Third-party check w/ $Apple (AAPL.US)$ employee on deploymen...

Third-party check w/ $Apple(AAPL.US)$ employee on deployment of AI. Seemed only sort of in-the-know so would discount highlights accordingly. Also $Apple(AAPL.US)$ employees are very siloed so commentary likely only applies to a single division.
Punchline is they are trying to get GenAI to power new features like app store review summaries. It has been expensive and slow-going. They sound all-in on $Alphabet-A(GOOGL.US)$ cloud for this which they find to be the most performant data lake.

Highlights
-Working on using GenAI to generate summaries of app store reviews. See lots of users go through reviews and then freeze (analysis paralysis) so would like to make this easier.
-Also want to use GenAI to create personalized app recommendations instead of just top 10 by category. Should be personalized by geography, type of person, etc.
-Mostly use $Alphabet-A(GOOGL.US)$ BigQuery for data lake – most performant. Have dedicated hardware within $Alphabet-A(GOOGL.US)$ that is encrypted.
-Run own foundation models open-sourced through $GOOGL Vertex platform. Not using closed models from Open.AI or Anthropic.
-Also use $Alphabet-A(GOOGL.US)$ GCP cloud monitoring.
-Use Pinecone for vector db
-Projects have been very expensive and still work