⁵To give some pointers and details on these steps: Desire was defined as a process that suggests new goals in Chapter 3. Such suggestions were seen to be possible by neural network computations in Chapter 7, resembling those performed by AlphaGo. In particular, such computations give fast approximations of valences, i.e. how rewarding near-by states are, preferably taking into account the whole future by approximating the state-values. Valence computations were further linked to the generation of interrupting (intruding) desire in the framework of the elaborated-intrusion theory of desire in Chapter 8. After the selection of one of such suggested goals, planning proceeds as already described in Chapter 3, including the idea of commitment or intention to the goal that has been selected among the possible high-value states. Note that in this framework, goals are set by the agent itself, based on predictions of rewards, which come from the outside world.