Deepfakes, ultrarealistic fake videos manipulated using machine learning, are getting pretty convincing. And researchers continue to develop new methods to create these types of videos, for better or, more likely, for worse. The most recent method comes from researchers at Carnegie Mellon University, who have figured out a way to automatically transfer the "style" of one person to another. "For instance, Barack Obama's style can be transformed into Donald Trump," the researchers wrote in the description of a YouTube video highlighting the outcome of this method. The video shows the facial expressions of John Oliver transferred to both Stephen Colbert and an animated frog, from Martin Luther King, Jr. to Obama, and from Obama to Trump.

Deepfakes, ultrarealistic fake videos manipulated using machine learning, are getting pretty convincing. And researchers continue to develop new methods to create these types of videos, for better or, more likely, for worse. The most recent method comes from researchers at Carnegie Mellon University, who have figured out a way to automatically transfer the "style" of one person to another. "For instance, Barack Obama's style can be transformed into Donald Trump," the researchers wrote in the description of a YouTube video highlighting the outcome of this method. The video shows the facial expressions of John Oliver transferred to both Stephen Colbert and an animated frog, from Martin Luther King, Jr. to Obama, and from Obama to Trump.

As part of the repository of data science tools I have been building over the past few weeks, I wanted to include some basic NLP / text-mining techniques in R. In thinking of a use case, I decided to kill two birds with one stone: teach myself a new skill, and build a cool gift for my SO! The result (link included below) is pretty basic, but hopefully you'll find it helpful! If you have ever wanted to work with text message data, are intrigued by NLP / text-mining in general, or you're the foremost expert on all things NLP (which I DEFINITELY am not), I hope there's something fun and insightful for all of you here. I look forward to your constructive criticism, and hope you enjoy this and/or any of other material I've assembled:)

In mid-September the European Union threatened to fine the Big Tech companies if they did not remove terrorist content within one hour of appearing online. The change came because rising tensions are now developing and being played out on social media platforms. Social conflicts that once built up in backroom meetings and came to a head on city streets, are now building momentum on social media platforms before spilling over into real life. In the past, governments tended to control traditional media, with little to no possibility for individuals to broadcast hate. The digital revolution has altered everything.

Considering music as a sequence of events with multiple complex dependencies, the Long Short-Term Memory (LSTM) architecture has proven very efficient in learning and reproducing musical styles. However, the generation of rhythms requires additional information regarding musical structure and accompanying instruments. In this paper we present DeepDrum, an adaptive Neural Network capable of generating drum rhythms under constraints imposed by Feed-Forward (Conditional) Layers which contain musical parameters along with given instrumentation information (e.g. bass and guitar notes). Results on generated drum sequences are presented indicating that DeepDrum is effective in producing rhythms that resemble the learned style, while at the same time conforming to given constraints that were unknown during the training process.

I implemented the Neural Processes paper in Pytorch, and found a few interesting things not discussed elsewhere as far as I can see. For example, the posterior is liable to collapse to a near point estimate, and this appears to be somewhat related to the formulation they use. I wrote this up in an informal blog post, thoughts/comments welcome!

He said the main part of its RL policy is implemented by a 4,000 dimensional LSTM network, which has roughly 100 million (10 9) parameters .He also said that in terms of Flops, it is like a honeybee brain. This means the OpenAI Dota2 bot has roughly the same number of parameters as honeybee brain (Assumption: real synapse weight can be represented by 1 Byte of memory, which is often the case in artificial neural network systems). If the former is more intelligent, it suggests that modern machine learning algorithm is more capable than natural selection at turning a certain size of parameters into intelligence. It is probable that the same algorithm can turn them into super-mice or super-cat intelligence.

I think that eager sucks. It does not fit my mental model. I am just worried that TF wants to attract PyTorch users, but a lot of the TF users actually prefer the current state.