AI is the most disruptive and transformational technology trend we have witnessed. And its unique requirements impact virtually every element of the technology stack – from silicon to compute platforms to software.
While most of the focus has been on the training aspect of AI – the shaping and modeling of datasets – inference, or what to do with trained data, is where the value of this workload is realized. The platforms that perform inference will be broadly deployed because inference happens everywhere and requires real-time responsiveness.
You can download the paper by clicking on the logo below:
Table Of Contents
- Situation Analysis
- Breaking Down AI – Training Is Temporary, Inference Lasts Forever
- The Unique Requirements Of Inference
- Purpose Built Should Be Purpose Designed
- Untether AI Delivers Inference Designed For Performance, Openness, And Efficiency
- Summary
- Figure 1: Inference For Sentiment Analysis
Companies Cited
- AMD
- NVIDIA
- Untether AI