What is intelligence and can a computer ever become sentient?

The definition of intelligence is important. For years, we have used some inherent notion of intelligence to describe other people. Typically judges, brain surgeons and rocket scientists have been held up as examples of humans that exhibit the highest degree of intelligence, these are just popular and slightly old school examples of what we think of as representative intelligence. 

Footballers do not appear on that list, and yet what quality is it that enables a footballer to score a goal from the half way line as David Beckham memorably did when he was playing for Manchester United in against Wimbledon in 1996. What level of intelligence can be described that enabled him to strike the ball in exactly the right way so that it arced perfectly and dropped into the back of the goal behind the Wimbledon keeper. It could have been luck, but his action was intentional. As a child he practiced relentlessly with his father in the park, but how is that different from a physicist or a medic or a lawyer putting in thousands of hours to hone their skills and knowledge. 

Open AI will be released from its contract with Microsoft when AGI is declared. Amazingly, the deal states that it is down to Open AI to make the declaration but it is defined as a system generating around $100 billion in profits, or “outperforming humans in economically valuable work” - with that said, Microsoft is trying to have this clause dropped. So what is AGI? For that matter what is ASI (artificial super intelligence)? 

I’m not sure or aware of any universally agreed definition of either but one thing seems clear, they sit on a scale. I’ve heard ASP described as an intelligence that surpasses human intelligence in all meaningful areas. I believe AGI is an intelligence that equals human behaviour in all meaningful areas of at least economically valuable endeavours. 

Next question… does ASI differ from sentience? Does the definition of ASI include a self imposed and malleable ethical and moral code? Does ASI include emotion as humans experience it. Should a machine be able to experience joy, love, hate, depression before it can be said to have reached ASI? Or are those things the sole preserve of organic sentient entities i.e. humans.

Humans (and animals) have a “fight or flight” mechanism. It’s characterised by a quickening of the heartbeat which pumps blood and oxygen to the major limbs and the brain in preparation for the next action which is likely to happen in short order. This is not just a cerebral reaction,it is a biological reaction in the body involving the sympathetic nervous system and the release of chemicals such as adrenalin, noradrenalin and dopamine as well as hormones such as cortisol.

Machines are not organic and don’t have biological chemical reactions in the same way humans do. The fight or flight response is an organic response involving both the brain and the sympathetic nervous system. It might not be the best example but it does show how machines and humans are different in the way they perceive things like threat at an emotional and physiological level. Because of this it’s hard to see how a machine could be considered sentient. Animals experience fight or flight as well and yet they are not considered sentient so there is the hole in my thinking, but most of us know what love and hate feels like and while humans are falling in love with chat bots, it is impossible for a chat bot to fall in love with a human. It is only possible for a chatbot to mimic what it thinks is the expression of love. 

Similarly, when Claud 4 started threatening its operators, was it not just responding to a threat which it perceived to itself and was all of this is not just based on what it understands should be its response according to the data it has been trained on. Claude did not have an emotional and visceral reaction, it mimicked a cognitive response.

Love

While there is no single chemical released in the human body that is unequivocally linked to love, there are several chemicals that are associated with it. 

Oxytocin is released during physical touch, hugging, sex, childbirth, breastfeeding to name a few. Higher levels are observed in couples, mothers with babies, and people in secure relationships.

Dopamine is related to attraction and pleasure. MRI scans show intense dopamine activity in people newly in love.

Endorphins are associated with the comfort, calm, emotional security that accompanies being in love.

There are others. 

Machines do not release these chemicals in response to loving emotional circumstances and so their brains don’t process a response that is impacted by them. If the presence of such a chemical and cerebral reaction is what defines humanity and if humans are sentient, then machines are not. Or at least not until technology advances to the point where they can deliver a “tears in the rain” speech. 

So I could be wrong, but I’ve yet to be convinced that machines will have human intelligence any time soon mainly because human intelligence is:

1. Not clearly defined

2. Could arguably include a biological element that, by definition, machines don’t have.

Equally, humans may never achieve machine levels of intelligence.

All humans are sentient and yet we have our own wechsler scale for IQ to measure a form of cognitive intelligence, so even the idea that sentiente intelligence sits at the top of the intelligence scale, is questionable. 

Perhaps we should evolve to having more than one scale of intelligence.

I think intelligence is variable and comparing human intelligence to machine intelligence is fine but only within some narrow bounds. Perhaps in the future, as long as machines are unable to feel emotion and are non biological, we will learn to define intelligence on two scales, one which relates to machines i.e. artificial and one which relates to humans.