John Searle (1980) did not agree with the Turing test either. The quasi-intelligent machine does not understand anything of what it computes and of the quasi-intelligent symptoms it produces. As an alternative thought experiment Searle proposes his Chinese room. A person who does not know Chinese is locked in a room. Outside the room is a second person, who can drop slips of paper with Chinese questions into the room through a letter slot. The person in the room does not understand what is written on the slips, but can, using a table, assign to these characters characters forming a Chinese answer and pass this slip outside. Even if the person outside the room believes that inside there is a human being who answers in Chinese, his behaviour does not by any means prove that he really understands Chinese. That is, even if a machine behaves intelligently and passes the Turing test, this does not yet mean that it understands anything. At best it shows only "weak" but not "strong intelligence". Searle has thus turned the question around. Unlike the Turing test, his concern is not to prove that an AI system possesses intelligence like a human being merely on the strength of its quasi-intelligent behaviour. Searle starts from the question whether it is at all necessary for the person in the room to know Chinese (i.e. to be intelligent in this sense), and arranges a setting in which the person actually acts like an AI system with "weak intelligence", which, however, makes no difference to the person outside. But "strong intelligence" is not at issue here.

In terms of epistemology and philosophy of science this means: in the sense of behaviourism, according to the Turing test, a machine that behaves as if it were intelligent, and in this respect shows symptoms of intelligence, would actually be intelligent. In the sense of hermeneutics it would – if one makes this distinction – be at most "weakly", but not "strongly" intelligent, for the occurrence of these symptoms is no proof that the machine understands what is said to it and means what it says.

The AI system ChatGPT seems to agree with this too, as the following AI interaction² shows. Incidentally, I deliberately use the expression "interaction" in order to emphasize that this is not communication between two human beings, although I myself play along with the communicative setting customary in ChatGPT and put this question to the system as if it were a person:

⇒ "I have a question in epistemology and philosophy of science. Recognizing, understanding, meaning, learning and doing something requires intelligence. Are you intelligent? Do you also understand what I write, and do you also mean what you 'write'? Or do you simulate intelligence with what you write?"

⇐ "I am an AI model that has been trained on huge amounts of text data. My answers are based on statistical patterns that I have learned in this data. This means that I have no conscious thoughts, no self-understanding and no understanding of what I write. I generate text only on the basis of probabilities and patterns present in the data.

I simulate intelligence by using algorithms, but I do not really understand what I write. The distinction between human intelligence and my 'intelligence' is exactly what makes the epistemological discussion about artificial intelligence so fascinating. It raises the question of what it really means to understand, to know and to think."

Could it be that I had already put ChatGPT on the right course with my somewhat suggestive question? At least the artificially intelligent statement proves that it is a good simulation of intelligence. What is still irritating, however, among other things, is the formulation "I am …". Even if this "I", as I said, belongs to the customary AI communicative setting, such language is not merely an incidental and cosmetic detail. The anthropomorphism creates a misleading reality that can stand in the way of an objective, technical, rational, responsible and meaningful use of AI. It may be assumed that staging an informal communication, as it were between two persons, suggests an authenticity that leads to information appearing more credible and being questioned less critically – in contrast to print media or online media with static content, which are received with more distance and more critically and in this respect promote a scientific approach.

But let us look more closely at ChatGPT's concise "self-knowledge" by considering, epistemologically and in terms of philosophy of science, "what it really means to understand, to know and to think". To understand means to grasp the meaning or sense of a sign. In principle anything can be a sign – and in this respect have meaning or sense – provided it is used as a sign (and in this respect is also a medium): for example a linguistic articulation (we mean and understand the sense of sounds), a behaviour (when we shake hands, we mean and understand this as a greeting, an affirmation or a farewell), any media artefact (we mean something by a text in a book, and we understand writing, and so on). Meaning and sense are what someone means when using a sign, and what we then – if all goes well – also understand in this way, in this meaning. If all goes well, for this is not so certain: signs are really only symptoms of something, i.e. we have in principle no certainty that we understand these symptoms, these signs, a behaviour, linguistic sounds, gestures and so on correctly – or that we mean them in such a way that there is a chance that another person understands them correctly, i.e. as we mean them.

Meaning and understanding signs is nevertheless possible – so the position of hermeneutics – on the basis of a pre-understanding, a preliminary understanding, which we, when we understand something, share with someone who means something, because we share a form of life, culture and socialization – not wholly in common, but enough for communication to be possible. Herein consists the hermeneutic circle: we must have a pre-understanding, that is, already understand or have understood something before we can understand something else. This circle between pre-understanding and understanding is the condition of the possibility of our being able to mean something and understand something. We understand only what is available in the potential of our pre-understanding.