Let's pull another one from the file -- Things Eliezer Yudkowsky says that I think are just nutty

[sniffnoy]

The point here being that these are not just things I think are wrong, but things I find so clearly ridiculous that I have to wonder how someone like him could think that. Also, I'm just talking about things he publically endorses and makes a big deal of, not things he happens to have written somewhere that nobody cares about.

There's basically just one, really. One was going to be his ideas about set theory and his insistence that ω₁ is absurd (and so theories that imply it are bad), and I guess also his not quite grasping just how qualitatively different first-order and second-order logic are, but after getting into an argument about it on LW not too long ago I'm no longer convinced this is fully nutty. It's still pretty silly, and the logic thing is kind of dumb, but it's not exactly nutty.

So that leaves just one: Timeless physics. Now, let me be clear on what I am not talking about. I'm not talking about the block universe perspective. I think he does a pretty good job at presenting the eternalist perspective. Saying that things are determined but it's meaningless to say that they're predetermined is certainly not something I would have thought of.

Which is why I don't understand why he has to muddy it up with "timeless physics". I'm going to join the many people who commented on his articles on this in saying: Huh? Why can't you just talk about a block universe? Special relativity already forces us to think in terms of a block universe, and, AIUI, later-developed physics even more so. So we should all be agreed on an eternalist perspective. But I fail to see how this "timeless" perspective yields any advantage over the far more straightforward block universe perspective and I can see quite a few disadvantages. And to be honest, most of what I've seen written about it -- by Yudkowsky or otherwise -- seems like just nonsense.

EDIT: Next paragraph, "linear" changed to "locally linear". I shouldn't so casually exclude CTCs. :)

Now I'll freely admit I'm going just by Yudkowsky and Wikipedia here -- I've never actually read The End of Time. But it doesn't seem like I really need to. I feel like I shouldn't even have to point out how insane this idea is, but since I do...

1. Locally linear ordering. Once you fix a coordinate system, time is locally linearly ordered. Probably like R, though it's still possible it might turn out to be like N (or something else). Point is, if the whole universe just consists of instants satisfying some equations of state, with similarity playing the role of time, there's no reason to expect anything like a linear ordering. Here I am implicitly taking a many-worlds (i.e. state-of-the-multiverse) perspective, as Eliezer does. From the Wikipedia article, it sounds more like Barbour takes more a perspective where a "now" consists of the state of a universe, not a multiverse, suggesting more of a branching tree structure. Doesn't matter. There's no reason to expect it to satisfy that either. (And if you're Eliezer Yudkowsky you're going to take a many-worlds perspective anyway.)
2. C'mon, if you can't even linearly order your time, how the hell are you going to differentiate with respect to it? Not only is time linearly ordered (with a metric, no less), but there are actual nontrivial relations between what happens at one instant and what happens a given amount of time later. What we call, y'know, physics. How are you possibly going to recover that from just equations of state + similarity relations?
3. That metric, by the way? It's important. If all we have is an *ordering*, then we shouldn't be able to tell the difference between different amounts of time, because you can easily take order automorphisms of R that don't preserve the metric. We have two objects connected by a metal rod, one hot, one cold... how are you going to get the heat conduction equation with just a similarity relation? Now you could say this only makes sense if the induced notion of time is continuous, if it's discrete you can just count, but we have good reason to believe time is continuous because of...
4. Lorentz invariance. Do I really need to say any more about this? Hell, that it's a reason to believe time is continuous is a minor point compared to it being a huge blow to timelessness all by itself. How are you going to explain nows being related to each other in such a way under changes of coordinates when you won't admit there is a time coordinate?
5. Let me finally add: To some extent, "timeless physics" is an oxymoron. If you only have equations of state -- if you can't predict what a given starting configuration will actually evolve into -- then to a large extent you don't really have physics.

Now I'll admit that much of the above is an argument from incredulity. But the facts are that the block universe perspective works, and the timeless perspective... might work with a lot more development but there's good reasons to expect it not to? Real numbers don't come from nowhere; you typically need a lot of assumptions to get them to appear. (Unless one of your assumptions is the common assumption "Let's use real numbers", but I'm ignoring such cases.) And that's me being generous; I honestly find it closer to a real denial of the facts rather than to undeveloped and unlikely.

So why does this bug me so much? Here's the thing: Most of what Yudkowsky writes about such things works perfectly well as a description of a block universe. He explains it really well, honestly -- to the point that I'm very tempted to think he missed Barbour's whole "No seriously I'm not talking about a block universe", and misunderstood the whole thing, and thinks timeless physics essentially is just a block universe, and that's all he's ever been talking about and just needs to get the terminology right. But it seems unlikely that's actually the case, because firstly are we going to believe he really made such a mistake, and secondly because he does eventually does lapse into timeless physics nonsense (by which I mean not that it agrees with Barbour, who I haven't read, but that it's nonsense; I can't make sense of it).

It has been quite a while since I read that particular sequence so I am mostly writing this without particular examples in mind, but let me pull up one I was thinking of -- not from the sequences but from HPMoR. Chapter 28; speaking of his achievement of partial transfiguration, Harry says

"I had to go all the way down to timeless physics before it took. Had to see the wand as enforcing a relation between separate past and future realities, instead of changing anything over time [...]"

Of course, these two things -- changing the state of something over time, and enforcing a relation between past and future -- are obviously equivalent, though they call to mind different intuitions. But my point is that enforcing relations between past and future, rather than changing something over time, is exactly the natural way to think of what the laws of physics do in a block universe! That doesn't suggest timeless physics at all! Hell, in timeless physics AIUI, the only thing enforcing a relation is their inherent similarity, and, well, I essentially already made this argument above in points 2 and 5; it becomes not about enforced relations, but rather questions of possibility (equations of state) and similarity. Which is not what's going on here! Again, it makes me wonder if he realizes what he's saying.

TLDR: Most of what Eliezer says about timeless physics works perfectly well as a description of a block universe, though he doesn't seem to notice it; the rest is nonsense; he should just discard the nonsense and admit he's been mostly working from a block universe perspective already.

-Harry