4 ms·

I don't understand what point you're making exactly. What categories do you mean? What do you mean by segmentation not necessarily being "an inside-out process

• mallowdram 1y ago
  The criteria for learning in this model has nothing to do with biological intelligence. Segmentation is a cog-sci hold-over of vision science and Marr and isn't how brain's perceive scenes/objects/events. There is a relatovely new approach to perception that ML has ignored that's integrative, coordinated, holistic. These new approaches, affective, coordinated-dynamical, ecological (optic-flow) are the likely routes to consciousness. What ML does with images like these are retrofit, they're imposed ad hoc on imagery as a pretend form of intelligence.
  • mallowdram 1y ago
    Inside-Out is a reversal of the stimuli model and a reversal of the cog-sci cognition model. https://academic.oup.com/book/35081 https://academic.oup.com/book/35081 The senses can't be excluded from consciousness or intelligence, otherwise the notion of intelligence is reduced from an arbitrary set of tests/criteria. Robotics and trained analogies, arbitrary ideas of affordance (which are not affordances) are definitely interesting, but they're not paths to intel. They're paths to homogenization posing as intelligence. This is the classic robotics idea of computer vision backing itself into a corner. https://docs.google.com/presentation/d/1Wkno8pKzWiav1a7c8IOrAlrr_AWUafn1_3aK1sLhK2k/edit?usp=drive_link https://docs.google.com/presentation/d/1Wkno8pKzWiav1a7c8IOr...