AI Conquers Poker

Here are five important AI stories from the week.

AI Conquers Multi-Player Poker (Facebook)

Facebook and Carnegie Mellon have built the first AI bot that is capable of winning multi-player poker (no-limit Hold’em), building upon the single-player success that Libratus demonstrated last year. Unlike chess and Go, poker is a game with hidden information - the AI cannot see the cards that are held by its opponents. This makes poker incredibly complex and challenging for an AI to win. Read the article for details on how the AI bot Pluribus conquered the game.

The Geopolitics of AI (Scientific American)

Whoever moves fastest in AI will be able to export its AI to the rest of the world. In other words, the race to AI is not a simply a technological race - it has significant geopolitical consequences. For example, will China’s AI or the U.S.’s AI be exported to Russia and India? To help establish proper rules of engagement, this article argues for an AI Trade Organization (AITO), modeled after the World Trade Organization.

ROI of AI in Enterprise (Fortune)

Compared to a year ago, business executives are less optimistic on how quickly AI will deliver a significant return on investment. Over 51% of those surveyed say it will take three to five years; a year ago, only 28% thought it would take that long. AI is harder to adopt than many believe but executives still firmly think it is worth investing in.

The Self-Driving Car Movement Cools (Fortune)

After laying off 190 autonomous vehicle employees earlier this year, Apple acquires engineers and intellectual property from struggling autonomous vehicle startup Drive.ai for less than $77 million. Drive.ai was once valued at $200 million. Both Apple’s moves and the declining fortunes of Drive.ai demonstrate just how sobering the progress in autonomous vehicles has been lately after years of optimism and hype.

XLNet Explained (ML Explained)

In the past 18 months, there have been many watershed moments in NLP. In December 2018, Google released BERT, trouncing previous records in many NLP tasks. A few weeks ago, researchers at Carnegie Mellon University and Google Brain released XLNet, which beat BERT and is now state of the art. This article does a beautiful job explaining just how XLNet achieved its superior performance.

More Stories Worth Reading and Watching…

XLNet, the Latest State of the Art in NLP (Borealis AI)

Major Misconceptions of China and AI (Fortune)

Ankur Patel