The curious comfort of machine intelligence

It would be foolish to underestimate Al. But take heart, there are parts of our humanity it can't replicate.

- Gwee Li Sui

I once heard a fellow writer declare with conviction, “There is nothing intelligent about AI!” These words still haunt me not so much because they startlingly ignore how ChatGPT, Gemini, Claude and the likes are works in progress and will only improve with time.

Rather, the words lay bare an old conceptual assumption that many continue to hold today. This is to think of intelligence as the sole purview of humans and, grudgingly, a few hundred other species. Intelligence is what allegedly characterises higher life forms and so can in no way describe inanimate things, which are stupid. The stone does not exhibit an awareness of its environment, nor does it dialogue with other stones.

In this belief, intelligence culminates in humankind, whose mental capacities are oriented towards a meeting with truth. An intelligent person does not just possess knowledge but also deepens it by blending in reason, memory, discernment, sensitivity, and some degree of creativity. To speak of artificial intelligence thus used to feel oxymoronic because intelligence was deemed inseparable from sentience.

All that began to change from the mid-20th century, when researchers asked a simple question about the nature of intelligence. Can the operations of the human brain be broken down into steps that an artificial construct is able to perform? Can the way we think be reverse-engineered and then simulated so that a silicon-based system may also be called intelligent?

This dream of machine intelligence has given programmers a Holy Grail to quest after for decades. The pursuit is kept on track by an assessment known as the Turing Test, named after its proposer Alan Turing, the goal of which is to determine whether a particular computational entity can show convincingly humanlike thoughts. Turing originally dubbed it the Imitation Game.