Bipeds need big brains. The arguments above, derived for spinal structures simpler than sea-snakes, apply to humans too. Two differences are crucial. First, the force of gravity now breaks the original symmetry of an aquatic environment. More importantly, humans balance upright, even sometimes on one foot, a horrific computational task (equivalent to balancing an inverted compound pendulum containing seven stacked ball-joints). This requirement alone demands a brain with far higher bandwidth and sophistication than even other quadrupeds. (The other uniquely human skill, of throwing, is also computationally difficult, but less obviously a consequence of the environment).
Bipeds must feel pleasure from good posture. Humans as a species evolved from quadrupeds, and individual human babies balance on four limbs before they manage on two. Obviously that balancing-act has evolutionary benefits--humans walk and run more efficiently than quadrupeds--but evolution's feedback has the timescale of lifetimes, too long to train a toddler. An individual creature needs an immediate incentive to want to stand upright, a kind of postural pleasure, to get the practice it needs to gain the skill. While the sensory quality of that pleasure can't be derived from mere anatomy, the gross mechanics can: the pleasure must lie in 'hip-opening,' the process of unbending the original right-angle joint between lower back and thigh to subtend 180° instead.
The mystery of the missing tail. Every mammal save the human (and perhaps chimps and bonobos) communicates with tails. The human tail is vestigial and invisible, yet we are the most communicative species of all. What happened to the original tail-wagging circuitry?
The native human communications protocol is vibratory. As complex representational structures, human brains need to communicate their vibratory eigenmodes to heal defects in their motor maps. This signal-processing requirement is agnostic to sensory channel. Mechanical (proprioceptive) linkage, cutaneous touch, sound, and visible tremors (or micro-expressions) all serve to inject high-bandwidth vibrations from one human to another, and all can be reciprocated with zero added latency or jitter between people in close proximity. The best model for this form of interaction is the natural collective resonant modes of coupled, actively-stabilized mechanical structures.
The native human computational paradigm is group sensation. Humans naturally live in social groups, among whom resonance is inevitable…most emotions are contagious within seconds because the inter-personal vibrator bandwidth is so high. The natural state of social groups is thus as collective computers, which synchronize perceptions, emotions, and interests as fast or faster than an individual on his own would change.
Certain vibratory patterns persist past individual lifetimes. Any tightly-coupled group of individual computational units will spread patterns over time (the principle behind the spread of viruses, whether virtual or real). These temporal patterns often can recirculate far longer than the memory-time of the individual units. So among humans, those vibratory communications which are easiest and most pleasurable to transmit will be amplified, forming a kind of "vibratory culture" in which any individual is but a small contributor. The same principles of spontaneous symmetry-breaking which allow brains to crystallize around low-entropy sensory data also allow collective vibrations to crystallize around low-entropy vibratory patterns like vocalizations and gestures. In the "medium" of human society, vibratory patterns can self-amplify and propagate, i.e. our favorite songs and dances last for generations.
Material patterns persist even longer. The patterns contained in man-made material objects (originally tools or cave-paintings) can spread in the same way as vibrational patterns: those objects/patterns which are easiest to construct and most attractive to copy will dominate over time, and can form a "material culture" analogous to vibratory culture. But vibratory patterns, like sound waves, persist only through collective dynamics and individual memories, lasting mere seconds during transmission before being remembered or forgotten. Material patterns, on the other hand, can endure years or centuries, thus gaining the same long-term competitive advantage as computer viruses which write to disk instead of RAM.