Instituut voor Taal- en Kennistechnologie
Institute for Language Technology and Artificial Intelligence

Harnad responds

It's difficult to determine whether Bringsjord's is another friendly commentary, or he's really changing the subject. I don't think we differ about connectionism. In my model, neural nets play a circumscribed role in a hybrid system. The hybridism is analog/symbolic, not connectionist/symbolic, however, and as I wrote, I don't much care whether neural nets are like room one (PAR) or room two (SIM) in my three-room argument, if they can do the job (TTT-scale categorization). The analog transducer projection is not optional, however, and can no more be replaced by a digital approximation than the world can. So much for discrete approximations.

Bringsjord can imagine cerebration without transduction. I don't see how that establishes its physical possibility (and that's why I think he's talking about another subject). I have a similar reaction to his ``argument from serendipity'': So chimpanzees might write Shakespeare by chance: What light does that cast on how Shakespeare wrote Shakespeare? And of course hypotheses conditional on passing the TT or TTT, by whatever means, do not yet have any true antecedents. How one is to go about finding a true antecedent is the name of the game here (my game, at any rate)! In this game, whatever it turns out to take to pass the TT, the TTT, or the TTTT exhausts all the empirical possibilities. Pure chance is not one of the possibilities worth considering.

It is, I think, easy to see why, if the TTTT -- requiring a system whose every observable property, behavioral and neuromolecular, is totally indistinguishable from those of our brains -- fails to capture mental properties, then we can certainly not hope to be any the wiser. It is only a bit more difficult to see that, although it too may be fallible, the TTT, rather than the TTTT, is all we have to go on in judging whether anyone else has a mind (since no one to date, amateur or professional, has been near expert enough about his neighbor's brain to base much on that -- and I assume this applies also to Selmer and Elizabeth). The only vexed question is whether the Blind-Watchmaker had anything stronger to go on. If He did -- if there are TTT-equivalent but TTTT-inequivalent options that somehow differ in their survival value, then we had better turn to the TTTT instead of just the TTT. My own hunch, however (and I can only repeat it), is that the TTT already narrows down the empirical degrees of freedom for success enough so that a mindless option need not be worried about (or not worried about much beyond the normal limits of scientific underdetermination).