Supercomputer Watson suggests new recipes for adventurous cooks

Would you like discover new recipes? Created by a supercomputer? You can now ask Watson, IBM's supercomputer, to suggest an entirely new recipe for you that combines ingredients you might never have thought of putting together.

"Chef Watson" is an experiment to test if Watson could create entirely new and surprising recipes and ingredient combinations. After reading thousands of recipes and learning about food pairing theories and the hundreds of thousands of complex relationships between flavor compounds, Chef Watson has helped shape the public perception of the potential for cognitive computing.

You can try Chef Watson for free here or as an smartphone app.

IBM supercomputer Watson beats humans in Jeopardy

In May 1997, IBM's Deep Blue Supercomputer played a fascinating match with the reigning World Chess Champion, Garry Kasparov. Overall, Deep Blue won 2 games, 3 draws, and one loss. (See IBM Deep Blue page here)

Chess is a very logical game with a large, but limited number of possible moves. The real challenge for computers is still naturally spoken language, with all it's subtle contextual meanings, puns, irony, and word plays. Now a team at IBM has taken on the challenge of natural language processing with the Watson project (See IBM Watson page here)

Watson is a supercomputer the size of ten refrigerators that is able to analyze natural spoken language and generate an answer in natural language. An IBM team of scientists spent four years building the Watson computer system that can rival a human in answering questions posed in natural language. The team decided to have Watson compete against human players in the popular TV show "Jeopardy" which is a good testing ground for it's language analysis capabilities. In a first machine vs man test round, Watson beat the two all-time best Jeopardy players Ken Jennings and Brad Rutter. More rounds will take place next month. Read more here: IBM supercomputer wins first Jeopardy dust up - Computerworld

Despite the impressive progress of IBM's Watson, it will still take awhile until we can naturally talk to computers, as envisioned by Apple back in 1987:

To follow Asimov's rules of robotics, one must break them.

A robotics research team from Slovenia wants to establish how robots should move when around humans in order not to hurt them. The first step in this study was having an industrial robot arm hurt humans (yes, they got human subject permission for that). The study aims to establish how fast robots can approach humans without causing them pain.

Robot arm punches human to obey Asimov's rules - tech - 13 October 2010 - New Scientist

Aiming to Learn as We Do, a Machine Teaches Itself (NY Times article)

Researchers at Carengie Mellon University are working on a computer that can understand natural spoken language, but it has to learn it first. The Never-Ending Language Learning system (NELL) learns by scanning websites for semantic relationships between words. Full article here.