Trusted AI

Artificial intelligence is used in an immense number of services that we use every day. A subdivision of AI are Conversational Agents, these are self learning programs that are able to communicate with humans. Current Conversational Agents like Alexa or Siri can assist with simple tasks, but more advanced assistants are being developed. Before being introduced to the public their trustworthiness should be addressed.

This futuristic, advanced AI is called general AI and can perform intellectual tasks like any human. the question is whether these systems can be trusted. Research presents different frameworks explaining that AI needs to become trustworthy. We concluded that these have not been applied to generalist AI. We synthesized these frameworks and applied them on our own CA, ‘Charlæ’, to explore their application. We researched how a synthesized framework for trust should be implemented in an advanced Conversational Agent.