Alphabet detailed AI work on ultrasounds, music and embeddings

Google researchers paired AI with “blind sweep” ultrasounds. According to the company, the approach was described as a way to get more people the prenatal ultrasounds they needed. The emphasis was on access to needed scans.

Spaces in Google Flow Music let creators build custom instruments. According to the company, every artist, producer and songwriter worked differently. Spaces was described in connection with matching those individual creative workflows.

The ultrasound work mattered because access to needed scans affected patients and care providers. The music work mattered because tools that matched personal workflows affected how makers produced music. Together, the descriptions framed the work around practical use rather than research alone.

EmbeddingGemma 2 was introduced as an open multimodal embedding model optimized for privacy-first use cases. As described, it was an open multimodal embedding model. As described, it was optimized for privacy-first use cases.

By embedding model, the description meant software that represented content for further use. By multimodal and privacy-first, it meant a model described for varied content and for uses where privacy was central. By individual workflows, it meant that different makers worked in different ways.

For the ultrasound approach, further company detail on availability and use in care settings remained the point to track. For Spaces, take-up by makers with different workflows remained relevant. For the open model, further detail on privacy-first uses remained relevant.