Usually, every month we write an article about the best and most promising AI research papers from that month. You can check them out here:

• January
• February
• March

This month (and the following) in addition to that, we bring you the best twenty AI articles we have found amazing that month. This collection of articles should give you an overview of what happened that month in the AI industry both from technical, business and from an ethical perspective. Have fun!

A Neural Weather Model for Eight-Hour Precipitation Forecasting

Near the end of this month, Google presented MetNet – an AI system that can predict precipitation up to eight hours into the future. In essence, it is a deep neural network capable of predicting future precipitation at 1 km resolution over 2-minute intervals. Really cool thing is that MetNet takes a data-driven and physics-free approach to weather modeling. This means it learns to approximate atmospheric physics from examples and not by incorporating prior knowledge.

Coronavirus Datasets from Every Country with Confirmed Cases

COVID-19 took the world by storm. It is having a huge impact on everything from health to the economy. Different AI techniques can help in different ways. For those looking to perform research about the virus, there are numerous datasets have been made available and this article is trying to gather them all. It is continuously updated since we are getting more and more information about coronavirus. This article collects coronavirus datasets from all countries with confirmed COVID-19 cases.

A Visual Guide to Evolution Strategies

This really cool article explains how evolution strategies work with the emphasis on the visual representation of them. Evolution strategies are a very interesting field and alternative to Reinforcement Learning. Even though they are less data-efficient than RL, they offer many benefits. There is no much math but there are a lot of links to other useful articles that can help you understand evolution strategies better. 

A visual debugger for Jupyter

You know how they say “programming is 30% coding and 70% debugging”. Well, we have great news for all developers that use Jupyter Notebook out there. Jupyter Notebook is now fully-fledged IDE. You don’t have to use Notebooks just for experiments and then go to some more appropriate IDE for production development. Literate programming is now a reality through nbdev and the new visual debugger for Jupyter.

MIT’s deep learning found an antibiotic for a germ nothing else could kill

Healthcare is getting transformed by the AI in general, and this is just one of the best examples of how. Guys and gals from MIT created a deep neural network that can acquire a broad representation of the molecular structure. This means that this network can find out new antibiotics. They showed that when the compound is injected in mice, it fights bacteria that no existing drug can eliminate.

Microsoft researchers create AI ethics checklist with ML practitioners from a dozen tech companies

The problem with AI is that it will do what we tell it to do. These algorithms can become biased simply because we didn’t take care of the type of data and we ourselves were biased. In order to avoid this, Microsoft Research, together with nearly 50 engineers from a dozen tech companies, created an AI ethics checklist. Some of the things might sound weird to you, for example, that your teams should “define fairness criteria.”

Google Believes Machine Learning Frameworks Need Five Key Things to Reach Mainstream Developers

While Machine Learning certainly is not in the early adopters’ stage anymore, it has not gone full mainstream yet, hasn’t it? In this article, Google presents the result of the survey that they took with TensorFlow.js developers. The goal of the survey was to determine the key elements that should increase the adoption of machine learning frameworks. Check out the results.

Kids’ brains may hold the secret to building better AI

This article contains a conversation between Sigal Samuel and Alison Gopnik. Gopnik has a really cool TED Talk about what babies think. In this conversation, they discuss how four-year-olds can learn things even the most intelligent machine can’t and how AI research can benefit from that.

An implant uses machine learning to give amputees control over prosthetic hands

Another huge victory for AI in the healthcare domain. Researchers have been working on mind control prosthetics for decades now and this time they manage to amplify nerve signals to the point where they can be translated into movements. All thanks to the AI. Check it out.

How to use Jupyter Notebooks in 2020

This is really cool three-part series on how Jupyter Notebook should be used in this modern-day and age. In the first article, the author focuses on the data science landscape and how the changes in the field affect the tools that we use. In the second article of the series, the author discusses all the things he likes about Jupyter Notebook and all the things he doesn’t like. Finally, in the third article, he writes down his wishlist. 

Reinforcement-learning AIs are vulnerable to a new kind of attack

Reinforcement learning is a powerful “third paradigm of machine learning” that already shocked the world a few times. DeepMind used it for AlphaGo and AlphaGo Zero. Multiple applications are using it for process control. However, these types of algorithms are vulnerable to adversarial attacks. Find out how.

Google launches Cloud AI Platform Pipelines in beta to simplify machine learning development

More news from Google. This month they presented the beta version of Cloud AI Platform Pipelines. This service is designed to deploy robust, repeatable AI pipelines along with monitoring, auditing, version tracking, and reproducibility in the cloud. Focus is on “easy to install” environment for machine learning workflows.  As they say, this could reduce the time spent on bringing products to production.

The 10 most innovative artificial intelligence companies of 2020

I am still blown away by ways AI is used out there. Since today it is used more or less everywhere, here is the list of the 10 newcomers who use AI in the most innovative ways.

Neuroevolution of Self-Interpretable Agents

This really interesting article, suggests neuroevolution ideal for training self-attention architectures for vision-based reinforcement learning tasks. In essence, it gets inspiration from inattentional blindspots, a process that causes you to miss things in plain sight. Even though that sounds bad, this process actually helps us to focus on important parts of our world and reject unnecessary information that can only confuse us.

Intel’s new neuromorphic Pohoiki Springs system sports 100M artificial neurons

Neuromorphic hardware is an experimental technology that attempts to mimic the way the human brain processes information. Big players in the field are hoping that this will speed up computations provide better performance than traditional servers for artificial intelligence and machine learning workflows. Intel presented its solution Pohoiki Springs, for which they claim that it can solve certain problems 1,000 times faster than a regular processor using 10,000 times less power.