AI Technology

HEROZ, Inc.(HEROZ) possesses a unique core technology of artificial intelligence (AI) that has defeated professional shogi (Japanese chess) players in public matches developed by our top engineers. HEROZ has accumulated such a machine learning technique through the development of smartphone apps such as “Shogi Wars”, “CHESS HEROZ” and “BackgammonAce”.
Meanwhile, as a supporting member of the Japanese Society for Artificial Intelligence, HEROZ keeps up with the trend of cutting edge AI, considering it as an important strategic field.
Currently, HEROZ focuses on planning, developing and operating intellectual games for smart phones, however, will apply and expand its cutting edge AI technology in various fields in the future.

History of Programming Intellectual Games (Mind Sports)

Intellectual games have been gaining attention for AI development from the old times.

Chess(since 1910s)

Possible variations to move = about10120

DeepBlue defeated a human player (1997)

Shogi(since 1970s)

Possible variations to move = about10220

AI developed by HEROZ’ engineer defeats an active professional Shogi player (2013)

Go(since 1970s)

Possible variations to move = about10320

AlphaGo defeated a human player (2016, 2017)

History of AI Development for Shogi

An innovation of machine learning takes place in 2004.
In the next 10 years, AI reaches to an equivalent level as the top human professional shogi players.

Computer Shogi Software A chart showing the improvement of AI

Innovation of Machine Learning

Development of AI for shogi enters into the world of automatic learning by machines and machine power (cloud, cluster connection).

Innovation of Machine Learning

Development of AI for shogi enters into the world of automatic learning by machines and machine

Before adopting machine learning

Before adopting machine learning

Developers are required to have a thorough knowledge and understanding on shogi and programming

Developers must capture the essential features of shogi and verbalize human minds

After adopting machine learning

After adopting machine learning

Knowledge on shogi is not required
Technology is applicable to various fields

Developers must view shogi as a drawing sheet and produce evaluation function for judgements and decision makings by machine learning from big data

An Image of Evaluation Function Production by Machine Learning

Computer Learns and Produces the Evaluation Function
f(x) for Judgements and Decision Makings

By developing intellectual games, HEROZ accumulates AI related methods such as machine learning with the
use of big data and deep learning that can be applied to and utilized in various fields in the future.