In https://creativisticphilosophy.wordpress.com/2015/05/16/how-intelligent-can-artificial-intelligence-systems-become/ I have argued that there is a limit to intelligence since a creative system can only process a very limited amount of information at any time without running into a combinatorial explosion. Large amounts of data require pre-existing knowledge on how to process them, i.e. algorithms, but if such knowledge does not exist yet and the system is trying to find out how to process some novel data, only small amounts of information can be processed.
Since our senses produce large amounts of data that tend to contain some novel information, we can only analyze small parts of that information. Automatic processing of the primary sense data using known algorithms (or neuronal networks that can be modelled as algorithms) can only provide a limited depth of analysis. As a result, only a small fraction of the information can be analyzed deeply at any given time, so the phenomenal overflow discussed, for example in https://scientiasalon.wordpress.com/2015/05/18/ned-block-on-phenomenal-consciousness-part-i/ and https://scientiasalon.wordpress.com/2015/05/20/ned-block-on-phenomenal-consciousness-part-ii/ might be inevitable for any such system, natural or artificial. A mechanism of attention might be necessary.