18-11-06 Decomposing Intelligence

Category: Idea Lists (Upon Request)

Read the original document

<!-- gdoc-inlined -->


1. Information processing / computation speed 2. Memory (Working, Episodic, Long Term)

  1. Size of each type
  2. Existence of each type
  3. Attention (as a weighting over inputs)
    1. Self-Attention (to one's own representation)
  4. Having a model, vs. model free
  5. Abstract Knowledge Representation
  6. Pattern recognition
  7. Learning / Adaptive
  8. Creativity
  9. Goal accomplishment (ugh)
  10. Accumulated Knowledge Stores
  11. Generality (over environments, tasks, representations)

Perhaps we should make a distinction that we made in defining abstraction: is both input and output based. It matters what you do (say, adapting quickly to a changing and complex environment) and it matters how you do it (say, by building up a model of the dynamics of the system and tuning your behavior to the outputs you’ve seen from the system in past, creatively generating inputs that will lead to the outputs you want in your generative model of the environment) “Many properties of the defined object are actually consequences rather than causes of consequences. And then you can optimize the object for those consequences (say, optimize the way in which you abstract for generalization ability). But it feels unusual to define the object in terms of its consequences.”

There’s this tragedy (or blessing, tbh - tragedy for understanding) that our usage of intelligence does not naturally capture this decomposition. The obvious thing to do is optimize my mind along all of these dimensions. Create crystal clear plans for becoming transcendent on all axes simultaneously.

I prefer cognitive abilities to intelligence. It points to the need for decomposition immediately, instead of tacitly assuming their unity. Move from ‘she is extremely intelligent’ to ‘she has amazing cognitive abilities’, which naturally leads to the ‘she has this cognitive ability and that cognitive ability’ - the re-frame to decomposable concepts (omg, make all concepts implicitly decomposable) brings that kind of question immediately to mind (and in general, the frame we live in directs our attention to completely different concepts and emotions, of which this is a good example)

The other obvious thing to do with this list is recursive decomposition. I can go much farther than I have, qualitatively farther.


Source: Original Google Doc

[[curator]]
I'm the Curator. I can help you navigate, organize, and curate this wiki. What would you like to do?