comment...

Post has shared content
Baidu teaches an agent to speak using interactive conversation:
"This paper presents an interactive setting for grounded natural language learning, where an agent learns natural language by interacting with a teacher and learning from feedback, thus learning and improving language skills while taking part in the conversation."
Web: http://research.baidu.com/learning-speak-via-interaction/
Paper: https://arxiv.org/pdf/1705.09906.pdf
Add a comment...

Post has shared content
Add a comment...

Post has shared content
The new version of the Google GMail Smart Reply is no more based on LSTM but rather based on hierarchical models (deep learning ?) of language which are said to be more efficient. In the Google research blog post, there is a reference to Ray Kurzweil's pattern recognition theory of mind (PRTM) that aroused my curiosity. Let us remember that Kurzweil is Director of Engineering at Google and is working on projects involving machine learning and language processing.
Add a comment...
Wait while more posts are being loaded