RobinZ comments on Diseased thinking: dissolving questions about disease - Less Wrong

236 Post author: Yvain 30 May 2010 09:16PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (343)

You are viewing a single comment's thread. Show more comments above.

Comment author: RobinZ 09 June 2010 12:15:22PM 2 points [-]

In a deterministic universe, the future is logically implied by the present - but you're in the present. The future isn't fated - if, counterfactually, you did something else, then the laws of physics would imply very different events as a consequence - and it isn't predictable - even ignoring computational limits, if you make any error, even on an unmeasurable level, in guessing the current state, your prediction will quickly diverge from reality - it's just logically consistent.

Comment author: Ganapati 12 June 2010 05:52:00AM *  0 points [-]

if, counterfactually, you did something else, ...

How could it happen? Each component of the system is programmed to react in a predetermined way to the inputs it receives from the rest of the system. The the inputs are predetermined as is the processing algorithm. How can you or I do anything that we have not been preprogrammed to do?

Consdier an isolated system with no biological agents involved. It may contain preprogrammed computers. Would you or would you not expect the future evolution of the system to be completely determined. If you would expect its future to be completely determined, why would things change when the system, such as ours, contains biological agents? If you do not expect the future of the system to be completely determined, why not?

Comment author: RobinZ 12 June 2010 01:49:53PM 1 point [-]

I said "counterfactual". Let me use an archetypal example of a free-will hypothetical and query your response:

Suppose that there are two worlds, A and A', which are at a certain time indistinguishable in every measurable way. They differ, however, and differ most strongly in the nature of a particular person, Alice, who lives in A versus the nature of her analogue in A', whom we shall call Alice' for convenience.

In the two worlds at the time at which A and A' are indistinguishable, Alice and Alice' are entering a restaurant. They are greeted by a server, seated, and given menus, and the attention of both Alice and Alice' rapidly settles upon two items: the fettucini alfredo and the eggplant parmesan. As it happens, the previously-indistinguishable differences between Alice and Alice' are such that Alice orders fettucini alfredo and Alice' orders eggplant parmesan.

What dishes will Alice and Alice' receive?

I'm off to the market, now - I'll post the followup in a moment.

Comment author: RobinZ 12 June 2010 03:13:06PM 0 points [-]

Now: I imagine most people would say that Alice would receive the fettucini and Alice' the eggplant. I will proceed on this assumption

Now suppose that Alice and Alice' are switched at the moment they entered the restaurant. Neither Alice nor Alice' notice any change. Nobody else notices any change, either. In fact, insofar as anyone in universe A (now containing Alice') and universe A' (now containing Alice) can tell, nothing has happened.

After the switch, Alice' and Alice are seated, open their menus, and pick their orders. What dishes will Alice' and Alice receive?

Comment author: Blueberry 12 June 2010 04:47:42PM 3 points [-]

I'm missing the point of this hypothetical. The situation you described is impossible in a deterministic universe. Since we're assuming A and A' are identical at the beginning, what Alice and Alice' order is determined from that initial state. The divergence has already occurred once the two Alices order different things: why does it matter what the waiter brings them?

I'm not sure exactly how these universes would work: it seems to be a dualistic one. Before the Alices order, A and A' are physically identical, but the Alices have different "souls" that can somehow magically change the physical makeup of the universe in strangely predictable ways. The different nature of Alice and Alice' has changed the way two identical sets of atoms move around.

If this applies to the waiter as well, we can't predict what he'll decide to bring Alice: for all we know he may turn into a leopard, because that's his nature.

Comment author: RobinZ 12 June 2010 05:02:30PM 0 points [-]

The requirement is not that there is no divergence, but that the divergence is small enough that no-one could notice the difference. Sure, if a superintelligent AI did a molecular-level scan five minutes before the hypothetical started it would be able to tell that there was a switch, but no such being was there.

And the point of the hypothetical is that the question "what if, counterfactually, Alice ordered the eggplant?" is meaningful - it corresponds to physically switching the molecular formation of Alice with that of Alice' at the appropriate moment.

Comment author: Blueberry 12 June 2010 05:15:34PM 1 point [-]

I understand now. Sorry; that wasn't clear from the earlier post.

This seems like an intuition pump. You're assuming there is a way to switch the molecular formation of Alice's brain to make her order one dish, instead of another, but not cause any other changes in her. This seems unlikely to me. Messing with her brain like that may cause all kinds of changes we don't know about, to the point where the new person seems totally different (after all, the kind of person Alice was didn't order eggplant). While it's intuitively pleasing to think that there's a switch in her brain we can flip to change just that one thing, the hypothetical is begging the question by assuming so.

Also, suppose I ask "what if Alice ordered the linguine?" Since there are many ways to switch her brain with another brain such that the resulting entity will order the linguine, how do you decide which one to use in determining the meaning of the question?

Comment author: RobinZ 12 June 2010 07:30:17PM 3 points [-]

I understand now. Sorry; that wasn't clear from the earlier post.

I know - I didn't phrase it very well.

Messing with her brain like that may cause all kinds of changes we don't know about, to the point where the new person seems totally different (after all, the kind of person Alice was didn't order eggplant). While it's intuitively pleasing to think that there's a switch in her brain we can flip to change just that one thing, the hypothetical is begging the question by assuming so.

Yes, yes it is.

Also, suppose I ask "what if Alice ordered the linguine?" Since there are many ways to switch her brain with another brain such that the resulting entity will order the linguine, how do you decide which one to use in determining the meaning of the question?

I'm not sure. My instinct is to try to minimize the amount the universes differ (maybe taking some sort of sample weighted by a decreasing function of the magnitude of the change), but I don't have a coherent philosophy built around the construction of counterfactuals. My only point is that determinism doesn't make counterfactuals automatically meaningless.

Comment author: Ganapati 13 June 2010 06:28:09AM -2 points [-]

The elaborate hypothetical is the equivalent of saying what if the programming of Alice had been altered in the minor way, that nobody notices, to order eggplant parmesan instead of fettucini alfredo which her earlier programming would have made her to order? Since there is no agent external to the world that can do it, there is no possibility of that happening. Or it could mean that any minor changes from the predetermined program are possible in a deterministic universe as long as nobody notices them, which would imply an incompletely determined universe.

Comment author: RobinZ 13 June 2010 11:46:50AM 3 points [-]

...

Ganapati, the counterfactual does not happen. That's what "counterfactual" means - something which is contrary to fact.

However, the laws of nature in a deterministic universe are specified well enough to calculate the future from the present, and therefore should be specified well enough to calculate the future* from some modified present*, even if no such present* occurs. The answer to "what would happen if I added a glider here to this frame of a Conway's Life game?" has a defined answer, even though no such glider will be present in the original world.