Abstract
The concept of intelligent software is flawed. The behaviour of software isdetermined by the hardware that "interprets" it. This undermines claimsregarding the behaviour of theorised, software superintelligence. Here wecharacterise this problem as "computational dualism", where instead of mentaland physical substance, we have software and hardware. We argue that to makeobjective claims regarding performance we must avoid computational dualism. Wepropose a pancomputational alternative wherein every aspect of the environmentis a relation between irreducible states. We formalise systems as behaviour(inputs and outputs), and cognition as embodied, embedded, extended andenactive. The result is cognition formalised as a part of the environment,rather than as a disembodied policy interacting with the environment through aninterpreter. This allows us to make objective claims regarding intelligence,which we argue is the ability to "generalise", identify causes and adapt. Wethen establish objective upper bounds for intelligent behaviour. This suggestsAGI will be safer, but more limited, than theorised.