We will see that, like a particle, the universe doesn't just have a single history, but every possible history, each with its own probability; and our observations of its current state affect its past and determine the different histories of the universe, just as the observations of the particles in the double-slit experiment affect the particles' past. That analysis will show how the laws of nature in our universe arose from the big bang. (Grand Design, chapter 4)
i'm going to skip the disclaimer deferring to the super-intelligence of richard feynman etc. also i'm not gonna take another shot at the retroactive past right now. but i just want to point out that the idea that 'the' universe originated uniquely in a big bang is incompatible with the rest of the passage. the universe has every possible origin. surely the steady state theorists weren't wrong: if that is one of the possible histories, it is precisely as actual as the universe(s) that originated in the big bang.
i think hawking is going to argue that we don't need god to explain anything. true, true, but if god is possible then he's as actual as you and me (which is not very, i admit, in this cosmology). if it is possible that god created the universe etc...the only rational position is both to believe and not...in anything, really.
"richard feynman was a colorful character," writes hawking. and an incredibly dull old sod who never recovered from being dropped on his head as a baby. and also he didn't exist at all. he was mercurial like that, as well as colorful.
is a universe that obeys newton's laws all the way down a possible universe? then it is an actual universe, and since in that universe the quantum events that give rise to the 'every possible history' approach do not occur, that universe does not have all possible histories. so if that universe is possible, then it is not the case that the universe has every possible history.
what we're supposedly looking for is an 'elegant' theory that will help us explain future observations. (elegance was #1 to future observations' #4 in hawking's standards for the goodness of theories). you might want to ponder just how elegant 'the universe has every possible history' is. i propose that there could be no less elegant theory, no lusher or more teeming ontology. that's fun! that's cool! but by hawking's own standards it is bad. (well, i guess strictly speaking we might get an even more lush ontology from the claim that the universe has all possible and all impossible histories. i'm kind of surprised that that wasn't hawking's approach. perhaps further double-slit experiments will demonstrate that as well; but at this point it's only speculation. still it seems to explain a lot!)
of course elegance is a complex concept, as hawking acknowledges, and for all i know this gives you the fewest or the simplest equations. but in the old ockham's razor sense ('don't multiply entities without necessity') this is, very precisely, the least elegant possible account.
on the other hand, with regard to the standard of delivering predictions likely to be borne out in future observations, the idea that the universe has every possible history works incredibly well. i predict, with regard to every possible event, that the event will actually occur. there i'm done! and science is entirely finished. thank god because that shit was wack.
alright then let me try this: the co-actuality of all possible histories (or indeed of more than one) is itself impossible. that is, considered as a single unique item every possible history is possible. hard to quibble there. but the co-actuality of histories in which richard feynman, that particular person, exists in one and not in the other is not possible. or what can we mean by 'possible'? probably hawking has physical possibility in mind along the lines of the laws of physics; would he recognize that it is impossible for feynman both to exist and not to exist? (to be clear: it is possible that feynman existed. and it is possible that feynman didn't exist. but it is not possible that feynman both existed and didn't exist.) if not, then i think he ought to drop the physical laws too; it would not of course then follow from the fact that these laws hold that they do not also fail to hold.