Current Problems with Care-Giving Robots

Although care giving robots can perform diverse chores such as reminding people to take medicine, calling hospitals in case of an emergency, and hold simple conversations [18], they still have issues with many necessary tasks.

For example, robots lack "common sense". To execute tasks such as bringing a bedridden user food and drink, they would have to be programmed to make sure a glass is right-side-up before attempting to fill it.

To solve the issue of common sense, the artificial intelligence community has been working on creations such as CYC. CYC is a general purpose "encyclopedia" of common sense knowledge.[19]

Another useful realm that has been plagued with trouble is the field of artificial conversation. However, we have yet to produce a robot with an adequate ability to converse. Most programs designed have the same abilities as children aged three to five. This is hardly enough to keep most people entertained.

Similarly, there are current issues with voice-recognition software. The industry is still in its infancy, but hopefully researchers will be able to improve the quality of recognition as time goes on. In the mean time, computers CAN recognize words people say, to an extent. In one personal instance, I tried to have the computer write the word "please" to a text file, only to have it be translated as "police". Such misinterpretations could be worked around with a "confirmation system". The program could ask the user to clarify the meaning via a touch screen, keyboard, or other interface, and use a neural-net style system to "re-learn" the user's way of speak. (For an example of a neural-net that "learns" based on the results a user puts into a computer, check out 20q.net, a 20-questions style game.)

Some problems that robotic builders face, such as getting two robots to navigate past each other in a corridor, are very difficult to program mathematically. However the same issue could be resolved by programming robots instinctively. For instance, each robot could be programmed to try to slide by each other, like how people turn to the side to get past each other in a tight space.

Another problem, aside from the issues of what we know how to do, is the issue of price. Currently, robots are impractical for home use because they are so expensive. One possibility would be to produce a general "robotics operating system" under which robotics commands could be run more easily. Something like this could save a lot of time in research and development. Next, a set of mass-market "off the shelf" components could be produced, allowing the robotics developers to get on with "the real stuff" instead of having to re-invent the "simple things". Thus, researchers could spend more time solving the problems of voice and image recognition instead of having to design and reconstruct robotic arms.[18]

At first glance, this doesn't seem like it would reduce the cost. However, mass-producing the parts would reduce the price robotics manufacturers would have to pay to have the parts custom made.[18]