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

Google said Friday that the company is expanding Gemini’s latest in-depth search mode to include 40 more languages.

The company launched In-depth research method earlier this month, allowing users of Google’s One AI premium plan to unlock an AI-powered search assistant of sorts. The in-depth functionality works in a multi-step manner, from creating a research plan to finding relevant information. Then, based on that information, the tool performs a search again to extract the knowledge. After repeating this process several times, a report is generated.

Languages ​​supported in 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, then summarize them in the native language without messing with the rules.

In a conversation with TechCrunch in early December, HyunJeong Choe, director of engineering for the Gemini app, said that while the company trains the model using clean data and trustworthy sources, Google’s AI overviews in… Native languages ​​like Hindi tend to be imprecise in summaries.

“We generally rely on local sources of data, and we also use Google search on the back end to verify that information. Additionally, we conduct assessments and audits of native language data before rolling out the model.

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

Jules Walter, product lead for international markets for Gemini, said the company has testing programs to obtain quality checks from local perspectives. He stated that the company generates data to train models. In addition, local teams review these datasets as well.

Earlier this week, TechCrunch reported that a contract company working to improve Gemini by evaluating responses passed along guidelines from Google stating that… Contractors are no longer allowed to skip rapid responses, regardless of their experience.

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

TechCrunch has an AI-focused newsletter! Register here Get it in your inbox every Wednesday.