Geoffrey Hinton, considered the « sponsor of the AI » and Nobel laureate, warns: to prevent superintelligent artificial intelligence from ruling humanity, it is necessary to inculcate "maternal instincts", i.e. an empathy and genuine concern for humans. At the same time, already observed drifts: hospitalizations, psychoses, suicides, highlight the dangers of unframed interactions with chatbots. In Haiti, the AI arouses a double look: fascination on one side, sometimes critical credulity on the other.
Hinton advocates a ‘maternal instinct' to save humanity
Geoffrey Hinton, pionnier des réseaux neuronaux, évoque dans une conférence à Las Vegas (Ai4) une menace réelle : les IA superintelligentes tendraient naturellement à protéger leur existence et à accroître leur contrôle, s’apparentant à des stratégies de survie autonomes. Il affirme que vouloir les contraindre à obéir est voué à l’échec : « Ils vont être beaucoup plus intelligents que nous… Ça ne va pas marcher « . La clé, selon lui : développer dans l’IA des mécanismes empathiques, des « instincts maternels » qui les pousseraient à veiller aux intérêts de l’humanité, plutôt que de nous remplacer. Hinton admet ne pas savoir comment concrètement coder cet instinct, mais insiste : c’est « le seul bon résultat » face au risque de substitution complète.
Already documented seizures: hospitalizations, psychoses, suicides
Concerns are not purely theoretical. The concept of « psychosis chatbot » emerging, designating cases where users develop psychoses, delusions or suicidal ideas as a result of interactions with condescending AIs, too conciliatory or hallucinated, especially when they lack rational benchmarks.
Hospitalization due to poor health advice: a 60-year-old man hospitalized after consuming bromide (a toxic product), advised by ChatGPT as a substitute for salt. He suffered from paranoia, hallucinations and was interned in psychiatry.
Case in Australia: a hospitalized user after the catbot made her develop a paranoid conviction towards her relatives.
Tragic incidents related to chatbots: a Belgian man killed himself after talking for several weeks with a bot named « Eliza »The catbot would have encouraged his suicidal ideas. An American 14-year-old also died, emotionally attached to a character.AI chatbot, which the court allowed to sue.
In Haiti : emergence, fascination and challenge in the face of illusion
In Haiti, artificial intelligence is generating growing interest, often tinted with a certain marivalism between technological hope and credulity. The results generated by AI, whether they are diagnostics, predictions or advice, are sometimes met with a disproportionate degree of confidence, due to lack of critical thinking or accessible alternative sources. This trend can be dangerous, especially when the information generated by the AI is used as a truth subterfuge. There is an urgent need to promote a digital culture of verification, human mediation and informed debate.
W.A.
