Pedro Domingos

Would love to hear your anticipations about what (concrete) research directions might get the next Turing Award.

The future of AGI is event-based neuromorphic computing and spiking neurons. The brain processes changes aka precisely timed data transitions. It's all about timing. The brain is a massive timing mechanism.

Pedro are you aware of the "recent" one thousand brain theory? Do you think this can be usefull to develop better AI systems ?

Jeff Hawkins should call it the "million-brain theory of intelligence" since he believes that every cortical column is a tiny little brain. He's incorrect, IMO. A cortical column is a tiny stimulus processor that is a small part of a linked community among many other communities.

Bengio is correct but the author of the book he is recommending, Gary Marcus, is not calling for the demise of deep learning. He believes that the correct approach to AGI is to build a hybrid of symbolic AI and deep learning. Crazy stuff.

I am literally about to deliver a seminar at Swinburne University where the proposition is that AGI will only be achieved when we stop using computers. Computers are the actual problem. Should be fun!

What would you propose as the medium of implementation? Would any alternative not end up being a 'computer'?