@CrazyMaharajah said in #37:
> or any resemblance thereto
Sorry to say, but they have the architecture of the connectivity of the visual cortex. It is based on connexionist view of the brain basic archictecture.. Hubel and Wiesel model of the visual cortext. (Cats have been serving science in that respect, experimentally, for their high visual sense of the world, and their size).
The spatialisation of all sensory input is quite ubiquitous, even sound has to be encoded spatial (across a layer of neurons) in the inner ear before the brain can make sense of it.
I think the neural network is not a misnomer in that sense.
The basic non-linearity between input and output is another hallmark of the wet neurons.
I grant you, though that the current silica implementations are wasteful of the instant spatial processing layer-wise (wide) that a full network processing step would have, instead of being ground one CPU register at a time over each CPU/GPU bandwidth chunks of all the weights.
It is not as much a misnomer as intelligence in AI.. but intelligence itself is not well defined either.
I think in any way, without words assigned: It does not feel like a human speaking, or not a very sincere one, or one grounded in the sensory reality.
> or any resemblance thereto
Sorry to say, but they have the architecture of the connectivity of the visual cortex. It is based on connexionist view of the brain basic archictecture.. Hubel and Wiesel model of the visual cortext. (Cats have been serving science in that respect, experimentally, for their high visual sense of the world, and their size).
The spatialisation of all sensory input is quite ubiquitous, even sound has to be encoded spatial (across a layer of neurons) in the inner ear before the brain can make sense of it.
I think the neural network is not a misnomer in that sense.
The basic non-linearity between input and output is another hallmark of the wet neurons.
I grant you, though that the current silica implementations are wasteful of the instant spatial processing layer-wise (wide) that a full network processing step would have, instead of being ground one CPU register at a time over each CPU/GPU bandwidth chunks of all the weights.
It is not as much a misnomer as intelligence in AI.. but intelligence itself is not well defined either.
I think in any way, without words assigned: It does not feel like a human speaking, or not a very sincere one, or one grounded in the sensory reality.