Google DeepMind Presents Research on AI That Reads Photos, Sound, and Text Together

At a glance

On October 27, 2026, Google DeepMind released a paper on 'EmbeddingGemma 2'. The research processes different kinds of information, such as text, photos, and sound, in one unified way. The article says this could change search and everyday services.

Key points

1. 1

   This is research on multimodal AI, which understands many kinds of information at once.

2. 2

   'Embedding' is a technique that turns data into a set of numbers so computers can measure how similar meanings are.

3. 3

   The article does not confirm specific figures such as model size, performance scores, or release schedules.

4. 4

   The article stresses that photos and voice contain a lot of personal information, so privacy protection and accountability standards should be set first.

Why it matters

As AI that handles photos and voice together becomes more common, everyday services such as finance, shopping, and civil complaints could change.