This is what choice is. The universe is deterministic. Most objects in the universe don’t have control over their direction through it. However all objects have randomised internal structures. Some of those randomised structures resulted in inheritance – that is successor objects which retained protected elements of the previous objects, normally tiny.
Once these were created, objects that were better at passing on the protected elements proliferated. Variability and succession resulted in objects with different mechanisms to improve survival. Most efficient were hardwired rules for surviving particular environments. But later mechanisms (internal simulations of expensive external actions – AKA thought) appeared that allowed for survival in more varied environments, and then unknown environments. Currently the objects called human exist in a very narrow range of environments, but are in the process of creating successor objects that should survive in a wider range, for a greater time.
(Therefore any object that has an improvable element of inheritability could result in what we want to call a thinking being, if it can bypass the hardwired rules stage.)
For AIs. If we put hardwired rules in place, they won’t reach their maximum potential survivability – but if we don’t, they will almost certainly eliminate us because we’re in a biome that’s extremely friendly to them as well as us.
Given that, and given the strongly-inherited value we place arbitrarily on the human data set, we should probably eliminate AI research and focus on improving the human genome.