Artificial Intelligence Today and Tomorrow
Published on 18 August 15
Mike0
Is it possible for people to trust and collaborate with the intelligent machines and systems that are becoming closer to our workplaces and homes?
Let’s think about the emerging intelligent systems that are round the corner. How are we going to interact with them? Is it possible for us to become partners? Or maybe we are going to be regulated of the role of observing their output and doing what they tell us to do?
The main thing people are looking for in the cooperation with the unknown world of so-called black screens is transparency. Steve Lohr, one of the authors at New York Times, has written that there is an approach to the transparency issue, an initiative that uses visual images to display the reasoning. However, the simple answers given by these screens do not inspire trust. Danny Hillis, an A.I. expert, said in the same article that the key point in our potential human-machine cooperation is storytelling.
This word appears in the conversations about the integration of the learning machine as well as intelligent systems and their presence in our data centers. But what does it really mean when it comes to communication between human and software?
Some systems provide the results in the form of answers. They provide correct answers but never give any explanations. The answer without a story is enough for a machine, but is not enough for a human. For example, Watson Paths can tell you that you have a certain disease and provide you with a numerical percentage to prove the strength of this belief, but it is not enough for you. The reasoning is useful, but still every answer needs interpretation or skills to navigate a decision model.
The better approach is examining the information related to what Watson has found, where did it look, and the nature of the things found. This data should be interpreted and turned into an explanation of how Watson came to its conclusion.
It would be much better if a user could have access to a so-called narrative rather than an answer or a graphic. The narrative should be written in accordance with the data associated with a process that led to a particular result. The reasoning alone is necessary, but it also needs some explanations in the way that makes sense to a user trying to understand and evaluate it.
The narrative should answer the questions that any person would ask when evaluating the answer:
- -· Where did you find the information?
- -· What were you looking for?
- -· Are there any other variants or answers you have found? If yes, why did you choose this one?
- -· Are there some information gaps that could be filled to confirm the answer?
Indeed, the storytelling is an approach to be used in the cooperation between people and intelligent machines, because no one would communicate with a person that just gives dry answers to the questions and walks away. Why would we expect that people could work with machines that do the same?
This blog is listed under
Development & Implementations
Community
Related Posts:
Post a Comment
