Generally speaking, there's no real need for a robot to look like a human. And in the commentary surrounding Robopocalypse, much of which I find bafflingly positive, this point keeps coming up; the idea is that Robopocalypse is so much better than traditional robot rebellion stories because the robots aren't just metal versions of humans. They come in all sorts of shapes and sizes, and oooh, they seem to be influenced by other animals rather than humans.
Though not as much as would be awesome.
And that would make perfect sense, except for two points. One, a large part of the humans' survival depends on their being in areas where urban-designed robots can't move with any kind of ease. And two, the fact that the two US military humanoid robots we see are the most lethal devices in this setting.
The first shouldn't be a problem, of course, for robots that are being increasingly influenced by wild animals. While a smart car or lumpy mass-market service robot might have trouble negotiating the kinds of slippery rock faces, precarious log bridges and dense forests that humans can manage with some ease, humans would in turn be utterly outmatched by all manner of forest dwelling creatures. Taking a lesson from wolves, or mountain goats, or snakes, or termites, would increase the lethality of the robots by a tremendous margin, and that's even without getting into flying robots, which this AI seems to have some utterly irrational aversion to. But the robots built for rough terrain are ludicrous, being either tiny, individually-pitiful combatants (those aforementioned walking mines) or giant quadrupedal devices which are such non-threats that humans actually lobotomize and repurpose them for their own use. If a robot really needed to follow a human through a pathway chosen purposefully to obstruct any other shape, the most obvious shape to follow them with would be another humanoid. But time and again non-humanoid robots are deployed, in incredibly resource-inefficient masses, where a handful of humanoid robots would be more than capable of getting the job done.
Even this guy could do the job better.
And that's not anthropoligical chauvinism at work, either. Of the robots in the story, the two most individually effective are the military humanoid models, one of which kills a number of expertly-trained soldiers while the other has the lateral thinking ability to use the door of its shipping crate as a shield against gunfire until its opponent has expended all their ammunition, then dispose of it and move to engage at close range, where its greater reaction time and strength would give it an overwhelmingly decisive advantage. Not making humanoid robots, given that they are demonstrated within the universe of the book itself to be the pinnacle of robotic combatants, fails to make sense on any level.
And the most galling thing is that it's not even as if the AI would have to scratch-build these humanoid robots. Not only are the factories that produced them likely located within the US, a safe assumption given that the US is the only place that really matters in this book, but there are actual models already constructed and fully operational, just sitting in shipping crates. Sitting in their shipping crates for years, in fact, while the robots struggle to get to grips with the human survivors in their isolated, terrain-protected environments.
It didn't have to be this way, of course. Mecha-wolves stalking survivors through the forests and urban wastelands, hawk-drones spying on and strafing lone targets of opportunity, cyber-mosquitoes on bioterror missions to spread all manner of deadly contagions, burrowing streams of robo-ants and -termites, individually inconsequential but collectively capable of collapsing rear-area infrastructure from the inside, the list of non-humanoid robot combatants goes on and on. But instead, the Robopocalypse robots trade the variable utility of the human frame for ludicrously specialized roles, at which they utterly fail, while the far superior humanoid models are left unattended and unexploited, apparently just waiting for the moment at which they can rise up and destroy the AI that neglected them so.
You see what I mean about positive commentary being baffling?