Mini thread about the cognitive science and neuroscience inspirations behind our new paper in which we learn concepts as 'cognitive programs' on a 'visual cognitive computer'. robotics.sciencemag.org/content/4/26/e… 
Cognitive programs is inspired by Barsalou's Perceptual Symbol Systems (PSS) theory, which is also relevant for discussions on symbolic-connectionist integration. @GaryMarcus citeseerx.ist.psu.edu/viewdoc/downlo…
PSS posits that concepts are learned as simulations in a perceptual system that is 'componential' (as opposed to holistic) and steerable by top-down attention. RCN, our generative vision model, is built to match the requirement specs: science.sciencemag.org/content/358/63…
Ullman's 'visual routines' is a seminal paper on learning concepts. Ullman pointed out the elementary routines need to be part of a visual processor, which inspired the instruction set of our 'visual cognitive computer'. courses.csail.mit.edu/6.803/pdf/ullm…
Over the years, several others expanded on these ideas. One example is the paper 'The Human Turing Machine' by Zylberberg, @StanDehaene, @Pieters_Tweet and Sigman ncbi.nlm.nih.gov/pubmed/21696998
A few other papers from @Pieters_Tweet were very informational. One is about using primary visual cortex as a cognitive blackboard, similar in spirit to the global workspace idea from @mpshanahan
Mandler's "how to build a baby" series. Image schemas for enabling pre-verbal thought and as a foundation for language.
pdfs.semanticscholar.org/0f3d/c8fda72b5…
pdfs.semanticscholar.org/0f3d/c8fda72b5…
For program induction, we used a combination of generative models (for program space) and discriminative models (input-conditioned predictions of arguments), and used the Explore-Compress framework.
Similar to Bayesian Program Learning from @LakeBrenden, our method learns to learn by using the already learned programs to guide the search for new programs. web.mit.edu/cocosci/Papers…
Finally, we borrowed the 'cognitive programs' name from Tsotsos. Note that this paper points out the important role of top-down attention. ncbi.nlm.nih.gov/pmc/articles/P…
