Timon Harz

December 23, 2024

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

Gemini’s in-depth research mode now supports 40 more languages, allowing users to generate comprehensive reports with AI assistance. However, challenges remain in ensuring the model’s accuracy and language-specific quality.

Google announced on Friday that it is expanding the Gemini AI’s in-depth research mode to 40 additional languages. Launched earlier this month, the in-depth research mode allows users of the Google One AI premium plan to access an AI-powered research assistant. This feature works through a multi-step process, starting with the creation of a research plan, followed by a search for relevant information. The tool then performs additional searches to extract more knowledge, ultimately generating a report after several iterations.

Gemini now supports languages including Arabic, Bengali, Chinese, Danish, French, German, Gujarati, Hindi, Indonesian, Italian, Japanese, Kannada, Korean, Malayalam, Marathi, Polish, Portuguese, Swahili, Spanish, Tamil, Telugu, Thai, Ukrainian, and Urdu. However, a key challenge for Google is ensuring the accuracy of the sources in these languages and generating grammatically correct summaries.

In a December interview with TechCrunch, HyunJeong Choe, director of engineering for the Gemini app, explained that while Google trains the model using clean, trustworthy data, there are still inaccuracies in summaries, especially in languages like Hindi. “We generally rely on native sources of the data, and we also use Google search on the back end to ground that information. Additionally, we run evaluations and fact-checks in native language data before rolling out the model,” said Choe.

Choe also emphasized that ensuring factual accuracy remains a significant challenge for generative AI models. "While the model already has a lot of information from pre-training, we are focusing on training the model to use that information correctly,” she said. Jules Walter, product lead for international markets at Gemini, added that the company conducts testing programs to ensure quality checks from native perspectives, with local teams reviewing the training datasets.

Earlier this week, TechCrunch reported that a contracting firm working with Gemini had received updated guidelines from Google, instructing contractors not to skip prompt responses regardless of their expertise. In response, a Google spokesperson clarified that contractors evaluate answers for content, style, format, and other factors.

Press contact

Timon Harz

oneboardhq@outlook.com

The logo for Oneboard Blog

Discover recent post from the Oneboard team.

Notes, simplified.

Follow us

Company

About

Blog

Careers

Press

Legal

Privacy

Terms

Security