Speech mining and business intelligence

Corporations have large amount of speech data. Mining these data will provide lots of information to the business operation. The most obvious one is call center recordings. Mining these data will help to improve the quality of call center service, reduce call drop and so on.

Leading companies in this domain are BBN Technologies and Nexidia. New startup includes Callminer, based in Florida.

Robot that opens the door and cleans the kitchen

Imagine a robot that has an arm and hand. It grabs the handle of a door knob, turns it and open the the door. Then the robot moves quickly into the door. Watching this makes me feel something surreal. This action feels remarkably like a human. It is a little scary to see how far a robot has come. Check out this video from Andrew Ng's group:

The ability to grasp objects indicates that robotics has come a long way from its early days. It is triumph of both computer vision and the mechanical engineering of robot arms.

The kitchen clean robot uses the same grasping ability to put away plates. The group Readybot created this prototype:

Commercialization of AI

Artificial intelligence and its branches (machine learning, natural language processing) are seeing increasing opportunity to impact people's real life. Here are a few fields that can apply AI:

1. Computer games
As games become more sophisticated, the characters in the game can react and play an important part in the game. Certainly there is a social aspect that is solved by multi-player game, and AI agents sorely lacking in that capability. Is there a increasing role of for AI character? or smart game evolution?

2. Internet application
1) One of the most successful product is Google Translate (http://translate.google.com). With statistical training over large data, with user feedback, it continues to improve its accuracy.
2) recommendation on E-commerce site. Whether you are buying books (Amazon) or renting a DVD (Netflix), you can get recommendations for similar products. Figuring out such "similarity" is the job of AI problem. It takes into your account of personal preferences, the community your belong to, and product features.
3) Mining blog and sentiment

3. Robots
The simple household vacuum robot needs not much intelligence. But it has the basic capability of planning and mapping. Warehouse robots has similar functions, but can have additional capabilities (such as using arm to fetch things). Entertainment robots (appearing in Museum, trade shows) may need natural language interface. Ultimately, the humanoid robots will appear in households. They need the full-range of intelligence, from planning to natural language interface.

4. Portable devices (cell phone, telemedicine device)
They interact with people on daily basis. What if we build basic intelligence into these devices so that they truly become our assistants?

5. Pattern discovery/Statistical mining
It's the power of data. Retailers, and manufacturers rely on good data to improve their service and production quality. Data mining (or machine learning with large data) is the ultimate answer.