Hacker News new | past | comments | ask | show | jobs | submit login

Ignore the downvotes: there is value in reading and attempting to understand some of this material, despite the inflammatory way the claims are made. Carl Hewitt is presumably in his 70s, so I'll forgive him for being an old man who probably doesn't give a shit what people think.

In particular, I find the discussion of "paradox" attacks to be quite illuminating. Bertrand Russell attempted to remove self-referential and self-applicable paradoxes through the embellishments of types and orders. Hewitt argues that Gödel basically "cheated" in his proofs because the Diagonalization Lemma violated restrictions on orders.

I'm not qualified to evaluate that argument on the basis of mathematical reasoning, but I do believe it points to a mismatch between the way computation is modeled in the abstract and the way it happens in reality. In the purely abstract world of natural numbers, there's really only one kind of thing: the natural number. Statements about natural numbers can be represented by natural numbers, statements about statements about natural numbers can also be represented by natural numbers, etc. While a given natural number can be parsed to compute if it is a valid encoding of a proposition of a certain order, it is always and forever a natural number and nothing else.

However, I'm not entirely convinced that this captures the nature of computation in reality. When computation is embodied in a physical system, is it possible that the output of some computation can itself manifest computational properties that are not mere compositions of the lower level computational model? Where the interactions between the "higher level" objects simply follow a different set of rules and thus form the foundation of a new formalism?

The notion of types seems to be a way to capture this, by introducing the possibility of distinguishing a "thing" from its abstract representation or description. If everything exists only within the world of the abstract representation, then a thing is indeed no different from its description, as any manipulation of the thing is equivalent to some manipulation of its description. But when the thing is, in fact, a physical object, it is clear that construction of the thing can fundamentally alter the model of computation. Why is this?

I suspect that it is because there are no "pure" or "inert" abstractions in the real world; everything is in fact performing computation all the time. So physical constructions are not just compositions of objects, but also compositions of the computational capabilities of those objects. I realize this probably sounds like navel-gazing at this point, but sometimes that can be a useful activity.




Thanks for your interesting ideas!

My main purpose is to educate. So I am pleased that you got something out of the article :-)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: