Loading...
-
Will not come by returning to the past
Failed paradigms like expert systems and semantic networks required putting in information one fact at a time and could not scale to practical applications. Symbolic AI did not work — So not really AI.
-
Unlikely to be achieved by Today's AI
Based on a recent survey, and extrapolating on today's AI technology, the consensus of today's AI experts is that AGI is decades away. The average guess for a 50 percent probability was 80 years in the future, This is consistent with the way Deep Learning experts talk about the challenges of AGI. They appreciate that AGI must have models of the world and they admit they don't know how to achieve it using artificial neural networks.
-
Will not come by combining past failed paradigms with today's limited approaches.
Neural-Symbolic AI is a recent trend give impetus by DRAPA's explainable AI project, it is based on trying to combine past and current AI techniques in hopes their limitations will cancel out.