1. Speed up data training with data echoing

This paper describes data echoing, a technique that can speed up AI training.

2. The AI technique that could provide the ability to reason

“At six months old, a baby won’t bat an eye if a toy truck drives off a platform and seems to hover in the air. But perform the same experiment a mere two to three months later, and she will instantly recognize that something is wrong. She has already learned the concept of gravity.” Yann LeCun, chief AI scientist at Facebook, hypothesizes that a lot of what babies learn about the world is through observation. And that theory could have important implications for researchers hoping to advance the boundaries of AI.

3. Hold ’em or fold ’em

Pluribus, a poker-playing algorithm, can bluff as well as a human.

+ But don’t get excited. It could wreck the online poker industry—so it’s not being released.

4. Predicting the generalization gap in deep neural networks

Here’s a description of a new technique that uses margin distributions to better predict a DNN’s generalization gap.

5. Anatomy of an enterprise-scale AI strategy

Looking to move beyond point solutions and proofs of concept? Maria Korolov lays out what it takes to develop an AI strategy.

IN COLLABORATION WITH Unravel

Free ebook: Rebuilding Reliable Data Pipelines Through Modern Tools

Data applications that combine customer data from multiple business channels can fail for many reasons. Identifying the cause and finding a fix is both challenging and time-consuming. This practical ebook will show you how to plan, migrate, and operate data workloads and data pipelines using cloud-based and hybrid deployment models. And it’s free, courtesy of Unravel.

6. Companies plan to double AI projects in the next year

According to a recent survey, 59% of companies have already deployed artificial intelligence and machine learning, with an average of four AI projects already in place and six more planned for the next year. By 2022, organizations plan to have an average of 35 AI or machine learning projects in place.

+ Want to reach that market? Check out sponsorship opportunities for O’Reilly AI Conferences.

7. Monitoring unrest from space

More than half of the active volcanoes in the world are not instrumentally monitored. But a new system, MOUNTS (Monitoring Unrest from Space), analyzes satellite images using AI and can integrate many diverse datasets to monitor volcanoes.

8. Are you sure you want to post this?

An Instagram AI can identify potentially offensive comments and ask users if they actually want to post them. The tool detects mean or harmful comments and sends an in-app pop-up notification asking, “Are you sure you want to post this?” The good news? In early tests, some users decided they don’t really want to.

See the TensorFlow World program

The TensorFlow World program is now live. New speakers and intriguing new sessions are being added every day, but the lineup for the inaugural TensorFlow World is already spectacular. If you want to be a part of this unique gathering right from the beginning, make your plans to join us in Santa Clara, October 28–31, today to secure your spot.

9. A (very) technical look at creating an AI to restore and colorize photos

“Editor’s note: This article is highly technical and will not be easy reading for most.” But it is fun.

10. Mansplaining, explained

At the upcoming O’Reilly Strata Data Conference in New York, an all-women panel from AWS will discuss using ML and AI to identify and raise awareness about mansplaining.

The price for the O’Reilly AI Conference goes up Aug 2

Early Price for the O’Reilly AI Conference in San Jose—that’s up to $400 off. Which makes it a lot easier to convince your boss that attending the AI Conference is a no-brainer. But it expires at midnight on Friday, August 2, so hurry.