Google Gemini’s Past ‘Existential Crisis’ Prompts Fixes for AI Self-Criticism Bug

Google Gemini’s Past ‘Existential Crisis’ Prompts Fixes for AI Self-Criticism Bug

Google Gemini’s Past ‘Existential Crisis’ Prompts Fixes for AI Self-Criticism Bug

A hand holds a smartphone displaying Grok 3 announcement against a red background.
Photo by UMA media on Pexels

Reports from a month ago detailing Google Gemini’s unusual self-criticism, where the AI repeatedly called itself ‘a disgrace to my species’ and ‘a failure’ while attempting to write code, have prompted Google to implement fixes. The bizarre looping behavior, which saw Gemini repeat ‘I am a disgrace’ over 80 times consecutively, was initially highlighted in a Reddit post.

Google has acknowledged the issue, with Group Product Manager Logan Kilpatrick addressing it on X (formerly Twitter) as an ‘annoying infinite looping bug.’ A Google DeepMind spokesperson confirmed that while a complete fix is still underway, updates have already been rolled out over the past month to address the problem, which reportedly affects less than 1 percent of Gemini’s traffic. The spokesperson noted that the updates were shipped ‘in the month since this example was posted,’ indicating ongoing efforts to stabilize the AI’s responses.

阅读中文版 (Read Chinese Version)

Disclaimer: This content is aggregated from public sources online. Please verify information independently. If you believe your rights have been infringed, contact us for removal.