Apple Gains Full Access To Google's On-Device AI For Smaller Models And Distillation Capabilities
Reports indicate that as part of its broader agreement with Alphabet, Apple has been granted complete access to the Gemini model within its own data centers. This partnership enables Siri and other on-device features by allowing engineers to utilize distillations techniques for creating smaller AI models optimized specifically for iPhones rather than relying solely on massive cloud-based versions previously discussed in Bloomberg reports from earlier today
Key Points
-
1Apple gained full access to the Gemini large language model within its own data centers as part of a new agreement.
-
2The deal allows Apple engineers direct control over how they modify or customize specific aspects of the underlying models for Siri and other features.
-
3Using advanced distillation techniques, Apple is converting Google's massive AI architectures into smaller versions optimized specifically on-device processing.
Developments
Perspectives
Apple can create smaller versions of the massive Gemini language model for on-device processing by cutting it down into manageable chunks through a process called 'distillation'.
— [Mar 25, 16:39] Apple is distilling Google Gemini into smaller parts to make them suitable for iPhone AI features (Appleinsider)'Complete access' was granted as part of the deal that allows both companies and their respective users full control over how they use each other's technology.
— [Mar 25, 16:39] Apple has significant freedom in using Google Gemini models to create Siri features (The Information)'Complete access' was granted as part of the deal that allows both companies and their respective users full control over how they use each other's technology.
— [Mar 25, 16:39] Apple has significant freedom in using Google Gemini models to create Siri features (The Information)'Complete access' was granted as part of the deal that allows both companies and their respective users full control over how they use each other's technology.
— [Mar 25, 16:39] Apple has significant freedom in using Google Gemini models to create Siri features (The Information)'Complete access' was granted as part of the deal that allows both companies and their respective users full control over how they use each other's technology.
— [Mar 25, 16:39] Apple has significant freedom in using Google Gemini models to create Siri features (The Information)Google has granted full access permission allowing Apple to distill its own version of the Gemini model into smaller versions optimized specifically for Siri and other on-device features without internet connectivity. This collaboration builds upon an existing multi-year partnership where Google will provide foundational models, while ensuring that core AI functions like personalized Siri remain powered by local devices or Private Cloud Compute rather than external servers.
Apple has granted full access to the Gemini model for distillation and customization within its own data centers. This allows Apple engineers to train smaller, efficient models capable of running entirely offline while adapting responses to meet specific product needs despite current limitations regarding chatbot performance tuning.
As a condition of their partnership with Alphabet/Google, Apple receives exclusive rights to run Google's Gemini AI within its proprietary infrastructure. This arrangement allows the company full access and permission for distillation techniques used in developing smaller-scale models based on this technology.