Citizen Sensor Network

From Knoesis wiki
Jump to: navigation, search

Citizen Sensor Network


Human Computation

One of the recent trends in computing has been to utilize human cognitive abilities over the Web to solve hard AI problems. These are problems that are not straightforward to solve with currently known computer algorithms. For example image identification and natural language processing are tasks that humans champion but the prevailing computer algorithms are childish at best. Such problems, if dispersed within a group of humans, can be solved far more quickly and accurately. Systems such as the Amazon Mechanical Turk already exploit the power of humans over the Web where a hard computation task (such as identifying entities from a satellite image) is distributed among human participants for a fee. This trend of using humans as computing devices is known as human computation.

The concept of human computation has been used in the early work related to interactive genetic algorithms. Although the conceptual idea has been around for a long time, recently it has gained attention due to the introduction of [www.captcha.net/captcha_crypt.pdf CAPTCHA] and the subsequent spinoffs. The most no-table human computation effort in the recent history can be attributed to the Google Image Labeler where the perceived objective for the participants is to make high scores in a game but the real objective is to utilize the human cognitive ability to provide better labeling for the Google’s image index. Another notable system is Amazon mechanical Turk where the participants are paid for providing their services and tasks can be uploaded for a fee. One successful use of mechanical turk is to provide on demand translation services. This highlights the importance of human computation as an alternative to pure artificial intelligence since completely automated approaches for natural language translation have not been successful to be of practical use todate. The anticipation is that the advancement of artificial intelligence would not reach the maturity to comprehensively pass the Turing test in the near future. Hence the field of human computation still remains an important research area, primarily as a practical alternative to solve hard computational problems.