Every two weeks, we find the most interesting data science links from around the web and collect them in Data Science Briefings, the DataMiningApps newsletter. Subscribe now for free if you want to be the first to get up to speed on interesting resources.
- A Recipe for Training Neural Networks
“Clearly, a lot of people have personally encountered the large gap between “here is how a convolutional layer works” and “our convnet achieves state of the art results”.” - How to hide from the AI surveillance state with a color printout
“Fortunately, as some researchers from the Belgian university KU Leuven have just shown, you can often hide from an AI video system with the aid of a simple color printout.” - Speedgate: World’s First Sport Generated by AI
“Developers from AKQA, a global innovation agency most known for working with some of the hottest brands and public figures, trained a recurrent neural network and a deep convolutional generative adversarial network on over 400 sports with the aim of creating a new and original sport.” - OpenAI MuseNet
“We’ve created Musenet, a deep neural network that can generate 4-minute musical compositions with 10 different instruments, and can combine styles from country to Mozart to the Beatles. MuseNet was not explicitly programmed with our understanding of music, but instead discovered patterns of harmony, rhythm, and style by learning to predict the next token in hundreds of thousands of MIDI files. MuseNet uses the same general-purpose unsupervised technology as GPT-2, a large-scale transformer model trained to predict the next token in a sequence, whether audio or text.” - Notes on AI Bias
“Such issues are not new or unique to machine learning – all complex organizations make bad assumptions and it’s always hard to work out how a decision was taken. The answer is to build tools and processes to check, and to educate the users – make sure people don’t just ‘do what the AI says’.” - Dear AI startups: Your ML models are dying quietly
“Be cautious, DL models die quietly. There is no error message, no 404 page, no notification. It might take months for you to detect that your model is quietly failing on you.” - IBM halting sales of Watson AI tool for drug discovery amid sluggish growth
“Citing lackluster financial performance, IBM is halting development and sales of a product that uses its Watson artificial intelligence software to help pharmaceutical companies discover new drugs, according to a person familiar with the company’s internal decision-making.” - One Model to Rule Them All
“The machine learning community focuses too much on predictive performance. But machine learning models are always a small part of a complex system.” - Selling data products is the wrong business model for AI startups
“The problem with trying to create universal data products in a competitive market is that everyone has access to the same information, and everyone responds to that information similarly.” - Using R To Analyze The Redacted Mueller Report
A lot of people are analyzing the Mueller report. - How your data scientists become bottlenecked
“Data scientists are worth their weight in gold but they can become ineffective and bottlenecked if they don’t know the customer, the business problem, or the data itself.” - Detecting Malaria with Deep Learning
AI for Social Good — A Healthcare Case Study - This Google Experiment Destroyed Some of the Assumptions About Representation Learning
“Initially, the Google study proved that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases both on the considered learning approaches and the data sets.” - Machine learning accurately classifies age of toddlers based on eye tracking
“We adopted a data-driven approach by using machine learning (Support Vector Machine (SVM) and Deep Learning (DL)) to elucidate factors that contribute to age-related variability in gaze patterns.” - NLP Learning Series: Part 4 – Transfer Learning Intuition for Text Classification
“In this post, I will try to use ULMFit model which is a transfer learning approach for text classification.” - AI Helps Classify Lung Cancer at the Pathologist Level
“Once the neural network was trained, the team evaluated the algorithm on a dataset of 143 images. According to the team, all evaluation metrics for the model were within 95% confidence intervals of agreement with the pathologist’s assessment.” - Speech synthesis from neural decoding of spoken sentences
“We designed a neural decoder that explicitly leverages kinematic and sound representations encoded in human cortical activity to synthesize audible speech. Recurrent neural networks first decoded directly recorded cortical activity into representations of articulatory movement, and then transformed these representations into speech acoustics.” - OpenAI bot crushes Dota 2 champions, and now anyone can play against it
Reigning International champions Team OG were soundly beaten over the weekend. - Amazon’s empire rests on its low-key approach to AI
Unflashy but high-powered machine learning powers everything from its fulfilment centres to the cloud - Name one thing in this photo
