Google is expanding Gemini’s in-depth research mode to 40 languages

Google announced Friday that the company is expanding Gemini's latest deep search mode to 40 additional languages.

The company launched deep search mode earlier this month, allowing users of the premium Google One AI plan to unlock a sort of AI-powered search assistant. The deep feature works in a multi-step method, from creating a research plan to finding relevant information. Then, based on this information, the tool searches again to extract knowledge. After repeating this process several times, it creates a report.

Languages ​​supported by Gemini include Arabic, Bengali, Chinese, Danish, French, German, Gujarati, Hindi, Indonesian, Italian, Japanese, Kannada, Korean, Malayalam, Marathi, Polish, Portuguese, Swahili, Spanish, Tamil, Telugu, Thani, Ukrainian, and Urdu.

The challenge for Google is to find reliable sources in a particular language and then summarize them in the native language without messing up the grammar.

In a conversation with TechCrunch in early December, Gemini app engineering director HyunJeong Choe said that while the company trains the model using clean data and trusted sources, the company's AI Google present in native languages ​​like Hindi tend to have inaccuracies in summaries.

“We typically rely on native data sources and we also use Google Search on the back end to support this information. Additionally, we perform evaluations and fact-checks in the native language data before deploying the model,” Choe said.

“Reality or getting correct information is a well-known research problem for generative AI in general. Although the model already contains a lot of information in pre-training mode, we focus on training the model to use the information in the right way,” Choe said.

Jules Walter, product manager for international markets for the Gemini app, said the company has testing programs to get quality checks from a native perspective. He mentioned that the company generates data to train models. Additionally, local teams also review these datasets.

Earlier this week, TechCrunch reported that a contracting company working to improve Gemini by evaluating responses had passed along Google's guidelines that contractors were no longer allowed to ignore quick responses, regardless of their expertise.

After this report was published, a Google spokesperson said that contractors not only evaluate responses based on content, but also look at style, format and other factors.

#Google #expanding #Geminis #indepth #research #mode #languages

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top