> What I find interesting about this hypothesis is that consciousness is just a stream of thoughts which we do not control.
I would define consciousness as that which keeps the body alive, reacting to changes in the environment and actions of other agents.
Consciousness is a perception-judgement-action-reward loop. It's a reinforcement learning kind of situation, where the agent, embodied in an environment, learns to act in such a way as to maximise its rewards. We have been programmed by evolution to favour certain kinds of rewards which are good for keeping us alive.
This definition is much more concrete and practical - it has been applied with some degree of success in designing AI agents. It gives a purpose and a mechanism to consciousness.
The consciousness loop is formed of sensing, judging, acting, learning from sensation and reward. Sensing is like deep learning - a neural net converts raw sensory signals in high level representations. Sensory learning is mostly acquired by reconstruction or self supervision. Judging is like RL - where an agent learns to predict the future rewards following its current situation and actions. It can be model based (reasoned) or model free (instinctual). Acting is like the feedback based motion planning algorithms in robotics. Reward learning happens in the background of all these processes. When you put all these systems together, and inside a body, inside an environment, you get consciousness.
In the end it's just the competition of genes for survival, self replication in a limited resource environment over billions of years.