Follow
Follow
News
By Rob Enderle for TDWI
Most AI implementations are cloud-based. Hybrid AI lets you split the AI processing load. We explore the advantages and use of this technology and what’s ahead.
One of the recent topics being discussed by all the chip manufacturers and some of the hardware OEMs is hybrid AI. Every processor company is rushing AI components to market to provide for this hybrid concept, but because Apple is largely vertically integrated, it introduced its product and initial solution to market first.
In a few months, we’ll begin to see many more solutions, so let’s explore hybrid AI, why it is important, and how it could change how client devices work.
Currently, most AI implementations are cloud-based with some movement to put them on premises, but that is hybrid cloud, not hybrid AI. Hybrid AI is when you split the AI load between the central resource (be it cloud or company data center) and the device itself.
This is done for some of the same reasons hybrid cloud is done: to increase reliability and to lower latency. Although Apple’s initial use of this capability is kind of AI light, focused on Siri and enabling Siri to function when the phone is disconnected, the real value will emerge as we move generative AI to this hybrid architecture because we will increasingly rely on AI as a productivity tool, an advisory resource, and a critical operational element.
This means that the AI must be able to function if the network isn’t available or goes down. For instance, take a medical AI application supporting first responders in a catastrophe. Whether we are talking 9/11 when the communications networks (except for the old BlackBerry network) failed, or the huge Maui fire where the cellular networks went out of service, first responders dealing with burn and injury victims would be crippled if they were used to getting advanced AI support but were suddenly cut off from it.
In wars such as the current one in Ukraine, the AI must continue to function. Otherwise, a simple network outage could cause a critical effort to fail. We were given an example of this when Elon Musk allegedly cut off Starlink and crippled a major Ukrainian offensive operation using drones.
These problems would extend to autonomous cars, robots, and other machines that will increasingly need a functional AI to continue to operate, or where a lower level of latency than can be provided by a hosted AI is required.
This initial Apple effort with Siri is binary, meaning it will work from the cloud or work from the phone as this technology advances. It will become more dynamic with loads shifting between the two extremes based on the needs of the device and the bandwidth and availability of the network. Future hybrid AI efforts…
Read the full article and find out what the evolution of hybrid AI looks like: Hybrid AI and the Apple iPhone 15: A Glimpse into AI's Future
Rob Enderle, The Enderle Group
An Internet search of media quotes validates Rob Enderle as one of the most influential technology pundits in the world. Leveraging world-class IT industry analysis skills honed at DataQuest, Giga Information Group, and Forrester Research, Rob seized upon the power of the information channel as a conduit to reach business strategists and deliver valuable, experienced-based insight on how to leverage industry advances for maximum business advantage.
Interested in AI? Check here to see what TechTalk AI Impact events are happening in your area.
