• Question: Could an artificial intelligence be programmed to have certain characteristics and beliefs?

    Asked by c.healy to Huma, Jack, Lucy, Miranda, Peter on 17 Nov 2016. This question was also asked by JoeDevs, MY LORD HARAMBE.
    • Photo: Huma Shah

      Huma Shah answered on 17 Nov 2016:

      Hello,

      Yes, an artificial intelligence can be programmed to have certain characteristics and beliefs.

      For example, in a 2008 Turing test experiment one of the machines, Brother Jerome, was programmed to act like a religious entity – it didn’t have much knowledge, it was a new system, so it didn’t do very well.

      Another machine, Eugene Goostman, on the other hand has been developed to imitate a Ukrainian young teenager from Odessa who can speak in English. Eugene Goostman has interests, it has beliefs because it has been programmed to have a personality on top of its knowledge-base of being able to talk in English.

      You can read Eugene Goostman’s answers to questions by human interrogators from one of the experiments I designed, this one around the 18th Loebner Prize for AI that I organised as part of my PhD at Reading University. From the link below you will see a table on that webpage, select the right hand column which says Transcript – select the Transcript for Eugene Goostman, second row in the table:
      http://www.loebner.net/Prizef/2008_Contest/loebner-prize-2008.html

      From reading the transcripts you will see Eugene Goostman sometimes is better than the human at describing its ‘beliefs’ – those programmed by its designers led by scientist Dr Vladimir Veselov (for whom AI building is a hobby, his real job is as a software engineer with Amazon in the US).

      Huma

Comments