TED Conversations

Ian Gwonsang Shin

undergrad Aeronautical Sceince, Korea Aerospace Univ.

This conversation is closed.

Could algorithmic approaches control all of the variables in reality?

I've always been interested in mathmatical approaches for solving business problems (Operation Researches, Management Science). Meanwhile, I saw a TED talk named 'How algorithms shape our world'. I partially agree with Kelvin Slavin. Algorithms are useful and effective. But, I want to ask something about it. Are they perfect?

Algorithms are also 'man-made' methods for solving problems or implementing methodology, even predicting the future (proficiency, feasibility of business investment, etc). There are tons of variables in reality and algorithmic apporaches couldn't calculate all the variables I guess. And there are also variables that couln't even be digitized. And algorithmic approaches have flaws in themselves. They could crash each other.

I'm not speciallized in computer engineering, so I'd appreciate any info you might have. Thank you! :)


Showing single comment thread. View the full conversation.

  • Oct 15 2012: I think the game changers for AI will be rationalizing, recognizing context, situational awareness, and applying lessons learned to present situations. It almost requires a machine to have an underlying philosophy of human interaction that is constantly updated based on sensory data collected. It a machine can achieve this, it may learn. It may be coming to one conclusion on the spot, then later get more information and learn the decision was incorrect, then learn to be less confident about a decision or to ask additional questions at a later time. Machines need to also understand the potential of their own bad decisions and bad sensor data, particularly when the decisions being made are radically different than past knowledge and experience. Machines need to develop a threshold for confidence in decisions relative to risk analysis. Saying nothing may be wiser than making the best guess at a given time. Machines need to be able to put situations in the context of nature, for example weather and he biological needs of humans.

    I think there are many areas where just a problem solving algorithm is not enough.

    Watson is headed down the right path. http://en.wikipedia.org/wiki/Watson_%28artificial_intelligence_software%29

Showing single comment thread. View the full conversation.