Iconary is an AI-driven,

Pictionary ''Pictionary'' (, ) is a charades-inspired word-guessing game invented by Robert Angel with graphic design by Gary Everson and first published in 1985 by Angel Games Inc. Angel Games licensed Pictionary to Western Publishing. Hasbro purchas ...

-style online game developed at the

Allen Institute for Artificial Intelligence The Allen Institute for AI (abbreviated AI2) is a research institute founded by late Microsoft co-founder Paul Allen. The institute seeks to achieve scientific breakthroughs by constructing AI systems with reasoning, learning, and reading capabi ...

. Publicly released in February 2019, the game is designed to encourage collaborative communication between a human player and the AI player ''AllenAI''. Iconary is the first demonstration of an AI system capable of playing a Pictionary-like game with a human partner. Researchers at the institute intended Iconary to highlight the potential of collaborative communication with an AI agent in a game setting, as contrasted with other highly publicized AI-backed games that place humans in adversarial roles against an AI system. Pictionary creator

Robert Angel Robert Angel invented the popular word guessing game Pictionary, which was self published in his tiny apartment in 1985. Since selling Pictionary, he has gone on to invent board games and products. Most of his life is spent drawing fruit and surfi ...

noted of Iconary "it’s not me against you, it’s us as a team against another team”. The ''AllenAI'' system behind Iconary can compose scenes based on a given phrase for the human player to guess, as well as attempt to guess a phrase depicted by the human player. The ''AllenAI'' system combines elements of

computer vision Computer vision is an interdisciplinary scientific field that deals with how computers can gain high-level understanding from digital images or videos. From the perspective of engineering, it seeks to understand and automate tasks that the hum ...

natural language understanding Natural-language understanding (NLU) or natural-language interpretation (NLI) is a subtopic of natural-language processing in artificial intelligence that deals with machine reading comprehension. Natural-language understanding is considered an ...

, and

commonsense reasoning In artificial intelligence (AI), commonsense reasoning is a human-like ability to make presumptions about the type and essence of ordinary situations humans encounter every day. These assumptions include judgments about the nature of physical objec ...

to compose scenes or to make guesses about a scene. The system was trained using data from collaborative games between two human players, using over 75,000 unique phrases and a constrained set of 1,200 icons. The phrases used in Iconary describe complex scenes or activities, e.g., “people jogging in the park” or “engineer riding a train”. The limited vocabulary of icons requires players to create novel combinations of icons to communicate a phrase to their partner, and to make use of abstract symbols (like arrows to signify motion) or contextual clues (like a stethoscope next to a figure of a person to signify "doctor") to successfully depict a concept.

References

{{reflist 2019 video games Browser games Drawing video games Video games developed in the United States Word puzzle video games