Hybrid AI

I was recently talking to my old friend David Terrar who runs the Cloud Industry Forum (CIF). The topic was generative AI and cloud computing, and how the two relate..

When David brought this up, my first thought was that the synergy between generative AI and cloud is pretty obvious. After all, services like Chat GPT and Claude AI were arguably ‘born in the cloud’. But then a second thought popped into my head: Just like cloud, the future of generative AI will inevitably be hybrid.

And once you start thinking like this, experience with cloud tells us there’s a huge amount to consider in terms of architectures, deployment models, distributed execution, integration, standards, and so many other things. Add in different business, compliance and workload requirements and you end up with a complex matrix of factors and dependencies that again, just like cloud, will be with us for the long-term and will be subject to continuous change.

As an example, I’ve received briefings from many IT suppliers offering different ways to combine local, private business data with the use of generic cloud based large language models (LLMs). This particular kind of hybrid approach ensures that output from generative AI reflects your specific business context and insights, not just the body of data the foundation model was originally trained on. 

As I’ve explored the various architectures that enable this and deal with other requirements, the inevitability of hybrid AI in an enterprise context has become increasingly clear to me. It will be normal for even a single application to make use of multiple models and data sets, sometimes distributed across multiple clouds, data centres, desktops and other edge devices. 

So is there synergy between generative AI and cloud computing? Well yes, but I would actually go beyond this and say the two are intimately intertwined.