The MIT Technology Review headline on artificial intelligence reads: AI just beat a human test for creativity. What does that even mean? I'll tell you what it means. It means we're in dire need of a better universally accepted definition of creativity. All those scare makers are just there to ride on the popularity of the terms "AI" and "Language Learning Model". AI does what AI does: it generates stuff. That's not creative at all---it's just a single part of creativity.

Shallow articles that have nothing to say aside, I was all the more disappointed when reading Mika Koivisto and Simone Grassini's¹ paper called Best humans still outperform artificial intelligence in a creative divergent thinking task, where the abstract mentioned that on average, AI chatbots outperformed human participants (going even further: while human responses included poor-quality ideas, the chatbots generally produced more creative responses.)

The term "creative" or "creativity" should not have been used in the above paper: the title itself indicates that it's all about divergent thinking. Of course AI is better at generating a more diverse set of so-to-speak original ideas: its gargantuan dataset is based on the ideas of millions of other humans! We could never possibly have that much experience, meaning we can never possibly come up with that many diverse answers---and that's okay. In an interview with software developers, while probing how they perceive creativity and creative problem solving, one participant said: "creativity is the brew of different inputs". Chatbots have had a lot more input than we'll ever be capable of processing.

If you dig deeper, you'll uncover even more evidence proving the misuse of the term creativity. Those "tests for creativity"? Those are either Torrance's Test of Creative Thinking (TTCT) or the Alternate Uses Task (UAT), both one-dimensional divergent thinking tasks developed in the seventies that have little to do with modern creative problem solving. Of course coming up with as many different alternative solutions as possible is still relevant today, but like I mentioned before, it's only one dimension of creativity. It begets all the others: collaboration, critical thinking, a creative mindset, tooling, etc (see what is creativity in software engineering?).

Chatbots, AI, Language Learning Models, code-generation tools---or whatever you want to call them---should be treated as just another way of getting input. Input that still needs to pass the different critical thinking phases: if you proudly claim never to blindly copy-paste Stack Overflow answers, then why are you blindly (and still proudly?) accepting Copilot's proposal to generate that piece of code? Perhaps it's not the best fit here. Perhaps the proposed code reminds you of another problem that you once discussed with a colleague, concluding that the best way to proceed here is to reject the proposal and implement a variant of the one you were reminded of. Without the proposal, you probably didn't remember. But without critically evaluating the proposal, that part of the code wouldn't have been as reusable and clear as it is now.

Granted, Koivisto and Grassini do make amends in the last sentence of their conclusion:

It should be noted that creativity is a multifaceted phenomenon, and we have focused here only on performance in the most used task (AUT) measuring divergent thinking.

That warning should have been placed in the beginning, in the middle, and in the end, as people are quick to generalize and some pages contain the word creativity more than 30 times.

In the academic field of cognitive psychology, a once generally accepted definition of creativity (something original, qualitative, and relevant) long evolved into a more multidimensional socio-cultural phenomenon. Unfortunately, in the field of computing (education), we found that this evolution hasn't seeped trough yet. That means many very recent published works are still relying on obsolete ideas of the concept of creativity! Long live the speeds at which interdisciplinary research travels.

In the above paper, the authors more or less refer to the aged definition:

The standard definition of creativity describes it as the ability to produce ideas that are, to some extend, original and useful.

Hence something hot in computing (AI) is wrongly correlated against another hot topic outside of computing (creativity) using old---and incomplete!---definitions and even more obsolete tests.

AI does not beat humans in creativity. AI beats humans in generating new subsets of something based on millions of records---records that human artists once painstakingly crafted, by the way, but the ethical aspects of all this is a whole different problem I'll leave for another rant. Creativity is more than generating ideas. Please stop equating divergent thinking with creativity. Thank you. End of message.