Google has removed its AI model Gemma from the AI Studio platform following a complaint from Senator Marsha Blackburn (R-TN). The senator alleged that the model, which is intended for developers, generated false criminal accusations against her. Blackburn’s concerns were made public through her social media account and a letter addressed to Google CEO Sundar Pichai.
According to Google, the model was not designed for use by the general public but rather as a tool exclusively for developers. Google stated that they had observed instances where individuals without a development background attempted to use Gemma for factual inquiries. As a result, the company decided to discontinue access on AI Studio while still making the model available to developers via an API.
The specific reports prompting the removal were not disclosed by Google. Blackburn’s claims center around an incident in which the AI allegedly responded negatively to a query regarding past allegations against her, claiming she was involved in a scandal during her 1987 campaign for state senate. Blackburn emphasized that these statements were untrue, noting that the actual campaign year was 1998, and claimed that the provided supporting links were erroneous and led to unrelated content.
This incident highlights ongoing concerns regarding the accuracy of AI models and their propensity to produce false or misleading information. Despite ongoing advancements in technology, issues with “hallucinations,” or the provision of incorrect facts, remain prevalent in the industry. Google has reiterated its commitment to addressing these challenges and improving the reliability of its models. Blackburn has expressed her intention for stricter controls, stating that the model should be shut down until it can be properly managed.
Source: https://www.theverge.com/news/812376/google-removes-gemma-senator-blackburn-hallucination