What I can find from Tom referring to the consciousness (which term the damned search function refuses to search for) of computers is here: viewtopic.php?f=9&t=3535&p=8063&hilit=a ... uter#p8063
He does not there enumerate 4 attributes but states:
A sufficiently complex self modifying computer will spontaneously develop consciousness, but that computer can also be pre-loaded with data defining all or part of some other computer's (one with extremely similar hardware) self developed consciousness. Immediately after that software load, the two computers may have very similar attributes, but begin to diverge into their own uniqueness as they learn and grow in their own ways (approach problems with multiple solutions that can be assessed in multiple ways). Same story with any digital consciousness system.
Perhaps someone remembers where he stated this. Here he describes what is represented within his model of the development of consciousness as AUO and then AUM: A sufficiently complex self modifying computer. That is what we effectively had with the interacting proto IUOCs of Tom's model interacting in more and more complex ways over the RWW. This constitutes a complex self modifying computer. The sufficiency of the complexity would be demonstrated by the fact that consciousness did arise as AUO. Large computer systems would have to be specifically programmed with a program setting up many subroutines with different capabilities that could communicate randomly initially with the ability to self modify that communication based upon results, and results that it would itself have to evaluate, not its human programmers. Normally computers are loaded with programs to do specific things such as simulating large natural systems. This would have to be a special program to simulate the conditions required for consciousness to arise. This would indeed be an interesting experiment. One factor not mentioned would be just how much time would be required for this initially random process to require to evolve to consciousness. Judging by the LCS, it could take more than the lifetimes of the experimenters before just the right functions came together.