Virtual worlds making artificial intelligence apps 'smarter'
Vendor says its technology will 'learn' by interacting with humans in virtual worlds
Computerworld - An artificial intelligence company next month will unveil plans for a program that will help its AI software "learn" by interacting with humans in virtual worlds.
Novamente LLC, a San Francisco-based AI company, plans to detail its new product and how it will operate it in virtual worlds next month at the Virtual Worlds Conference and Expo in San Jose.
Ben Goertzel, Novamente's CEO and chief scientist, said that virtual worlds are an ideal environment to "embody" AI software, to give it a form to allow it to "learn" by interacting with virtual animals and humans.
While many existing AI applications today -- like those used to predict the movement of a stock or to play chess -- focus on a specific task, the new AI programs do not conceptualize who or what they are so they can expand to solve additional problems, he said.
"If you want to make an AI [system] that has real self-understanding, that knows that it is there, that you are there and can generalize things from one context to another, [you should] give it some sort of body," Goertzel said "Giving an AI [application] a body seems to be the best way to get it to really learn and understand. Once you launch AI in a virtual world ... you can have a hundred thousand or a million teachers for it."
Some researchers use robots to embody AI programs, but Goertzel said that process requires complex programming capabilities. "Working in a virtual world lets you get the general advantage of embodiment but without the hassle or costs that comes from physical robotics," he added.
Goertzel, declined to say which virtual worlds Novamente and its partner The Electric Sheep Co., a maker of add-on software for virtual worlds, intend to use for the project.
He did note that the project will embody the AI programs as animals. For example, an AI system could be put into a virtual world as a dog, which a human could teach to play soccer, he said. "You could potentially teach it soccer through example by showing it what you are doing ... and then by reinforcing it. By interacting with you, observing and get rewards or correction from you, it can learn something. This could happen even if the guys who programmed it had never heard of soccer."
The company plans to use a parrot to help the system develop language skills as avatars in the virtual world talk to it, he added.
"Most of the AI out there now is based on rules that people program in," Goertzel said. It is not going to do anything that will surprise the programmers. Our system ... can learn new behaviors by interacting, with human characteristics, and they can do things that the programmers would have never imagined they can do."
Read more about Web Apps in Computerworld's Web Apps Topic Center.
- The Business Value of Continuous Delivery Download this whitepaper to learn more about the business value of Continuous Delivery and see why it could be a game changer for...
- Coding with JRebel: Java Forever Changed With JRebel, developers get to see their code changes immediately, fine-tune their code with incremental changes, debug, explore and deploy their code with...
- Ten Factors Shaping the Future of Application Delivery Download this research report conducted by Enterprise Management Associates (EMA) to learn how those that are seeking to accelerate application delivery are leveraging...
- Adobe Creative Cloud FAQ The following are answers to common questions about Adobe® Creative Cloud™ for teams membership, purchasing, security, and storage.
- Keep Servers Up and Running and Attackers in the Dark An SSL/TLS handshake requires at least 10 times more processing power on a server than on the client. SSL renegotiation attacks can readily...
- On Demand: Mastering the Art of Mobile Content Management Mobile device usage in the enterprise has skyrocketed, and it continues to escalate. IT must answer to users who demand access to their... All App Development White Papers | Webcasts