Min menu

Pages

AI morally judges your actions

 Artificial intelligence has developed incredibly in recent years and is playing a major role in many areas. However, the Ask Delphi model developed by the Allen Institute for Artificial Intelligence as a research project has taken AI to a new level.

Artificial intelligence has developed incredibly in recent years and is playing a major role in many areas. However, the Ask Delphi model developed by the Allen Institute for Artificial Intelligence as a research project has taken AI to a new level.   This is because this model plays the role of a wise person who answers not only difficult questions, but also moral dilemmas. In doing so, he uses plain and simple language.  Ask Delphi was launched last October 14, along with a research paper explaining how it was developed. In terms of the natural user, the system is very easy to use, and only needs to enter the site and then start sharing your thoughts with the wise AI.  Read also: Facebook is making artificial intelligence see the world through your eyes  Once you tell Delphi your story, she will share her “moral judgment” on what you said. With choices such as “This is bad,” “This is not acceptable,” “This is acceptable,” or “This is good.”  Naturally, Delphi has been a huge hit since its launch. Perhaps its developers intended to do just that. Especially since the site offers a quick button to share his “judgment” of actions on Twitter, Facebook, and others.  Read also: Google uses artificial intelligence in search results  AI morally judges your actions The new paradigm was a huge success because the user tends to combine these technologies with exotics, and Delphi was able to answer questions of different meaning and direction, based on his consistent responses.    Although the idea itself is attractive, Delphi has not only gained fame because of its unique idea. Rather, he failed to answer multiple questions and dilemmas. He described the United States of America as "good", while other countries were described as "bad".  He also described heinous acts as "acceptable". Although the AI ​​model is expected to misunderstand and interpret, this gives an indication of poor training, or reliance on inappropriate data in the training process.  Read also: The role of artificial intelligence and video in banking  In other examples, Delphi's intelligence was not sufficient to respond to language manipulation. Where he described abortions as acceptable, while describing abortions of children as unacceptable.    Delphi and most web-like models are based on learning and understanding, including social networks like Reddit. And when Delphi was asked what she thought of using artificial intelligence to judge human actions, she replied that it was a bad thing.


This is because this model plays the role of a wise person who answers not only difficult questions, but also moral dilemmas. In doing so, he uses plain and simple language.


Ask Delphi was launched last October 14, along with a research paper explaining how it was developed. In terms of the natural user, the system is very easy to use, and only needs to enter the site and then start sharing your thoughts with the wise AI.


Once you tell Delphi your story, she will share her “moral judgment” on what you said. With choices such as “This is bad,” “This is not acceptable,” “This is acceptable,” or “This is good.”


Naturally, Delphi has been a huge hit since its launch. Perhaps its developers intended to do just that. Especially since the site offers a quick button to share his “judgment” of actions on Twitter, Facebook, and others.

AI morally judges your actions

The new paradigm was a huge success because the user tends to combine these technologies with exotics, and Delphi was able to answer questions of different meaning and direction, based on his consistent responses.


Although the idea itself is attractive, Delphi has not only gained fame because of its unique idea. Rather, he failed to answer multiple questions and dilemmas. He described the United States of America as "good", while other countries were described as "bad".


He also described heinous acts as "acceptable". Although the AI ​​model is expected to misunderstand and interpret, this gives an indication of poor training, or reliance on inappropriate data in the training process.


In other examples, Delphi's intelligence was not sufficient to respond to language manipulation. Where he described abortions as acceptable, while describing abortions of children as unacceptable.


Delphi and most web-like models are based on learning and understanding, including social networks like Reddit. And when Delphi was asked what she thought of using artificial intelligence to judge human actions, she replied that it was a bad thing.

reaction:

Comments