What could such a difference be pinned on at all? When it comes to intelligence, we only ever see the surface, at best intelligent behaviour, a symptomatology of intelligence that can be measured by certain criteria, whether human or so-called artificial. And if both show the same symptoms – and this is undoubtedly more and more often the case – do they really still differ, and if so, in what? This question was also the focus of a paper by Alan Turing (1950). At that time the question did not have today's topicality; intelligent systems did not really exist yet, nor did the terminus technicus "artificial intelligence". Nevertheless, the question whether artificial intelligence is possible already existed in theory. Turing devised a test scenario for demonstrating intelligence. Several persons carry out the test. Each person interacts with two players, A and B, only in writing and without seeing them, one of the two players being a machine and the other a human being. If at least 30% of the persons carrying out the test are convinced that they are talking to a human being although it was the machine, then the machine has passed the Turing test. In other words: if both A and B show symptoms of intelligence in their answers and no essential differences in their intelligent behaviour can be discerned, then the machine too has passed the "intelligence test", i.e. it possesses intelligence. Turing himself presupposes this as a matter of course anyway, so that for him the question whether a machine could have intelligence is actually already obsolete.

Of course this test would only make sense if a considerably reduced concept of intelligence were presupposed, namely that outwardly recognizable symptoms of intelligent behaviour suffice to demonstrate intelligence. The absurdity of such a reduction of intelligence, and hence of this "intelligence test", is evident already from the fact that AI has meanwhile advanced so far that everything has tipped the other way. In a Turing test it would now no longer be a matter of whether it can be proved that AI has at least as much intelligence as a human being, but whether it can pretend that it has no more intelligence than a human being. That is, AI would then have to be meta-intelligent, as it were, able to control its symptoms of intelligence so that it is not exposed as artificial on account of its superintelligence; it would have to turn its intelligence down a few degrees, to human level so to speak, for example by reducing the complexity of its output and building a few human rough edges into the stylistic smoothness (put negatively: the blandness) of its diction, and so on. One more sarcasm marginaliter may be permitted: it is to be expected that in future student theses (and presumably not only these but other academic work too) will increasingly not merely be garnished with small doses of AI text modules but realized in their very substance with AI texts on a larger scale, and that the demand for tips and guides for prompts, and also for usable concealment strategies, will accordingly rise – which may in turn be generated artificially intelligently, and so on. But to bring this to an end, it is clear: simulating is not good; it is better to find one's own style in thinking and language, to reflect in one's lifeworld, political and personal context, and to enjoy authenticity and creativity.