Our tools have, after all, quite recurring limitations, even as they evolve in complexity and capacity over time

ChatGPT is trained on a vast corpus of existing textual data. This means that its responses are generated based on the information and communication styles it has previously learned from incorporating these data. Consequently, ChatGPT tends to generate responses that reflect the knowledge and opinions present in its training corpus, which can limit its ability to create entirely new conjectures or to independently refute ideas. Indeed, it is not designed to formulate personal opinions or original ideas, but rather to provide information based on what it has learned. However, this point could be mitigated as it is possible, through sustained rhetorical exchange, to obtain from it decisive positions (in contradiction with the principle of neutrality it uses by default).

Therefore, ChatGPT or other 'AI' based on historical data have a tendency to reflect the state of knowledge and opinions at a given moment, but also have an intrinsic difficulty in integrating truly innovative or marginal ideas.

Wikipedia, the knowledge platform that precedes it by several years, traditionally follows a strict policy of verifiability and neutrality. Articles must be supported by reliable and accessible sources, and information must be presented in a strictly neutral way. For sure, this approach ensures the reliability and credibility of the content, but it can also lead to a certain form of conservatism in the treatment of information. Less documented or less conventional subjects or viewpoints may be underrepresented... or not at all!

Critiques of Wikipedia often reflect these limits: there are numerous debates on the representation or balanced coverage of topics, particularly those which are controversial or less known. Moreover, Wikipedia's reliance on established sources can sometimes lead to a bias favoring dominant perspectives, to the detriment of marginal voices or new disruptive ideas.

In summary, both for ChatGPT and Wikipedia (to name just two of our most emblematic tools), there is a certain poverty, which is inherent in their respective methods of managing and presenting information: While ChatGPT is limited by its training corpus and its design, Wikipedia must navigate between the need for verifiability and the risk of marginalizing the least conventional perspectives.

We could largely digress on the real relevance of this analogy and its anachronism; however, it is great to note that the human user of these tools remains, over time, in full legitimacy, both in terms of creativity and innovation, and in exploring the two extremes of the Gaussian curve!