Week 2 - Video 8 - Technical tools for AI teams (optional)

Summary

The video introduces common tools and infrastructure used by AI teams, explaining the terms so nontechnical colleagues can better understand engineers’ work.

Open-Source Machine-Learning Frameworks

Teams use tools such as TensorFlow, PyTorch, Keras, MXNet, CNTK, Caffe, Paddle, scikit-learn, R, and Weka to build AI systems more efficiently.

Research and Code Sharing

AI research is often published on arXiv, while GitHub is a major repository for open-source software. Reusing available code can speed up development, but teams should check its license before using it in a product. The speaker also encourages interested viewers to explore GitHub.

CPUs and GPUs

A CPU is a computer’s general-purpose processor. GPUs were originally designed for graphics, but their parallel-computing capabilities make them well suited to training large neural networks. NVIDIA is a major GPU provider; Qualcomm and Google, with its TPUs, are also developing specialized hardware for AI.

Cloud, On-Premises, and Edge Computing

• Cloud: Rent computing resources from providers such as AWS, Microsoft Azure, or Google Cloud Platform.
• On-premises: Buy and operate computing equipment locally within an organization.
• Edge: Process data near where it is collected—for example, inside a self-driving car or smart speaker. This can improve response time and reduce network data transfer.

The video does not compare the detailed pros and cons of these deployment options, but suggests looking up further articles on cloud versus on-premises and edge deployments.

Main speaker/source: A course instructor from SK Collectives.