3 ms·
> A machine with a conscious (simulated) mind can feel wonderful on the first fine day of spring and grow depressed as winter sets in No, that's a human perspe
by visarga 10y ago
> A machine with a conscious (simulated) mind can feel wonderful on the first fine day of spring and grow depressed as winter sets in
No, that's a human perspective, based off our bodily needs for sun, air and nature. It's got nothing to do with conscious AI.
Instead of "consciousness", which is a slippery word, meaning many things to many people, with implications in neurology, physics, AI and religion, we should use the RL framework.
In reinforcement learning, you only need to define 3 things: perception, action and reward. They are clear cut, don't leave so much to interpretation and have direct applications in AI agents and robots; human brains use RL as well.
If we ask ourselves: can we make a general RL agent? Can we master perception, action selection and learning from sparse rewards? Well, it's a whole different game when put in these concrete terms. We don't get lost in qualia, dualism, monism, what it's like to be a bat, hard problem, and other armchair consciousness philosophy concepts.
Perception/action selection/reward is enough to describe consciousness.