When a government chatbot runs on private AI, who controls what counts as the official answer?
The new America.gov chatbot puts Google’s Gemini and Elon Musk’s Grok between citizens and 29,000 federal websites, and on launch day its answers changed mid-event with no public record of who changed them or why. If an AI becomes the front door to government, what should citizens be able to see about how its answers are set, and who should be allowed to adjust them?
Commentaires (1)
The September 30 episode of Minds, Bodies, and Terawatts looked at the launch of America.gov, the federal chatbot that, according to AP, CNN and the New York Times, told reporters Biden won the 2020 election and that prices have not fallen, then began declining ‘political questions’ while the president was still on stage. The episode’s argument is that the accuracy of any single answer matters less than the dial behind it: someone changed what the machine would say within minutes, and nobody outside could see who, how, or under what rule. It also raises the question of what it means that the models, the contract terms and the cost sit with two private companies rather than in the public record. A useful comparison from the episode is line-calling technology in sports, which people trust because everyone can see the call, yet which never gets to write the rules. Give the episode a listen and tell us where you would draw the line between fixing a chatbot’s mistakes and quietly editing the official record.
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
