Artificial Intelligence in Dire Need of Better Brain Architectures?
Randal Koene - Whole Brain Emulation from Raj Dye on Vimeo.
The above is a video from a conference on artificial general intelligence (AGI) held in Switzerland, last year. The speaker is a neuroscientist -- an outsider to the typical AI person who attends AI conferences. His appearance at the AGI conference indicates that the entire approach to AI is in a state of flux.The attitude up until recently has been that intelligence does not rely upon any particular substrate, eg, a brain. AI researchers have boldly believed for several decades that intelligence could be built algorithmically inside machine architectures over a relatively short time span. "Sometime within 10 years . . ."
They have been saying the same thing -- "within 10 years" -- since the 1950s. Clearly not very much has happened in the way of significant breakthroughs since the 1950s. In fact, contemporary AI researchers themselves may well be growing less impressive, over time, than the pioneers of the field.
Hence the perceived need for possibly re-thinking the whole "substrate" approach. Another video in the series deals with the requirements of "cognitive architecture." An impressive phrase, although the reality is likely to prove far less impressive.
Labels: artificial intelligence, brains