JVM, the Java Virtual Machine, is old and bad for the AI-native use. A Python UDF in Spark has to cross the JVM boundaries back and forth, and slows things down so much, no matter how you slice it, Amdahl wins.
1 verified member voice · addressed to jvm-java