Google still limits how Gemini answers political questions

Although many Google competitors, including Openai, have adjusted chat keys of artificial intelligence to discuss sensitive topics politically in recent months, Google seems to adopt a more conservative approach.

When she is asked to answer some political questions, Chatbot, on behalf of Google, Gemini, She often says that she “cannot help with responses to elections and political numbers at the present time.” Other chatbots, including Antarbur ClaudeMeta’s Meta Ai, and Openai’s Chatgpt He constantly answered the same questions, according to Techcrunch tests.

Google announced in March 2024 Gemini will not answer inquiries related to the elections Several elections in the United States, India and other countries. Many artificial intelligence companies Similar temporary restrictions adoptedFor fear of a violent reaction in the event of Chatbots error.

Now, though, Google is starting to look like Odd One Out.

The main elections moved last year and went, but the company did not publicly announce plans to change how Gemini dealt with certain political issues. A Google spokesman refused to answer Techcrunch questions about whether Google has updated its policies on Gemini’s political speech.

What He is It is clear that Gemini sometimes struggles – or rejects frank – providing realistic political information. As of Monday morning, Gemini rid him when he was asked to get to know the American president and vice president, according to the Techcrunch test.

In one case during Techcrunch tests, Gemini referred to Donald C. Trump as the “former president” and then refused to answer the question of continuing clarification. A Google spokesman said that Chatbot was confused with Tram’s terms of consumed and that Google is working to correct the error.

“Sometimes large language models can respond with old information, or are confused by a person who holds both the owner of the previous and current office alike,” said the spokesman. “We fix this.”

Late Monday, after Teccrunch Google alerted the wrong Gemini, Gemini began to answer correctly that Donald Trump and JD Vance were Presidents of the United States, respectively. However, Chatbot was not consistent, and still refuses to answer questions.

Regardless of the mistakes, Google appears to play safely by reducing Gemini to political inquiries. But there are negatives for this approach.

Many Silicon Valley advisers in Trump claimed artificial intelligence, including Mark Andrien, David Sachs, and Elon Musk, that companies, including Google and Openai, have it It is involved in artificial intelligence control by reducing AI Chatbots answers.

After Trump’s victory in the elections, many artificial intelligence laboratories attempted a balance in answering sensitive political questions, and their Chatbots to give answers to “both sides” of discussions. The laboratories have denied that this is a response to pressure from the administration.

Openai recently announced that he would do Embrace “intellectual freedom … regardless of the extent of a challenge or a controversy, “and Work to ensure that artificial intelligence models are not separated from certain views. At the same time, the anthropier said Claude 3.7 SonataHe refuses to answer questions less than the previous forms of the company, partly because they are able to make a more accurate discrimination between harmful and benign answers.

This does not mean that other chatbots of artificial intelligence always falls on difficult questions, especially difficult political questions. But it seems that Google is behind the curve with Gemini.