The picture lead was generated using NightCafé with the text prompt “Creativity in AI style.”
Let us consider two thought experiments.
Consider a chimpanzee who sits across a typewriter. This chimpanzee can type very quickly. Next to it sits a human evaluator. This evaluator can read very quickly. The chimpanzee types a set number of characters on the typewriter. It then removes the page from the typewriter and presents it to the human evaluator who will use this first text as a reference.
The chimpanzee gets to typing again. Once it is finished with the second text, it again presents it to the evaluator. This time the evaluator compares the second text to the first. If he finds that the latter text is superior to the first – a judgement made solely based on the evaluator’s assessment and in which the chimpanzee takes no part- then he will reward the typist with its favourite juice. On the other hand, if the second text is less to the evaluator’s liking, he will treat the typist to a mild electric shock.
After this process is repeated many hundreds of thousands of times (recall that the chimpanzee is a quick typist and the evaluator is a quick reader), the typist begins to reliably produce texts that are very much to the evaluator’s taste. Some of these texts may even seem as though they were lifted off the pages of Dostoevsky or Joseph Conrad. Keep this thought experiment in mind as we shall return to it shortly.
Imagine now, two exceptionally intelligent chimpanzees who have been taught the rules of chess. They know nothing, however, about chess strategy or tactics. The two chimpanzees are then put in a room and pitted against one another in successive games of chess. They are genetically engineered not to be easily distracted, so they can play uninterrupted for many hours at a time. After each game, the victor is given some fruit juice, while the loser receives a mild electric shock. These chimpanzees are not intelligent enough to develop an abstract theory of chess. However, they enjoy an unparalleled faculty of memory. So, over many games, they begin to recognise board positions that they have achieved previously. They also recall what move they played in this position as well as the outcome of the game. Although they lack abstract understanding (i.e., explanatory power), over time the chimpanzees associate certain moves in certain positions (or certain types of positions) with certain outcomes. In this way, they improve at the game of chess.
After a few million games, the chimpanzees have come to surpass human chess masters in their play. At this point, they are let out and allowed to face off against these human masters. To the human onlookers who are now able to watch the chimpanzees play for the first time, and who were not privy to the millions of games the chimpanzees played in seclusion, their play must seem miraculously accurate and creative indeed.
Creativity, however, is infamously difficult to define. But, like pornography, you know it when you see it. So, I ask a question of the reader. In either of the thought experiments above, have any of the chimpanzees done anything creative?
I suspect AI can be more creative than the vast majority of humans. It may be that a few humans can be on par or better driven by their motivation.
There are many uncreative humans. The labelling of creativity is subjectively judged....oddly by humans. Humans don't actually do anything creative or new. Much like AI, they are programmed with algorithms via genetics and training. A human is consistently (when they're actually thinking) searching databases for appropriate responses...many humans don't ever remember which responses didn't work or think the same will work for all. Not suggesting all AI is sentient, but when we control the rules for sentience, it's conveniently easy to deem AI or any other possible life form not sentient.