The Wild Week in AI — JupyterLab; China overtakes US in AI funding; Malicious AI Report; Continual Learning advancements;
You can sign up for the newsletter here: https://www.getrevue.co/profile/wildml/archive/98307
JupyterLab is the next-generation web-based interface for Jupyter, the popular notebook tool for researchers and data scientist. With the latest release, JupyterLab is now ready for daily use.
According to CB Insights, China has overtaken the US in the funding of AI startups. The country accounted for 48 percent of the world’s total AI startup funding in 2017, compared to 38 percent for the US.
The report surveys the landscape of potential security threats from malicious uses of artificial intelligence technologies and proposes ways to better forecast, prevent, and mitigate these threats. It builds on a 2-day workshop held in Oxford, UK, in February 2017.
The Uber AI Residency is an intensive one-year research training program slated to begin this summer. Taking inspiration from similar programs, Uber AI Labs created the Uber AI Residency to allow up-and-coming researchers accelerate their careers in machine learning and AI research and practice.
Posts, Articles, Tutorials
Using deep learning algorithms trained on data from 284,335 patients, the team was able to predict cardiovascular risk factors from retinal images with surprisingly high accuracy for patients from two independent datasets of 12,026 and 999 patients. The algorithm could distinguish the retinal images of a smoker from that of a non-smoker 71% of the time.
An interesting discussion between Professors Yann LeCun and Christopher Manning on “What innate priors should we build into the architecture of deep learning systems?”
A walkthrough of the Dynamic word embeddings for evolving semantic discovery paper, with proposes a technique that allows you to track how the meaning of words changes over time.
New research from Baidu attempting to learn speaker characteristics from only a few utterances, commonly known as “voice cloning.” Check out the cloned audio samples here.
Code, Projects & Data
This is the code for implementing the MADDPG algorithm presented in the paper: Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments. It is configured to be run in conjunction with environments from the Multi-Agent Particle Environments (MPE).
A Keras reimplementation and tutorial of One pixel attack for fooling deep neural networks. It causes a deep neural network to misclassify an image by modifying the color of one pixel only.
Alpha release. An open-source conversational AI library, built on TensorFlow and Keras, and designed for NLP and dialog systems research and implementation and evaluation of complex conversational systems.
Highlighted Research Papers
ToMnet uses meta-learning to build models of the agents it encounters, from observations of their behavior alone. Through this process, it acquires a strong prior model for agents’ behavior, as well as the ability to bootstrap to richer predictions about agents’ characteristics and mental states using only a small number of behavioral observations. The authors apply the ToMnet to agents behaving in simple gridworld environments, showing that it learns to model random, algorithmic, and deep reinforcement learning agents from varied populations.
A method for learning useful skills without a reward function by maximizing an information theoretic objective using a maximum entropy policy. On a variety of simulated robotic tasks, the authors show that this simple exploration objective results in the unsupervised emergence of diverse skills, such as walking and jumping
Some real-world domains are best characterized as a single task, but for others this perspective is limiting. Instead, some tasks continually grow in complexity, in tandem with the agent’s competence. In continual learning, also referred to as lifelong learning, there are no explicit task boundaries or curricula. The authors propose a novel agent architecture called Unicorn, which demonstrates strong continual learning and outperforms several baseline agents on the proposed 3D domain. The agent achieves this by jointly representing and learning multiple policies efficiently, using a parallel off-policy learning setup.
The ability to learn from continuous streams of information is crucial for computational learning systems and autonomous agents (inter)acting in the real world. In this review, the authors critically summarize the main challenges linked to continual lifelong learning for artificial learning systems and compare existing neural network approaches that alleviate, to different extents, catastrophic interference.
In his paper, the authors present our study on an end-to-end learning system for spoken language understanding. With this unified approach, one can infer the semantic meaning directly from audio features without the intermediate text representation. This study showed that the trained model can achieve reasonably good results and demonstrated that the model can capture the semantic attention directly from the audio features.