AI refers to simulating human intelligence processes by machines, which demands learning from constantly changing data, reasoning to make sense of data and self-mechanisms to make decisions.
Since human intelligence is rooted in sensing & learning from the environment, consequently full AI requires simulating human senses (sight, hearing, smell, taste, touch), simulate learning and processing of data (machine and deep learning) and simulation of human responses (robotics).
In this keynote, Yoshua Bengio summarizes three key points when ‘looking forward’ to what "deep learning cognition" requires:
- a world model which meta-learns causal effects in abstract space of causal variables
- necessity to acquire knowledge and encourage exploratory behavior
- bridge the gap between system 1 and system 2 ways of thinking, with old neural networks and consciousness reasoning
