L’generative artificial intelligence needs large computational performance: according to what he explains Qualcommto generate it you need a hybrid approach. With workloads distributed across clouds and edge devices, the future of AI goes hybrid.
Generative artificial intelligence, the future is hybrid according to Qualcomm
Generative AI (or GenAI) is a form of artificial intelligence that produces content of various types (text, images, video, music, etc.) based on prompt calls requests. As demand and processing costs for this technology increase, adopting a hybrid architecture becomes imperative distribute workloads between cloud and edge devices.
A hybrid architecture of AI allows for run and coordinate AI processes between cloud and edge devices – such as smartphones, cars, PCs and IoT devices. This not only helps in the calculation, but also allows for personalized experiences.
But the main advantage is the economic saving. In fact, it is estimated that the search based on generative AI has a cost per query 10 times higher than the traditional one. And that’s just one of the many uses of generative AI.
But in addition to reducing costs, it offers additional benefits in terms of performance, privacy, security and global scalability.
These architectures can have different offload modes for balance processing between cloud and devices, depending on factors such as model and query complexity. By performing the simplest calculations on devices and the most complex ones on the cloud, for example. But consider that smartphones already run AI models with more than a billion parameters and over 10 billion will arrive in the near future.
For this reason, the hybrid approach looks like the future of generative artificial intelligence: more information in the Qualcomm whitepaper.