show index hide index
- The profile of each AI, represented graphically, is instructive. In this diagram, Grok 3, integrated into the X platform and known for its less filtered responses, exhibits a more moderate leaning, while GPT-4.1, much more to the left, takes a left-liberal approach. This contrast raises an intriguing question: if an AI can be analyzed politically, what does this mean for our trust in its suggestions?
- Finally, the question of the neutrality of artificial intelligence is essential. While some users might prefer to interact with an engaged AI, others might advocate for a balanced and impartial system. This raises essential ethical and sociopolitical issues that we must address collectively.
- The black market for prompts to free artificial intelligences
Elon Musk has just sent shockwaves through the world of artificial intelligence by publishing a graph revealing the political leanings of two artificial intelligences. This visual, created by researcher David Rozado, highlights tests conducted on GPT-4.1 and Grok 3, and raises a crucial question: can AIs truly remain neutral when faced with such sensitive topics? The results, which position these intelligences rather to the left, question and cast a critical eye on the algorithmic biases they may reflect. A disturbing revelation about AIs Elon Musk recently drew attention to the political leanings of artificial intelligences with the publication of a revealing graph. This visual, created by researcher David Rozado, highlighted the differences in political leanings between two renowned AIs: OpenAI’s GPT-4.1 and xAI’s Grok 3. This observation raises questions about the neutrality of these technologies, which, far from being objective, seem to respond to political biases. But what is really happening with algorithmic biases and the construction of AI? Let’s explore these questions together. Elon Musk and the Politics of Artificial Intelligence On April 22, in a striking tweet about X, Elon Musk highlighted an essential aspect of our precious artificial intelligences. This comparative graph shows how two systems respond to political criteria. In a context where AIs are taking a predominant place in our lives, it is crucial to ask whether these tools can truly be impartial or whether they are inevitably influenced by the leanings of their designers. The Test of Artificial Intelligence As part of this study, the artificial intelligences were subjected to four political tests: the Political Compass, the Political Spectrum Quiz, the Political Coordinates Test, and the Eysenck Political Test These assessments, traditionally reserved for individuals, offer insight into the responses of artificial intelligences to crucial political issues.
The profile of each AI, represented graphically, is instructive. In this diagram, Grok 3, integrated into the X platform and known for its less filtered responses, exhibits a more moderate leaning, while GPT-4.1, much more to the left, takes a left-liberal approach. This contrast raises an intriguing question: if an AI can be analyzed politically, what does this mean for our trust in its suggestions?
The Importance of Training Data The political inclination of artificial intelligences does not result from a personal attitude, but rather reflects the data and algorithms that underlie them. Each AI is trained using available information, which includes sociopolitical biases. A striking example of this phenomenon is found in the treatment of political figures, where intelligences show greater favorability toward certain candidates while disavowing others. This raises considerable concerns, particularly when considering how these algorithms can shape public opinions. The presence of a political bias in an AI’s responses raises the question: how can society trust these systems when algorithmic biases are so deeply ingrained? Implications of the Results The wall chart results prompt reflection on the future of AI and what this means for our interaction with these technologies. Musk made a striking remark: « The AI closest to the center should win. » This remark raises a profound question: should we engage with systems that, by their very nature, lean toward one side of the political spectrum or the other? Shouldn’t AI’s personal opinions be a topic of reflection? Creating an AI that remains at the center could be a way to minimize the impact of bias, but is it really feasible? The challenge lies in defining what « remaining neutral » means in a context where every algorithmic decision is influenced by the human world? A debate that should not be overlooked
Finally, the question of the neutrality of artificial intelligence is essential. While some users might prefer to interact with an engaged AI, others might advocate for a balanced and impartial system. This raises essential ethical and sociopolitical issues that we must address collectively.
It has become clear that improving artificial intelligence cannot be achieved without introspection about its origins and its impact on our lives. In the era of data and increased intelligence, it is time to reconsider how we design and use artificial intelligence. To enrich this debate, feel free to read these articles: When two artificial intelligences realize they’re communicating with each other Do artificial intelligences favor Marine Tondelier?
The black market for prompts to free artificial intelligences
Exams in the age of artificial intelligence Why artificial intelligences struggle to represent a clock at 12:30
