Apple gave a long explanation in a 500-word blog post: On-device deployment offers several advantages over a server-based approach. Firstly, it prioritises user privacy as all input data remains on the user's device. Secondly, it allows users to use the model without needing an internet connection post-download. Lastly, local deployment empowers developers to cut down or eliminate server-related expenses.
This development aligns with Apple’s partnerships with other companies. Apple's introduction of OctaneX, a GPU rendering feature for Apple M1 and M2, last year, has led to speculations about a potential partnership with OTOY. Octane reportedly utilises OTOY's Render (RNDR) token, known for decentralised GPU-based rendering solutions. This gains significance with Stability AI's launch of StableLM, an open-source competitor to ChatGPT.
OTOY's founder, Jules Urbach, hinted at RNDR's large GPU for AI inference, soon to be available on Apple Neural Engine. Emad Mostaque, Stability AI's founder, expressed interest in running Stable Diffusion on mobile devices. If Apple joins forces with OTOY, discussions with Mostaque may be on the cards, paving the way for intriguing developments in the AI space.
Read the full story here.
A Symbiotic AI Life
Oracle is setting ambitious targets to become a $65-billion company by 2026, and a major part of this strategy is its investment in Cohere, a generative AI startup. Unlike its competitors like OpenAI and Anthropic, Cohere is solely focused on enterprise needs, providing generative updates across Oracle's products and services.
The Cohere-Oracle partnership aims to leverage generative AI capabilities within Oracle's suite of SaaS applications, such as Oracle Fusion Cloud Applications Suite and Oracle NetSuite, catering exclusively to enterprise customers. This strategic partnership positions Oracle to excel in the generative AI landscape, targeting enterprise success.
Read the full story here.
GPUs aren’t Efficient
If $1 is spent on a GPU, another dollar goes towards the energy costs for running that GPU in a data centre. This shows the enormous energy cost of running these GPUs in data centres. As per research, AI data centre server infrastructure, along with operating costs is said to cross $76 billion by 2028.
Companies are exploring specialised data centres for generative AI workloads in suburban areas, reducing connectivity lag and costs. Additionally, innovations like NVIDIA's liquid-cooling system aim to enhance energy efficiency, and cool data centres efficiently in high-temperature conditions. Renewable energy sources and nuclear energy investments by tech giants might offer more sustainable solutions in the future.
Read the full story here.
Microsoft's Zero-Waste Goal
Комментариев нет:
Отправить комментарий
Примечание. Отправлять комментарии могут только участники этого блога.