• This technology can make the Mona Lisa talk (sort of)
  Machine learning researchers have produced a system that can recreate lifelike motion from just a single frame of a person’s face, opening up the possibility of animating not just photos but also paintings. It’s not perfect, but when it works, it is — like much AI work these days — eerie and fascinating...

• The AI gig economy is coming for you
  The artificial-intelligence industry runs on the invisible labor of humans working in isolated and often terrible conditions—and the model is spreading to more and more businesses...

• Training a single AI model can emit as much carbon as 5 cars
  In a new paper, researchers at the University of Massachusetts, Amherst, performed a life cycle assessment for training several common large AI models. They found that the process can emit more than 626,000 pounds of carbon dioxide equivalent—nearly five times the lifetime emissions of the average American car (and that includes manufacture of the car itself)...

Python & R + SQL. Powerful & Shareable

Mode Studio combines a SQL editor, native Python & R notebooks, and visualization builder in one platform. Connect data from anywhere and analyze with your preferred language. Layer custom visualizations with HTML, CSS, and JS or use out-of-the-box charts.
Sign Up - Free Forever.

• Hearing your touch: A new acoustic side channel on smartphones
  Hackers may soon be able to tell what you’re typing - just by hearing you type; ML analyzes the soundwaves from each keystroke...

• How To Learn Data Science If You’re Broke
  Over the last year, I taught myself data science. I learned from hundreds of online resources and studied 6–8 hours every day. All while working for minimum wage at a day-care....

• MelNet: A Generative Model for Audio in the Frequency Domain
  Existing generative models for audio have predominantly aimed to directly model time-domain waveforms. MelNet instead aims to model the frequency content of an audio signal. MelNet can be used to model audio unconditionally, making it capable of tasks such as music generation. It can also be conditioned on text and speaker, making it applicable to tasks such as text-to-speech and voice conversion...

• Notes on the Limitations of the Empirical Fisher Approximation
  I was debating with myself whether I should write a post about this because it's a superbly written paper that you should probably read in full. There isn't a whole lot of novelty in the paper, but it is a great discussion paper that provides a concise overview of the Fisher information, the empirical Fisher matrix and their connectinos to generalized Gauss-Newton methods...

• Once again, a neural net tries to name cats
  Last year I trained a neural net to generate new names for kittens, by giving it a list of over 8,000 existing cat names to imitate. Without knowledge of English beyond its list of cat names, it didn’t know what letter combinations to avoid. So I decided to revisit the cat-naming problem, this time using a neural net that had a lot more context...

• High Frequency Component Helps Explain the Generalization of Convolutional Neural Networks
  We investigate the relationship between the frequency spectrum of image data and the generalization behavior of convolutional neural networks (CNN). We first notice CNN's ability in capturing the high-frequency components of images. These high-frequency components are almost imperceptible to a human...

• DeepMind’s AI has used teamwork to beat humans at a first-person shooter
  In a paper published in Science yesterday, DeepMind showed how it had let AI programs loose in a modified version of the 3D first-person video game Quake III Arena. The team used an algorithm called “For the Win,” which trains a host of agents in parallel using reinforcement learning...

• A Quantitative Approach to Product Market Fit
  This article will first go into greater detail of why we believe a quantitative approach to product-market fit is important. Then it will outline in detail three types of analyses we employ at Tribe Capital to understand product-market fit in a given company...

• DeepMind’s plan to make AI systems robust & reliable, why it’s a core issue in AI design, and how to succeed at AI research
  Today, I’m speaking with Pushmeet Kohli. For the last two years, Pushmeet has been a principal scientist and research leader at DeepMind. Before joining DeepMind, he was a partner scientist and director of research at Microsoft Research and before that he was a postdoctoral associate at Trinity Hall in Cambridge. He’s been an author on about 300 papers which have between them been cited at least 22,000 times. So thanks for coming on the podcast, Pushmeet...

• How To Define A Convolutional Layer In PyTorch
  Learn how to use PyTorch nn.Sequential and PyTorch nn.Conv2d to define a convolutional layer in PyTorch, via a screencast video and full tutorial transcript...

• Deep Learning Models
  A collection of various deep learning architectures, models, and tips for TensorFlow and PyTorch in Jupyter Notebooks...

• Introducing FastBert — A simple Deep Learning library for BERT Models
  FastBert is the deep learning library that allows developers and data scientists to train and deploy BERT based models for natural language processing tasks beginning with Text Classification...

• Guesstimation: Solving the World's Problems on the Back of a Cocktail Napkin
  "Guesstimation enables anyone with basic math and science skills to estimate virtually anything--quickly--using plausible assumptions and elementary arithmetic"...
  For a detailed list of books covering Data Science, Machine Learning, AI and associated programming languages check out our resources page
  .
   

  ---

   
  P.S., Want to reach our audience / fellow readers? Consider sponsoring - grab a spot now; first come first served! All the best, Hannah & Sebastian