With the rise of AI, it is the turn of the challenges of our humanity, raising ethical issues and redefining our relationship to intelligence.
The Turing test, imagined in 1950 by the British mathematician Alan Turing, asked a provocative question: “Can machines think?” To answer it, he proposed an experience: if a program is able to converse with a human without the latter realizing that he is talking to a machine, then the computer can be considered “intelligent” (1). This test, long remained a simple philosophical exercise, took a concrete turn with the rise of generative artificial intelligences and the quest for IAG (general artificial intelligence like human intelligence) by tech giants. But today, an inversion of the paradigm is emerging: it is no longer humans who test the intelligence of machines, but the machines that seek to verify that they interact well with humans.
A necessity linked to the proliferation of automated agents
With the rise of generative AIs like Chatgpt, Gemini or Claude, digital platforms are invaded by bots capable of emitting convincing messages, simulating opinions, or even creating digital false identities. Faced with this, AI must be able to distinguish if the interlocutor is well human, especially in critical contexts: computer security, fraud detection, moderation of content or online games.
This is where the “reverse Turing test” appears, also called Reverse Turing Test. The most famous case is that of Captchas (Completyry Automated Public Turing Test to Tell Computers and Humans Apart), these tests that we must pass to prove that we are a human, often by identifying red lights or by copying blurred letters (2). Paradoxically, it is a machine that judges our humanity here.
From behavioral detection to semantic analysis
Today, AI develops much more subtle techniques than simple capthas. In the context of cybersecurity or the moderation of forums, the AI scan user behavior models: striking rhythm, lexical variability, logical inconsistencies, or the absence of emotions typical of human conversation. The AI can analyze these clues to determine if it converses with a human or a bot.
For example, in certain instant messaging systems, AI assesses the level of Small Talk, hesitations, typographic faults or humor, as many characteristic features of a human conversation. AI can also compare the responses to large dialog databases to check if the style is generated automatically (3).
Between ethics, bias and surveillance
If this inverted test seems necessary in many contexts, it raises fundamental ethical questions. What data does AI collect to assess our humanity? Can detection algorithms be biased towards certain writing styles, or towards non-native users of a language? Does the illusion of “bot that guesses if you are human” risk becoming a generalized surveillance tool?
The paradox is that the more efficient AIs become in human simulation, the more difficult it becomes to distinguish bots … and the more the AI must be correctly equipped to detect other AI. We are then witnessing an algorithmic arms race between generators and humanity detectors.
When the machines test our reality
The reverse Turing test is not a simple logical reversal; He embodies a new border in our relationship with machines. When these evaluate our authenticity, it is also our own definition of humanity which is put to the test. Are we still able to prove that we are human, without adopting the codes that the machine expects from us?
In a world where the machine becomes a judge of our cognitive identity, it becomes urgent to rethink not if the machines may think, but how they perceive human thought and what they will do with it. Are we still the authors of our words, or simply the variables of an algorithm that observes us?
(1) Turing, AM (1950). “Computing Machinery and Intelligence.”
(2) Von Ahn, L., Blum, M., Hopper, NJ, & Langford, J. (2003). “Captcha: Using Hard Ai Problems for Security.”
(3) Hosanagar, K. (2019). “A Human’s Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control.”
