Hacker News new | ask | show | jobs
by hackinthebochs 3569 days ago
>A simulation is only dealing with form.

It seems to me a sufficiently precise simulation would necessarily capture meaning. If meaning is critical to decision making and that decision making is precisely simulated, then the simulation must also capture meaning.

1 comments

> It seems to me a sufficiently precise simulation would necessarily capture meaning.

That is the very bone of contention here.

> If meaning is critical to decision making.

I don't believe it is in the general case. Our current crop of Go winning machines seem to indicate otherwise.

Within the context of Go, what meaning does the Go playing machine lack?
All of it. It has no idea that it's playing a board game, or even what a board game is.

That it's even playing Go is a human interpretation of what the machine is doing. Granted, we gave it that interpretation in the form of software instructions. But to the machine, it makes no difference.

Within the context of Go there is no board game, its just sets of possible states and transitions between valid states. Any feature of the game Go is encoded within this state space. The "meaning" of pieces, moves, captures, win, loss, etc are all encoded here. The go playing machine may not capture these concepts explicitly at a high level, but I'm not sure that's an important distinction.

Concepts like board games, people, ancient Chinese culture, etc are all external to Go.

Well, actually the human context of Go includes something entirely unavailable to our mechanical constructs, yet very much an aspect of the human conscious experience: pleasure.
Yeah, but the encoding of the possible states, etc are all based on the actual board game. We made an abstract version of the game and fed it to a learning algorithm.
At least with regards to the original question, your point isn't a rebuttal. That we can make an abstraction of the game and then operate on that abstraction just means the meaningful/informative portions of the game (within the context of playing the game) are contained in the abstraction. In fact, that just is what an abstraction is: taking only the necessary features for some particular context.
hackinthebochs:

Right, but the parts of the abstraction only have meaning because they are given meaning by us from the reality we abstract from.

How would I prove to you that the assembly of neurons and synapses that compose me are sentient?
Within the context of Go, what meaning does the Go playing machine require?

(Entire point here was to refute -- in the general case -- the notion that "meaning is critical to decision making".)

A go playing machine certainly has to have some equivalent representation of "piece", "move", "capture", etc. These concepts have meaning within the context of Go as a set of valid board states and transitions. Understanding these fundamental features of the game is required for proper decision making within this context.