AI invented a new sport, learned to generate animation basing on a text, and outplayed the eSports team. Weekly news digest AI invented a new sport, learned to generate animation basing on a text, and outplayed the eSports team. Weekly news digest

OpenAI Five has once again beaten people in Dota 2, Disney taught a neural network to generate animation basing on texts, and Chinese citizens will get free medical aid in exchange for their personal data consequently used to train AI. Read about the latest news from the world of artificial intelligence further in the digest.

Artificial intelligence outplayed
OG eSports team

OpenAI Five managed to outperform the OG eSports team in Dota 2. The match was streamed on Twitch.

The developer team explained that specifically this online game was an excellent fit for deep reinforcement learning due to its complexity level. Remarkably, artificial intelligence experienced 45,000 years of play that lasted for 10 months in real time. On average, AI trained for 250 years per a day.

Previously, OpenAI beat the Dota 2 team comprised of three former professional players (Merlini, Blitz, and Fogged), one active professional player (MoonMeander), and one commentator (Capitalist).

Disney trained AI to generate
animation from texts

Disney’s research department developed an artificial intelligence algorithm that can generate short animated videos using text information.

Authors created a modular neural network comprised of three parts. The first conducts text parsing by separating words of characters from scene descriptions. The second processes natural language simplifying complex sentences. The third creates three-dimensional videos that roughly reflect the processed scenario. The program was trained using one thousand scenarios.

As a result, 68% of 22 volunteers that assessed AI-generated videos called them “quite meaningful”.

Artificial intelligence invented
a new kind of sport

Artificial intelligence developed by AKQA design agency came up with a completely new kind of sport – Speedgate. The neural network set up the rules having analyzed 400 existing sports. Unrealizable conditions were removed.

Two teams with six members in each participate in the new game. The playground looks like a football ground with the difference in the number of gates – there are three of them.

As soon as players of one of the teams kick the ball into the central gate, they can target the end gates. Every scored goal brings two points. The team member that catches the scoring ball and kicks it back through the gate straightaway, gets three points.

The ball has to move around every three seconds, so players cannot stand still for long.

Engineers designed a drone
to look for lost people

Engineers of the Swiss air rescue organization Rega have developed a drone helicopter that can search for lost people autonomously.

The machine is equipped with numerous embedded sensors including a daylight camera, thermal camera, infrared camera, and phone-tracking sensor. It allows Rega to scan vast areas looking for people lost in the woods or mountains.

Algorithms of the drone helicopter analyze data from sensors in real time and define potential locations of people. Besides, the drone reports the obtained data to the operator.

Chinese citizens get free medical
aid in exchange for data

In China, 130 private medical companies offer village dwellers free medical aid in exchange for full access to their medical history. They act as part of WeDoctor project that belongs to one of the largest Chinese companies – Tencent. The company uses the obtained data to train artificial intelligence the basics of therapy.

Even today, physicians can enter the patient’s symptoms and AI will predict the diagnosis with up to 90% accuracy basing on data of 5,000 symptoms and 2,000 diseases.

Learn more about the AI-technology ►►►

Related news