Lhe recent launch of the ChatGPT “conversational robot”, developed by the OpenAI company, immediately triggered an avalanche of comments. What was most often noted is that the system, designed to answer written questions, offers answers with syntactic quality and consistency “bluffing”. However, this observation was nuanced by the fact that the result is still in its infancy; a number of advances would still be necessary for it to resemble a result produced by a human. But it is precisely there that lies our great illusion: that of believing that these are devices using a language similar to ours.
This is the reason why it is necessary to go and see what are the springs of these technologies, far from the speeches forged by the digital industry, so often taken at face value. What characterizes these statements is that they are only the production of algorithms feeding on statistical analyses, then taking their sole source from already existing registers. In this they are unrelated to what so-called “natural” language supposes.
Because what is specific to human language is that it stems from a tension between a vast lexicon, made up of words and grammatical rules, and our ability to generate formulas. And this, in a relationship to time that is not exclusively attached to the past, but is part of a dynamic combined with the present and constantly becoming. When we speak or write, we constantly draw from a phraseological ocean, while adjusting ourselves, in an indeterminate way, to a specific context each time. Any locution, written or spoken, stems from an outpouring which invariably exceeds any prior schematization.
This dimension is absent from the mechanical verb, the result of settings that only respond to specific functionalities, for example in the personal assistant Siri (owned by Apple), which tells us “What can I do to you?” »or in connected speakers such as Alexa (developed by Amazon), whose sole purpose is to guide our decisions for mainly commercial purposes.
Double transformation
Rather than asking ourselves naively if these systems will soon replace us in the writing of texts – a sign, then, of a definitive renunciation of the use of our own reason –, we see the civilizational model which, at the bottom noise, establishes itself? That which proceeds from a double transformation of our relationship to language. On the one hand, so-called “generative” artificial intelligences, endowed with the power of speech, with airs allegedly identical to ours, are gradually being delegated the task of managing our relationships with others and many of our current tasks – a faculty which, however, conditions our right to pronounce ourselves in the first person and to conduct ourselves according to our judgment within a free and plural society.
You have 47.66% of this article left to read. The following is for subscribers only.