Last Week in AI
The newest developments this week in AI research and development.
Every week, Invector Labs publishes a newsletter that covers the most recent developments in AI research and technology. You can find this week’s issue below. You can sign up for it below. Please do so, our guys worked really hard on this:
From the Editor: The Challenge of Training at Scale
Training is one of the frequently overlooked elements of building machine learning solutions at scale. While training machine learning models seems relatively simple conceptually, it gets really complicated when applied to large models or to a large number of models. As a result, most organizations struggle getting past certain level of scale in their machine learning solutions given the infrastructure costs.
Despite the proliferation of a new generation of machine learning frameworks, the techniques for scaling the training of models have remained mostly constrained to large technology companies such as Google, Microsoft, Amazon, Uber or Facebook. The great thing is that those companies have started open sourcing some of their methods in the form of libraries and frameworks. Just this week, Uber and OpenAI open sourced Fiber, a new framework for parallelizing computations and training in machine learning models. Contributions like this are going to be key to streamline the implementation of large-scale machine learning solutions.
Now let’s take a look at the core developments in AI research and technology this week:
AI Research
Massively Scalable Reinforcement Learning
Google Research published a paper proposing SEED RL, a reinforcement learning architecture that can scale to thousands of machines.
>Read more in this blog post from Google Research
Training Self-Driving Cars in a Simulation
MIT AI researchers published a paper proposing a simulated environment with virtually infinite steering possibilities in order to train different policies in self-driving vehicles.
>Read more in this article from MIT News
Coordination and Collaboration in Multi-Agent Environments
Researchers from MIT and Harvard collaborated in a paper proposing a coordination method that facilitate collaboration multi-agent AI environments.
Cool AI Tech Releases
Uber-OpenAI Fiber
Uber and OpenAI have partnered to open source Fiber, a new platform for distributed and scalable training.
>Read more in this introductory blog post
A-Z of AI
Google and Oxford University launched a new portal explaining the fundamental concepts of AI in a very intuitive way.
>Check out the A-Z of AI portal here
Cortex
Cortex is a new open source framework for rapid machine learning deployment.
>Read more in their GitHub repository
AI in the Real World
COVID-NET
AI Researchers from University of Waterloo developed a deep neural network that can help spot signs of Covid-19 in chest x-rays.
>Read more in this coverage from MIT Technology Review
Supercomputers to Fight Coronavirus
IBM is cooperating with the White House and the Department of Energy to offer coronavirus researchers access to supercomputers.
>Read more in this coverage from VentureBeat
Human Compatible AI
TheNextWeb published a comprehensive analysis AI-legend Stuart Russell’s new book.
