Google Research’s Post

Established metrics for assessing the syntactic accuracy of automatic speech recognition models miss a critical aspect of performance: comprehensibility. Learn how we use an #LLM to determine if a transcript accurately captures the intended meaning → https://goo.gle/3zJ8rdu

It’s no secret the human brain is a complex structure. But did you know we’re using #AI to map out its complicated pathways? 🧠 → https://lnkd.in/gEFz5X_W Discover the advancements our Connectomics team and Harvard University researchers are making to understand the brain.

Did you know it took more than 10 years to map the 302 neurons in a roundworm's brain by tracing the wiring by hand? Our #AI algorithms, coupled with TensorStore, an open-source C and #Python software library to store and manage massive multi-dimensional datasets, have enabled us to reconstruct bigger brains of more complex creatures. More on the latest research → https://lnkd.in/gEFz5X_W

"These AI tools enable us to study systems at a scale that wasn't possible before," says Komal Singh of Google Research, on our partnership with Geena Davis Institute. Help us create a future where everyone feels seen on (and off) screen. →https://lnkd.in/g6CMqQFV

Today on the blog, we demonstrate the feasibility & benefits of learning from human feedback for text-to-image generation by obtaining rich human feedback, then training a model to predict this feedback to enable improved image generation. Learn more →https://goo.gle/4c9i2Zz

We're back with Research@ London. 🇬🇧 The event covered a range of research projects, from PaliGemma, our open vision language model, to Med-Gemini, AI for Breast Cancer, Genomics, Health Equity LLM Toolbox, and Connectomics. Here's a big round of applause for our speakers and attendees for making this another successful event. 👏

Can AI help us understand representation in media? → https://goo.gle/4bkDr0z Google Research uses #AI to analyze massive amounts of media content, uncovering hidden patterns in representation. We partnered with the Geena Davis Institute & University of Southern California to inform a more equitable media landscape.

We believe in the importance of supporting diverse perspectives and empowering innovators who are harnessing AI to make a positive impact on the world. So we're thrilled to celebrate the 2024 Google for Startups Founders Fund recipients — a group of promising Black and Latino-led companies pushing the boundaries of what's possible with AI. From healthcare to education, these exceptional startups are tackling real-world challenges head-on. Learn more about this year's cohort below.

Introducing the Meeting Information Seeking Dialogs dataset (MISeD), which can be used to fine-tune model agents that support natural language conversations about meeting recordings, so users can catch up on meetings they may have missed. Learn more at https://goo.gle/4caXu2O