In contrast to several competitors, such as OpenAI, which have refined their AI chatbots to handle politically charged discussions in recent months, Google seems to be adopting a more cautious stance.
Testing by TechCrunch revealed that when prompted with specific political inquiries, Google’s AI chatbot, Gemini, frequently states that it “cannot assist with questions about elections or political figures at this time.” In contrast, other chatbots like Anthropic’s Claude, Meta’s Meta AI, and ChatGPT provided consistent responses to similar queries.
In March 2024, Google declared that Gemini would refrain from addressing election-related inquiries in the lead-up to elections in the U.S., India, and several other nations. Other AI firms imposed similar temporary restrictions, fearing potential repercussions if their chatbots provided inaccurate information.
However, Google now appears to stand out from the crowd.
With last year’s major elections having concluded, the company has yet to publicly communicate any changes regarding Gemini’s treatment of certain political discussions. A representative for Google did not provide answers to TechCrunch’s inquiries regarding updates to Gemini’s policies on political discourse.
What is evident is that Gemini occasionally falters or outright declines to provide accurate political information. As of Monday morning, Gemini hesitated when asked to identify the current U.S. president and vice president, as determined by TechCrunch’s investigations.
During one of TechCrunch’s test sessions, Gemini referred to Donald J. Trump as the “former president” and subsequently declined to respond to a follow-up question for clarification. A representative from Google explained that the chatbot was baffled by Trump’s nonconsecutive terms in office and noted that efforts are underway to rectify this issue.
“Large language models can occasionally provide outdated information or be confused when a person has held both former and current office,” the spokesperson explained via email. “We are addressing this.”
Following a notification from TechCrunch about Gemini’s incorrect answers, the chatbot accurately identified Donald Trump and J.D. Vance as the current president and vice president of the U.S., respectively. However, the chatbot still inconsistently answered other queries and at times refused to provide responses.
Despite its shortcomings, Google seems to be prioritizing caution by restricting Gemini’s responses to political inquiries. However, this strategy has its drawbacks.
Several of Trump’s advisors from Silicon Valley regarding AI, such as Marc Andreessen, David Sacks, and Elon Musk, have accused companies like Google and OpenAI of practicing AI censorship by restricting their chatbots’ responses.
After Trump’s electoral victory, many AI laboratories attempted to find a middle ground for answering politically sensitive questions, programming their chatbots to present “both sides” of various debates. These labs have denied these adjustments were influenced by external pressure from the administration.
OpenAI recently stated its commitment to uphold “intellectual freedom, irrespective of how challenging or controversial a subject may be,” while ensuring its AI models do not suppress certain viewpoints. In contrast, Anthropic mentioned that its latest AI model, Claude 3.7 Sonnet, avoids refusing questions more frequently than its previous versions, aided by its enhanced ability to differentiate between harmful and benign answers.
This is not to imply that chatbots from other AI labs consistently provide correct responses to challenging questions, especially those of a political nature. Nevertheless, Google seems to be lagging with Gemini.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
