It imagines

The Haikonen machine can readily imagine things it has never observed. For example, having learned what a yellow ball is and having seen it rolling down a slope, next time it sees (or thinks of) a yellow ball, it can imagine a ball of a different color or a ball rolling up a slope. This imagination happens as follows: the distributed signal representation "yellow ball" evokes another distributed signal representation "ball", but the "yellow" signal line in the signal array is turned Off and another is turned On. Or, a "ball rolling" sequence is evoked from memory, then the machine reverses the temporal sequence.

It has flows of mental images

Say the machine's camera is now focusing on a yellow ball in front of it. From afar, all we see is a machine whose built-in cameras focus on a ball, it seems to do nothing much. But move closer, look at the machine's electronic brain (by watching signals on probes which the machine's builder has attached to signal lines in the brain), and we will see a busy series of events taking place all over the brain. The first thing that happens, is the emergence of a signal pattern from the camera's preprocessor circuit, whose "roundness" and "yellow" lines are on, among others. The next thing is the "ball" sound pattern, the "ball rolling" knowledge, and the visual memory of a "blue ball" which it has seen before, etc. Then these representations themselves may in turn evoke yet others: the "roll faster down a steeper slope" knowledge, the "ball rolling uphill" imagination, the "several balls rolling together" imagination, etc.

Say previously a yellow ball was hurled at the machine, damaging its skin. As part of the building of the machine, it was given a rule that says "If faced with something that has the potential to cause damage, then be prepared to move away from it". The pattern of the yellow ball evokes from memory the pattern depicting this unhappy encounter, and in turn this pattern will condition the machine's motors to move the machine backwards, by associatively evoking a distributed signal representation priming such a move. This priming, in turn, causes results which some internal sensors report to the brain (a motor sensor reports that higher current is pouring in, a brake sensor reports that the brake is tensing up, etc.).

The first wave of associative evocations from the visual preprocessing circuitry, and the evocations from them in turn, and the next wave of evocations, and so on, as exemplified above, constitute the machine's flow of mental imagery.

A cow probably has a limited flow of inner images too, triggered by blades of grass. But if it does not know about this flow, then the cow is not conscious. Does the Haikonen machine know that it has this flow, or at least some parts of it,? We shall see, further below.

It has emotions

At any given time, in any situation, numerous distributed signal representations are active because they come in from sensors or evoked or imagined. Lots are happening now. We know that the machine should pay more attention to the possibility of the ball being hurled at it, rather than on imagining non-existent purple balls rolling uphill. But how does the machine know that? The main part of the answer, in cases like this, is emotions.

Haikonen sets out the basics for the emoting by a combination of neuron-level rules and global rules.

The neuron-level rules are about Match, Mismatch, and Novelty. Say the machine is watching a yellow ball, and nothing happens. The neuron group processing this signal array outputs a "Match" signal, in addition to the signal array itself. This is because a signal array fed to its associative inputs represents a yellow ball seen a split second before, and this associative input matches the main input array. Now the ball starts to move away. Immediately after it starts moving, the new input array is different to (smaller than) the pattern at the associative input, and a Mismatch signal is generated, at least for a while, attracting attention. If now the machine looks at a square box, something it has never seen, then the associative inputs contain nothing. With nothing to compare to, the neuron group outputs a Novelty signal.

Apart from Match, Mismatch, and Novelty, the machine also has rules to build its capability to know what is pain and what is pleasure, plus what is good and what is bad. A few rules are something like: "Pain is Bad", "Tearing of skin's sensors is Pain", and "Avoid Bad things". Another set of  rules, which gets evoked when the battery is getting charged, reads something like: "Battery level getting from low to higher, is Pleasure", "Pleasure is Good", and "If it feels Pleasure, keep on doing it ".

In nature, presumably these rules are from evolution, with individuals and species stumbling upon such rules thriving and passing on in their genes as they reproduce. Whether the Haikonen machine can learn these rules by evolving, or they must be given by its builder, is not clear. The former would be more satisfying.

Together, these reactions form the machine's emotions and their associated behaviours - from Fear (Bad and Pain => Withdraw) to Curiosity (Good and Novelty => Approach) to Desire (Good and Pleasure => Approach), etc.