Brilliant conference so far, and it’s only the first day (*just* a workshop session). We’ve had four ‘tutorials’ today:
Hutter described the AIXI model, which is a theoretically best case version of a goal-driven intelligent system.
Looks discussed program learning and ways in which this can be implemented. Program learning involves a system ‘discovering’ a program via altering its own code.
Goertzel talked about developments of AI agents interacting with virtual environments versus robots interacing with real world environments and discussed the advantages and disadvantages of both appraoches.
Talked about whether or not we can find correlations between some of the concepts encountered in AI/AGI an some of the neuronal mechanisms that are occurring in the brain. The specific concept under scrutiny was reinforcement learning. Koene demonstrated that collections of neurons can be modelled by cortical minicolumns and by connecting thes together in large groups, they simulate the same behaviour as is seen in trials undertaken by monkeys given a reward / no reward mechanism for recongnising certain images.
Another thing I love is the lecture theatre/conference venue (The USI). There are power sockets at every seat and there is free, easy-to-connect wifi. I don’t think I’ve been to a single conference where both those criteria have been fulfilled before. Lunch is provided on-site, and coffee and snack breaks are scheduled often. Those are really the only things I ask in terms of the conference venue to be honest. Power, wifi, coffee, lunch.
Yesterday I had a great meal at one of the local Lugano pizzaria restaurants, and talked about AGI and what it actually means.
Anyway, more soon, it’s buffet-networking time now…