Open-Minded AI
|
Open-Minded AI |
 |
Large-scale collaborative projects harvesting both networked computer power and human brain power in an effort to build AI on the Web.
|
|
|
|
|
|
|
Elsewhere on the Web
|
• Artificial Intelligence in the World Wide Web
• Researchers Muster Online Volunteers for Collective Brainpower
|
|
|
| |
|
Teaching common sense to a computer must be one of the most difficult problems for the years to come. I'm sure that people at Cycorp would agree with this statement: they invested nearly 350 person-years of effort over 17 years to create a knowledge base that would supply the basic knowledge needed to be applicable to many different applications. By building a sufficiently large knowledge base with this general knowledge, it is hoped that the system will be able to learn or create new inferences by itself. Their main product, Cyc, is built upon a core of over 1,000,000 hand-entered assertions. As you can imagine, feeding this amount of information into a computer is a very tedious task. It involves equipping Cyc with the ability to cut through the layers of ambiguity inherent to the world we are living in. For example, the sentence "Napoleon died in 1821 - Wellington was saddened" required two months to enter into the knowledge base all the information needed to explain the concepts of life and death!
Dr. David Stork, chief scientist of Ricoh Silicon Valley and founder of the Open Mind Initiative, estimates that in America alone something like 100 million mouse clicks have been wasted on a game of solitaire. Applying the idea of distributed computing not to the idle processor cycles but to the idle human brainpower, Stork and Push Singh of MIT's Media Lab started collecting data contributed by non-expert Web users or "netizens". In their own words, "...a key ingredient needed for the development of AI software is data. There is theoretical and experimental evidence that it is the lack of data that is retarding development of many systems such as speech recognizers, handwritten character recognizers, common sense reasoners and many others." So, starting with their first project, Open Mind Commonsense, the goal was to make it easy and fun for people to work together, giving computers the millions of little pieces of ordinary knowledge that constitute "common sense". Everyone has common sense (or should I said almost everyone?), so everyone could participate. Non-expert users answer simple questions, fill in the blanks, describe the relation between two words, describe pictures, all in order to build a database of common sense facts.
Open Mind Web site now encourage visitors to teach a computer to recognize speech or handwriting, and not just for fun. Financial incentives like lotteries and e-coupons might be incorporated in the future. But making such projects more fun still remains one of the main goals for Stork and his team. Building "true AI" just by playing games on the Net? I'm sure many of us would join.
Handwriting recognition requires large sets of training data, as do other Open Mind projects. Led by Dr. Lambert Schomaker of the UNIPEN Foundation, the Open Mind Handwriting recognition project seeks to collect data and develop recognizers for a variety of handwritten data. Netizens view handwritten characters or handwritten words in a browser and type in a transcription. The unique feature of all such projects is the use of a technique called "interactive learning", where only the tasks that are most informative to the system are presented to users.
The Open Mind Speech project is another important part of the Initiative aiming to develop free speech recognition tools and applications, as well as collect speech data from netizens.
Recently announced "Animals" is an experimental program demonstrating the overall capabilities of and design issues related to the Open Mind Initiative. It is a simple game in which the current player thinks of an animal and the computer tries to guess it.
Open Mind isn't the only project of its kind: The Mindpixel Digital Mind Modeling Project claims to be the world's first program attempting to become artificially conscious by talking to Internet users. Its goal is to collect and validate 1 billion MindPixels - simple true or false statements - over the next 10 years. Its founder, Chris McKinstry, thinks that quality control is crucial for such large-scale distributed projects. His system starts by asking users for a mindpixel, and continues with a request for validation of 20 statements entered by other volunteers. The validation process ensures the correctness of the provided information. The volunteers are rewarded with the equity in what could eventually become a for-profit corporation licensing Mindpixel Corpus or Corpus derived products.
I know that many of the established AI researchers will disagree with such methods for collecting common sense knowledge and question the ideas behind the projects described above. However, the Web-based front ends for the Open Mind Initiative and the MindPixel project are very simple and easy to use, so you can try and judge for yourself.