This is my take. With two sensors connected crosswise to musculature that pushes forward, you get goal seeking for targets in front and avoidant behaviour for targets behind.
It's what our ancient ancestors had. And they were successful, and elaborated on that theme, and by the time they were complicated enough to start thinking "hey wait, why is our nervous system crossed up like this" it was too much bother to uncross it.
You're just making shit up. There's zero evidence for this.
You also don't seem to understand that inputs (senses) are also crossed, meaning that in your hypothetical inputs-directly-tied-to-outputs scenario, the movement produced would be in exactly the opposite direction to what you're describing, i.e. toward the predator.
And even if you found a way to resolve these problems, movement is far more complex than you're describing, and practically requires a more complex structure between sensation and reaction to coordinate muscular contractions which produce movement. The sort of reflexive muscular contraction you're describing would look more like a seizure than a contraction which produces useful movement.
I'm all for preferring the simplest explanation which explains our observations, but your explanation doesn't explain what we observe in any way.
