Searle's argument against AI and emergent properties--part 2

If b-wetness is not the same as wetness then what use is this? The point is that b-wetness is an emergent property of a physical system, just as wetness is. There is a difference between the properties of wetness and b-wetness -- different programs are needed to detect them -- but there is no profound difference in the sort of property that is being detected: they are both physical properties of a system that can be found by a physical process detector. Someone might say, "Wrong! All cases of wetness involve the same general physical processes while b-wetness might involve many different types of physical processes." I would say this is wrong. Wetness involves a certain general classification of physical process -- that described by the wetness detection program. B-wetness involves a general classification of physical process -- that described by the b-wetness detection program. We might see differences between systems in which b-wetness is detected -- for example, the difference between a glass of water and the behaviour of electrons in a computer, but we can also see differences between systems in which wetness is detected -- for example, the difference between the movement of water molecules and the movement of ethanol molecules. There is no respect in which one type of property is "physical" while the other is not. If you can say that wetness results when "basically the same sort of physical processes are occurring in a system" then I can say b-wetness also results when basically the same sorts of physical processes are occurring in a system. The processes might intuitively appear profoundly different but that is our bias: the processes producing b-wetness in a glass of water and producing b-wetness in a computer simulation of water have a lot in common, that commonality being described by the b-wetness detection program.

To this, someone could reply that "I don't get it" and that the processes involved in wetness are really the same, while the processes involved in b-wetness are different, but any such distinction is artificial. The wetness detection program would detect wetness in a large number of different systems, such as the same glass of water at different times or places, different glasses of water, glasses or bottles of different substances, fluids spilled on tables, etc. The behaviour of matter that produces wetness has a particular description which applies to a set of different systems. The behaviour of matter that produces b-wetness also has a particular description which applies to a set of different systems. The set of systems which have b-wetness will be much larger than the set of systems which have wetness, but that is just a matter of degree. The scope of b-wetness's definition is clearly much wider that that of wetness. We might argue that b-wetness is somehow more abstract than wetness, or more removed from the basic physical world than wetness, but this does not in itself mean that b-wetness is not an emergent property of a system. If being applicable to lots of systems means that something is not an emergent property then it could just as easily be argued that wetness is not an emergent property because it is claimed to be possessed by completely different physical systems. The formal description of any property will have a kind of "bandwidth" of systems to which it applies and it is just more obvious for b-wetness.

Someone might insist that I do not understand that wetness obviously involves the same physical processes while b-wetness is an abstraction. In reply I would suggest that we arrange all the possible detection programs in a row. On the left-hand side we have the programs that have "narrow bandwidth" -- the ones that are very specific about what criteria have to be met for a property to be detected. The wetness detection program will not be on the extreme left because it has to have at least sufficient bandwidth to detect wetness in different glasses of water. The detection programs on the left-hand side will be much more specific than this -- only detecting the property in a singular situation. As we go to the right the definitions of properties become more general, the corresponding detection programs detecting properties in larger sets of situations. As we keep going to the right, eventually we come to the detection program for the wetness detection property and if we keep going we encounter programs more general than the wetness detection program, eventually encountering the b-wetness detection program. As we go further we encounter still more general programs. The point is that as we go from left to right there is no point at which things suddenly change and at which the detection programs stop detecting things that are not emergent properties of a physical system. Wetness and b-wetness are just different emergent properties with different levels of generality. If you disagree with this my challenge to you is to tell me exactly at what point, as we go from left to right, the properties being detected suddenly become non-physical and "abstract," "artificial" or "made-up."

"You don't get it do you? Computers and brains do not involve the same processes."

Some readers may still think it obvious that a glass of water and a glass of ethanol involve the same general kind of process -- movement of molecules -- while a glass of real water and a computer simulation of water involve different processes -- the movement of water molecules in a glass and the movement of electrons, in a different way, inside the computer. I am not doing something as naïve here as deluding myself by making some crude mechanical analogy between the movement of electrons and the movement of water molecules. The only thing that justifies us in saying that a glass of water and a computer simulation of water involve the same physical processes is that the description of those processes could be expressed as a physical process detection program and used to detect b-wetness both in the glass of water and in the behaviour of matter in the computer. If this seems contrived it is for this reason:

There is a one-to-one correspondence between what is happening in a glass of water and a glass of ethanol. There is a simple one-to-one relationship between elements of the processes in the glass of water and the glass of ethanol: a single ethanol molecule is equivalent to a single water molecule, even if the behaviour is not exactly the same. The general description of wetness can be expressed as "X does various things" (let us assume those things are stated) and to get the description of what is going on in the glass of water we substitute "water molecules" for X and in the case of the glass of ethanol we substitute "ethanol molecules" for X. This one-to-one correspondence does not exist for b-wetness. Instead the general description of the b-wetness, expressed as a physical process detection program would have to work in a more complicated way to "find" b-wetness in both a glass of water and a computer simulating water. There is, however, nothing special about such one-to-one correspondence: it merely relates the description of the processes particularly directly to the system. It may seem "obvious" to us that physical processes defined so that there is such one-to-one correspondence to systems are "physical" because we are able to understand such processes easily, but to an entity much more intelligent than us much more abstract processes, in which the relationship between the description of the process and the elements of the system is very abstract, may appear trivially simple and "obvious."

Next page >