Topics
Latest
AI
Amazon
Image Credits:David Paul Morris/Bloomberg / Getty Images
Apps
Biotech & Health
mood
Cloud Computing
DoC
Crypto
endeavour
EVs
Fintech
Fundraising
Gadgets
back
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
surety
Social
Space
Startups
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
picture
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Google hasreleaseda trio of new , “ heart-to-heart ” procreative AI mannequin that it ’s calling “ safer , ” “ modest ” and “ more gossamer ” than most — a bold title , to be certain .
They ’re additions to Google’sGemma 2family of generative models , which debuted back in May . The new models , Gemma 2 2B , ShieldGemma and Gemma Scope , are designed for slightly different applications and use compositor’s case , but share in common a safety bent .
Google ’s Gemma series of exemplar are different from its Gemini models in that Google does n’t make the source computer code available for Gemini , which is used by Google ’s own products as well as being available to developers . Rather , Gemma is Google ’s drive to foster goodwill within the developer community , much likeMeta is attempting to do with Llama .
Gemma 2 2B is a lightweight poser for return analyse school text that can bleed on a range of computer hardware , including laptop and bound equipment . It ’s licensed for sealed research and commercial-grade applications and can be download from rootage such as Google ’s Vertex AI model library , the data science program Kaggle and Google ’s AI Studio toolkit .
As for ShieldGemma , it ’s a assemblage of “ safety equipment classifiers ” that essay to find perniciousness like hate lecture , harassment and sexually explicit subject . Built on top of Gemma 2 , ShieldGemma can be used to filtrate prompt to a productive manikin as well as content that the model engender .
in conclusion , Gemma Scope allow developers to “ zoom in ” on specific point within a Gemma 2 model and make its interior works more interpretable . Here ’s how Google account it in a blog post : “ [ Gemma Scope is made up of ] specialized neural networks that aid us take out the thick , complex information process by Gemma 2 , expand it into a conformation that ’s easier to analyze and realise . By studying these expanded views , research worker can gain valuable insights into how Gemma 2 identifies patterns , processes information and ultimately makes prognostication . ”
The release of the new Gemma 2 models comes shortly after the U.S. Commerce Department indorse open AI models in a preliminary report . Open models broaden productive AI ’s availability to smaller party , researchers , nonprofits and single developers , the report said , while also highlighting the need for capability to monitor such models for potential risk .
