Intel's AI Hardware Accelerators and open ecosystem push to democratise generative AI

Generative AI has the ability to change the way we live and work, but it necessitates complicated computation. By working with industry partners to promote an open AI ecosystem, Intel hopes to make this technology available to everyone. 

Generative AI requires a lot of computing power, which makes it important for hardware manufacturers to step up in a big way. Intel, with its Deep Learning Training Processor, the Intel Habana Gaudi2 is stepping up to the challenge in a major way.

ChatGPT, a creative AI chatbot, emphasises the importance of hardware and software solutions that allow AI to realise its maximum potential. An open ecosystem enables developers to create and implement AI anywhere while balancing power, price, and speed.

Intel is optimising open-source generative AI tools and libraries to allow better performance on its hardware accelerators. Hugging Face, a leading open-source machine learning library, revealed that Intel’s Habana Gaudi2 outperformed Nvidia’s A100-80G by 20 per cent when performing inference on the 176 billion parameters BLOOMZ model. 

On the smaller 7 billion features BLOOMZ model, Gaudi2 performed three times quicker than A100-80G. Hugging Face Optimum Habana is a library that makes it easier to run big language models on Gaudi accelerators.

Furthermore, on 4th Gen Intel Xeon Scalable CPUs with built-in Intel AMX, Stability AI’s Stable Diffusion, a generative AI model for text-to-image creation, now operates 3.8 times quicker. 

This acceleration was accomplished with no code modifications, and auto-mixed accuracy using the Intel Extension for PyTorch with Bfloat16 can further decrease delay to less than 5 seconds.

Intel’s 4th Generation Xeon processors provide a long-term and energy-efficient answer for large-scale AI tasks. With built-in accelerators such as Intel AMX, these CPUs can improve inference and training performance by 10x across a variety of AI use cases, while also increasing performance-per-watt by up to 14x over the previous iteration. 

This method enables a build-once-and-deploy-everywhere plan with adaptable and open solutions.

While generative AI has the potential to greatly improve human skills, it must be developed and deployed in a human-centered and responsible manner. 

To guarantee ethical practises and minimise ethical debt, transparent AI control via an open ecosystem is required. Intel is dedicated to democratising AI by investing in technology and fostering an open environment to satisfy the compute requirements of all facets of AI, including generative AI.

Intel is betting big on AI and is pushing to democratise access to computing and tools, including big language models, in order to lower expenses and increase equity. Personalized LLMs are being created for ALS people in order to enhance communication. 

Intel promotes an open ecosystem to cultivate confidence and guarantee interoperability through a multidisciplinary strategy that focuses on amplifying human potential through human-AI cooperation and energy-efficient solutions. An open strategy is the path forward for AI.

from Firstpost Tech Latest News https://ift.tt/3YV2NDr