This kindcof problem has just come up again. How do we make the most out of a conversation? In specific contexts. With defined goals? With people, or combined with intelligent agents? Note here the inclusion of 'safety critical', so also an element of defined risk.
U of T Experts in AI Explore a Classical Problem of Computer Science
U of T News By Nina Haikara
Researchers at the University of Toronto (U of T) in Canada are exploring synergies between fast, effective algorithms in artificial intelligence for automated planning and program synthesis, or generating a computer program automatically from a specification of user intent. U of T's Alberto Camacho says he used linear temporal logic (LTL) to develop a practical tool for synthesis, expressing the user's intent in instructions similar to English-language instructions. "I can say something to you, and if you misunderstand, we can have a conversation back and forth [to clarify]," says U of T's Sheila McIlraith. "But when we care about synthesizing safety-critical systems...it really matters that the system understands what we're asking it to do." Camacho says the team is applying automated planning algorithms to LTL synthesis. The program that is produced is correct by construction, and every run of that program satisfies the LTL statements with which it is provided. ... "
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment