🌐 Gemini’s global leap
Google announced the expansion of Gemini’s in-depth research mode to 40 languages. Initially launched in early December 2024, this feature is part of the Google One AI premium plan. The tool uses a multi-step process to create detailed reports by repeatedly searching, analyzing, and summarizing data. Newly supported languages include Hindi, Arabic, Chinese, French, and Swahili. This expansion marks a significant step toward making AI tools accessible globally, but challenges with maintaining factual accuracy in summaries remain.
🧠 Training for precision
HyunJeong Choe, director of engineering for Gemini, acknowledged issues with factual accuracy in native-language summaries, especially in languages like Hindi. Choe explained that Gemini’s model relies on clean data and trustworthy sources. Google also integrates backend checks using its search capabilities. Additionally, native data evaluations aim to improve reliability. These steps highlight the complexities of training AI to deliver consistent, accurate insights in diverse languages.
🔍 Rigorous quality control
To ensure quality, Google employs testing programs with native language experts. Jules Walter, product lead for international markets, noted that local teams review datasets and provide feedback for improvement. The company also generates training data to help models learn better. Google’s stringent quality checks aim to overcome AI’s challenges in understanding cultural nuances and context, critical for accurate output across its 40 supported languages.
📜 Contractor guidelines spark debate
A recent report by TechCrunch revealed updated guidelines for contractors working on Gemini. Contractors can no longer skip evaluating responses, regardless of their expertise. Google stated these evaluations consider not just content but also style and format. This development has sparked discussions about the workload and evaluation standards in AI development. It underscores the human effort behind building robust AI tools like Gemini.
Will Gemini redefine global research?
