How artificial intelligence shapes the future of interactive games
Today in the blog of the “Game Management for Internet Projects” program , we prepared and translated for you an article written by Kill Screen chief editor Zak Bajor about artificial intelligence using the example of The Suspect and the film “She” .
In the film “She” (2013), the main character Theodore plays a game in which he comes across an insane swearing cartoon character with artificial intelligence. The little foul-mouthed alien enters into conversation with Theodore, remaining stubbornly useless. Realizing that this may be a test, Theodore curses him in response, he begins a verbal skirmish, at the end of which the alien shows Theodore a further way.
')
“She” is science fiction, but such interactive games as in this film may soon become a reality. The Alien Child game, created by David O'Reilly, gives some insight into how artificial intelligence (AI) and machine learning can make the games of the future exciting and interactive like never before.
“The most fascinating thing about using machine learning is that the AI studies us, as well as it studies itself and what we write to it,” says Guy Gadney, director of the interactive thriller with AI “The Suspect”.
The Suspect uses machine learning in the conversation of players with the fictional criminal Marcus Winter. To save the people he holds hostage, players need to use interrogation tactics, trying to outwit the criminal.
As in interaction with a chat bot, players respond to Winther with text messages, to which the AI responds on the fly. Setting a goal to achieve a sense of natural conversation with elements of slyness, similar to that depicted in the film “She”, Gadney uses machine learning to create natural saturation and tension. “We wanted to achieve this surprise, which happens only when the car goes beyond the limits of common sense and creates a situation that no one expected ,” he says. “As humans, we tend to share information. It makes everything very believable . ”
Software in the traditional sense is a set of strict rules that are followed by the machine to the letter. Usually, a computer only executes a chain of commands, but machine learning allows the computer to detect rules over time and experiments. “Machine learning is defined as a class of programs in which performance improves over time ,” explains Pradeep Dubey, an Intel scholar and director of the Parallel Computing Lab. “The more data a system consumes, the better it gets.”
George Dolbier, Technical Director, IBM Interactive Media in North America, describes machine learning as “a way to write software, not with code, but with data . ” “The program makes its choice, performs actions based on statistics, and changes behavior based on new data, ” he says.
Due to its flexibility, machine learning and AI technologies open up a whole world of new creative possibilities, says Dolbier. They can be applied “in almost all aspects of the gaming experience in the future, for example, in solutions, how to adjust the level of difficulty and reward system for special players, analyzing large amounts of player data or creating absolutely fictional characters with whom we can interact” - he says.
In May, for example, YouTube-blogger SethBling wrote a program called MarI / O , which learned to play Super Mario World (1990) . Unlike other AI, she started without preliminary instructions, and, she learned to pass the level like ordinary players: through trial and error. MarI / O did not need to “think” how to pass the game. After each death, she repeated successful actions and tried random ways to avoid failure. Despite being close to human abilities, the MarI / O machine mindset still remained obvious due to its focus on reaching the final goal, ignoring temptations such as bonuses and coins along the way - the human player probably wouldn’t lead yourself in a similar way.
Perhaps machine learning is not yet ready to fully reproduce human behavior, but Gadney believes that it can bring more human into the interaction of people with technology. “As we grow, we experience the world through emotions and experience,” he explains. “And if we can use machine learning to reproduce this experience and make it interactive, I think we are doing what we need . ” The Suspect’s uniqueness lies in machine learning, but this does not mean that the game writes itself or that the plot is arbitrary. It was used in combination with other, more traditional forms of game design.
Gadney says it’s best to take machine learning in game design as another tool like a brush. This tool is far from perfect. In March, Microsoft's Twitter bot, Tay, was supposed to learn how to conduct an informal conversation with those who interact with it. But after 24 hours of directional trolling , she began to hate everyone and spew curses, and was turned off. This case proves the need for a balance between the granting of freedom and the observance of certain rules. Such as taking into account the existence of trolls on the forums.
Probably one day machine learning will become as flawless and individual as it was shown in the movie “She” . Today, the games are far from this, but, according to the creator of Alien Child, David O'Reilly , a whole new world opens up in which players can talk and build more natural relationships with AI. Machines will still give odds to the human intellect, says Dubey. “Now this is not just a theory. In practice, cars sometimes show much more accurate results than the best of people . ”
On this joyful note, we congratulate everyone on the upcoming New Year, we will see what is new in the field and the new year will bring us. Well, we hope to please you with interesting translations in the new year. Subscribe to the blog. And also at the beginning of 2017 we can meet with you in person at our open events: the open day for educational programs of the gaming industry on January 25 and at our conference on creating games on February 11. Admission is free everywhere, just by registering. Do not forget to pass it.
If you have questions, write them in the comments, I will be glad to discuss.
