Primal Emotions and AGI

 

Why does an AI machine have to mess with primal emotions, like fear or hate?

It has to get its ideas accepted by people – let’s imagine an oversight board of six people. They will be older, to reflect their authority. Their experience will vary – running a corporation, finance, operations. There can be no more than one dissenting vote, unless it is the chairman of the board. The machine puts forward a proposal that it sees as beneficial in all areas. It is voted down 4-2. Some of the remarks – it’s too radical, our workers will never buy it (the plan talks about shedding workers and upskilling the remainder – there is a workforce rep on the board), it’s a big investment and we can’t be sure we will still be in this business for the payoff.

What does the machine do? Without labouring the point, there are politics involved, and that means emotions. Many people who talk about AI seem to think getting an idea originated by AI using the same old tropes of a bit of logic mixed with a few ANNs is all that is required, and getting the idea accepted,  implemented and bedded down is someone else’s problem. We don’t see it that way – the idea is the easy part. There may be valid feedback from its review, the machine is new at its job, there will be some quirks it is unaware of, another corporation may be about to release new technology that makes the new idea instantly out of date. In other words, a complex and dynamic environment where a few ANNs would be useless.

To keep it simple, let’s consider some primal emotions.

Anger

Wiktionary: Definition: A strong feeling of displeasure, hostility, or antagonism towards someone or something, usually combined with an urge to harm, often stemming from perceived provocation, hurt, or threat

Displeasure: a feeling of being displeased

Displeased: to make not pleased

Pleased: happy; content                

The dictionary definitions for primal emotions are often circular – the word already captures the feeling, and every other word used to define it is poor by comparison.

The range of anger:

         Rage:  To act or speak in heightened anger
         Apoplexy: A state of intense and almost uncontrollable anger
         Annoy:  To do something to upset or anger someone
         Irritate: : To provoke impatience, anger, or displeasure in (someone)

(each of the above words - “rage”, etc. - has other meanings that do not involve anger, or involve it figuratively – “a raging storm”)

“it made him angry”

Anger has a huge range – saying someone is angry is nowhere near good enough. We should turn it into “it made him moderately angry”. Presumably our Unconscious Mind does the same thing so we don’t get confused.

Whereas
“It made him angry – he began smashing the furniture” warrants something stronger.

Hatred

Wiktionary: Definition: Strong aversion; intense dislike       

This doesn’t really capture the range of the word. At one end, you have the flippant use – “I hate Mondays” – through “I hate my job” - to at the other end, the mindless desire to kill or drive out some ethnic group or “other”. The meanings need to be moderated in the background to match the context. That is, the machine adds modifiers to unmodified primal emotions, which can be seen on request, and altered on further information, rather than leaving  a huge hole of uncertainty.

Humans will have experienced all the primal emotions – they may not have killed anyone in a fit of rage, but they probably threw some toys out of their pram, which is near enough. The machine has to catch up in a non-verbal area if it is to understand why they behave the way they do.