We think because we define ourselves by the positions that we hold and by taking joy in the fact that we have arrived at those positions on our own. If you care more being correct than being meaningful, then you have a faulty picture of the world we are currently in and what we need to do in it to succeed. Wanting to be correct as opposed to wanting to be meaningful is really not even being sensitive to the context of the game we are playing at present. It is like playing football when actually the game that is going on is table-tennis. The desire to be correct prevents us from actually understanding the flow of the conversation and the kind of intervention it needs us to make.

Striving to make human cultural material (images, text, music, video, food, clothes, adornments or habitats) with AI is playing a game of deception. A game of deception that makes one appear to be smarter and more capable than one is. This may lead to one of two different eventualities. Either we (collectively) as a society learn to see through appearances and learn to value people for what they are or we go on living in a world in which this appearance-enhancing deception goes on having a hold on us and makes us do things which are deluded, is bad for us and is bad for everything we do.

Human cultural material makes the beholder construe an image of the mind that might have produced it. This image might be false but it is important in how it makes us perceive the author of that material. In the case of AI-created human cultural material this image is a mirage. It is faulty. This image is responsible for the erroneous perception of intelligence in AI, in what is but a carefully engineered mimicry. In this text I propose that AI should steer clear of trying to replicate human cultural material (an effort which is deceitful and not really helpful to either furthering the cause of human civilisation in any way) and stick to code-generation. Code can basically be understood to be a specialised way of instructing machines to perform actions.

Of all the use-cases of AI technology of the past few years — I find code-generation to be least problematic because at the end of the day — it really makes sense for a machine to be instructing another machine. If most of the funds and engineering resources in the field focussed on this one use-case dedicatedly, it would get much better, much faster.