Here is a snippet from it:
Actually, both of those objections seem irrelevant to me. The problem here is that apparently everybody is imagining that science is something other than what it is! Is there any question that what science is is a prescription for obtaining demonstrable knowledge about our universe (as opposed to a warm fuzzy feeling of understanding that is untestable)? So does that not mean that science is about motivating new experiments and organizing the data of existing ones? That just seems perfectly obvious to me. So the problem with speculating about variations in the vacuum energy is not that it is "giving up" on the problem (what demonstrable solution is being given up on there?), and it is not that the other values are known to be inherently unobservable (what other values can we know we can't observe?)-- it is simply that we have not observed them, nor is this approach giving us any clever ideas about how to observe them. That's sufficient, right there, to say it isn't science!

One does not invoke a contrary philosophy to argue against anthropic philosophy, one simply defines what science is and demonstrates that anthropism does not provide any of the proven benefits of science. To me, the crucial point is that the whole reason we invented science was to provide one mode of inquiry into truth that we could actually demonstrate, whose goals were testable and whose primary purpose was not simply to allow us to pretend we know more than we do. The latter is the hallmark of unscientific modes of inquiry-- including anthropic thinking. Why wouldn't Carroll count that as the main argument against it?

Carroll instead sets up the strawmen and shoots them down thusly:

I must agree that science does not first decide how reality works, and then force it to conform. But I would point out that this is precisely what anthropic thinking does, because it first posits that there has to be a reason that the variables are what they are, and then says the only we reason we can think of is anthropic. Both of those steps involve an unscientific fallacy-- the first is to state that it is part of a scientific principle to say we have to be able to understand something, and the second is to say that the only possibility we can think of is the correct one. Fortunately, real science does not rely on weak natural philosophy like that, it relies on experiment.
The problem here is that Weinberg is clearly wrong, and in just the way that Deutch's "black box" image suggests. But that doesn't make Deutsch right about multiverses as science! Yes, we don't just want predictions, we want understanding. I've always said that, I've said that Occam's razor is the beating heart of science, I've pointed out the inadequacy of "google science" (quite analogous to Deutsch's black box) where we just interpolate every answer from a body of stored existing data. But none of that suggests we need multiverses. The flaw in multiverse "science" has nothing to do with the idea that we only want to make predictions, it has to do with what we are going to call understanding, and how science can tell when and if we really do understand something.

This is not to suggest that scientific understanding is all that matters, it just says it is all that is science. I typically find Weinberg (like Dawkins) to be hopelessly positivist (the only truth worth having is the scientific truth, in effect). I did like Deutsch's point about the weakness of "selfish gene" approaches, and I have also asked the question, how do we know a gene wants to survive? Maybe they all crave oblivion, and the ones that don't achieve it are the failures. But I felt there were a few problems with "Deutch's Law:"
The first problem with it is that it would not seem to support his view of science as being forced to look for philosophical explanations. I would challenge him to cite even one single problem, even one, that was ever "solved" by philosophy! To me, his law states that no philosophical questions are interesting, which seems to contradict the way his view on science is being described here. Unless, that is, he will define "solved" as "convinced himself, in the absence of any actual experimental data on the subject".

The other problem with his "law" is, how do we know if a problem is soluble or not? Are there any examples of a problem that was known to be soluble before it was solved? Even one? If not, I claim that the only way we know if a problem is soluble is by solving it. But then the only problems we can know are interesting are ones that are already solved, and an already solved problem is no longer a problem at all. So the intersection of what is known to be interesting, and what is known to be a problem, is the null set using Deutch's Law. That's not a very useful law!