From the course: Advanced AI Analytics on AWS: Amazon Bedrock, Q, SageMaker Data Wrangler, and QuickSight

Unlock this course with a free trial

Join today to access over 24,700 courses taught by industry experts.

Optimizing energy efficiency in AI analytics workloads

Optimizing energy efficiency in AI analytics workloads

From the course: Advanced AI Analytics on AWS: Amazon Bedrock, Q, SageMaker Data Wrangler, and QuickSight

Start my 1-month free trial Buy for my team

Optimizing energy efficiency in AI analytics workloads

“

- [Instructor] Energy efficiency is a very interesting problem in the world of AI and ML pipelines. And this visualization quantifies the computational inefficiency propagation in these AI and ML pipelines based on this empirical research shown in the 2017 paper. If we look at a base comparison analogy, this is like an automotive scenario, right? Where you have a 73-times efficiency difference between Rust and Python. You could think of a Chevy Suburban, for example, a 1968 Chevy Suburban that has, you know, 10 miles per gallon. And this is like Python, which is a high-interpretive overhead. So it's really a language designed for, you know, experimentation and interactive interpreter, not for runtime. If we look at a Prius Prime, this has the equivalent of 730 miles per gallon. And so this is a zero-cost abstraction, right? It's a hybrid car that has extremely low mileage, and it's a futuristic design. So both paradigms can be optimized, but you still have architectural constraints…

Contents