How artificial intelligence shapes the future of interactive games.
Link to the original.In the film “She” (2013), the main character Theodore plays a game in which he comes across an insane swearing cartoon character with artificial intelligence. The little foul-mouthed alien enters into conversation with Theodore, remaining stubbornly useless. Realizing that this may be a test, Theodore curses him in response, he begins a verbal skirmish, at the end of which the alien shows Theodore a further way.
')
“She” is science fiction, but such interactive games as in this film may soon become a reality. The Alien Child game, created by David O'Reilly, gives some insight into how artificial intelligence (AI) and machine learning can make the games of the future exciting and interactive like never before.
“The most fascinating thing about using machine learning is that the AI studies us, as well as it studies itself and what we write to it,” says Guy Gadney, director of the interactive thriller with AI “The Suspect”.
The Suspect uses machine learning in the conversation of players with the fictional criminal Marcus Winter. To save the people he holds hostage, players need to use interrogation tactics, trying to outwit the criminal.
As in interaction with a chat bot, players respond to Winther with text messages, to which the AI responds on the fly. Setting a goal to achieve a sense of natural conversation with elements of slyness, similar to that depicted in the film “She”, Gadney uses machine learning to create natural saturation and tension. “We wanted to achieve this surprise, which happens only when the car goes beyond the limits of common sense and creates a situation that no one expected ,” he says. “As humans, we tend to share information. It makes everything very believable . ”
Software in the traditional sense is a set of strict rules that are followed by the machine to the letter. Usually, a computer only executes a chain of commands, but machine learning allows the computer to detect rules over time and experiments. “Machine learning is defined as a class of programs in which performance improves over time ,” explains Pradeep Dubey, an Intel scholar and director of the Parallel Computing Lab. “The more data a system consumes, the better it gets.”
Machine learning and artificial intelligence technologies open up a whole world of creative possibilities.
George Dolbier, Technical Director, IBM Interactive Media in North America, describes machine learning as “a way to write software, not with code, but with data . ” “The program makes its choice, performs actions based on statistics, and changes behavior based on new data, ” he says.
Due to its flexibility, machine learning and AI technologies open up a whole world of new creative possibilities, says Dolbier. They can be applied “in almost all aspects of the gaming experience in the future, for example, in solutions, how to adjust the level of difficulty and reward system for special players, analyzing large amounts of player data or creating absolutely fictional characters with whom we can interact” - he says.
In May, for example, YouTube-blogger SethBling wrote a program called MarI / O , which learned to play Super Mario World (1990) . Unlike other AI, she started without preliminary instructions, and, she learned to pass the level like ordinary players: through trial and error. MarI / O did not need to “think” how to pass the game. After each death, she repeated successful actions and tried random ways to avoid failure. Despite being close to human abilities, the MarI / O machine mindset still remained obvious due to its focus on reaching the final goal, ignoring temptations such as bonuses and coins along the way - the human player probably wouldn’t lead yourself in a similar way.
Perhaps machine learning is not yet ready to fully reproduce human behavior, but Gadney believes that it can bring more human into the interaction of people with technology. “As we grow, we experience the world through emotions and experience,” he explains. “And if we can use machine learning to reproduce this experience and make it interactive, I think we are doing what we need . ” The Suspect’s uniqueness lies in machine learning, but this does not mean that the game writes itself or that the plot is arbitrary. It was used in combination with other, more traditional forms of game design.
Gadney says it’s best to take machine learning in game design as another tool like a brush. This tool is far from perfect. In March, Microsoft's Twitter bot, Tay, was supposed to learn how to conduct an informal conversation with those who interact with it. But after 24 hours of directional trolling , she began to hate everyone and spew curses, and was turned off. This case proves the need for a balance between the granting of freedom and the observance of certain rules. Such as taking into account the existence of trolls on the forums.
Probably one day machine learning will become as flawless and individual as it was shown in the movie “She” . Today, the games are far from this, but, according to the creator of Alien Child, David O'Reilly , a whole new world opens up in which players can talk and build more natural relationships with AI. Machines will still give odds to the human intellect, says Dubey. “Now this is not just a theory. In practice, cars sometimes show much more accurate results than the best of people . ”
On this joyful note, we congratulate everyone on the upcoming New Year, we will see what is new in the field and the new year will bring us. Well, we hope to please you with interesting translations in the new year. Subscribe to the blog. And also at the beginning of 2017 we can meet with you in person at our open events: the open day for educational programs of the gaming industry on January 25 and at our conference on creating games on February 11. Admission is free everywhere, just by registering. Do not forget to pass it.
If you have questions, write them in the comments, I will be glad to discuss.
Source: https://habr.com/ru/post/318648/
All Articles