3 reasons why AI needs to be conscious

1 Our systems need to know what 'good' looks like and act accordingly, without us leading them every step of the way. In the context of this model of how consciousness works, that means defining their 'valence' function. This takes the place of a program and a programmer in determining what they pay attention to, what they do in response to different values of the attended data, and what they learn.

2 We need to understand what our systems are doing and to be able to communicate with them about it. This model of consciousness says that the grammar of language closely mirrors the mechanisms of consciousness; and that the words we use relate directly to the concepts we distinguish. Collaboration with our systems means that we must share a grammar of thought, and common concepts; deep learning systems fail abysmally in this regard.

3 Systems need to use times when they are not busy to improve their own performance. Consciousness lets them do this by reflecting on their past performance and planning ahead for possible future activities. Furthermore, because the mechanisms of consciousness use attention to select for active processing only information that represents change to the current situation, many cognitive cycles are freed up for reflection and planning.