Re:Objects of consciousness
by Animats on Saturday July 04, 2015 @02:02PM
(#9830960)
Same here. As soon as I read "shifting from evolutionary biology to quantum physics", I knew there was going to be a problem.
The line "Perception guides adaptive behavior; it does not estimate a preexisting physical truth." is reasonable enough. Evolution may have generated visual recognizers that are task-oriented rather than object oriented ("predator", rather than "fur in bushes"). It's more important to have recognition of likely predators from partial information than correct recognition of a range of animals and other moving objects. That's a useful idea, and suggests that training deep ANNs on predator/non-predator data sets, without trying to distinguish the predators, is a worthwhile line of research. It would be interesting to see what visual model of "predator" emerged.
The "Markovian kernel" concept is not too useful. It's a table of probabilities with one row for each state of the world (which, of course, would be combinatorially huge) and a column for each possible consciousness state. This might be conceptually similar to Searle's "Chinese Room", where understanding is faked by pure lookup. But Searle and these guys go off in different directions.
The line "Perception guides adaptive behavior; it does not estimate a preexisting physical truth." is reasonable enough. Evolution may have generated visual recognizers that are task-oriented rather than object oriented ("predator", rather than "fur in bushes"). It's more important to have recognition of likely predators from partial information than correct recognition of a range of animals and other moving objects. That's a useful idea, and suggests that training deep ANNs on predator/non-predator data sets, without trying to distinguish the predators, is a worthwhile line of research. It would be interesting to see what visual model of "predator" emerged.
The "Markovian kernel" concept is not too useful. It's a table of probabilities with one row for each state of the world (which, of course, would be combinatorially huge) and a column for each possible consciousness state. This might be conceptually similar to Searle's "Chinese Room", where understanding is faked by pure lookup. But Searle and these guys go off in different directions.
The paper goes on, but not to anywhere useful.