While gen AI’s promise is immense, the associated costs of procuring, training and fine-tuning LLMs can be astronomical, with some leading models costing nearly $200 million to train before launch. This does not include the cost of aligning for the specific requirements or data of a given organization, which typically requires data scientists or highly-specialized developers. No matter the model selected for a given application, alignment is still required to bring it in-line with company-specific data and processes, making efficiency and agility key for AI in actual production environments.
Red Hat believes that over the next decade, smaller, more efficient and built-to-purpose AI models will form a substantial mix of the enterprise IT stack, alongside cloud-native applications. But to achieve this, gen AI needs to be more accessible and available, from its costs to its contributors to where it can run across the hybrid cloud. For decades, open source communities have helped solve similar challenges for complex software problems through contributions from diverse groups of users; a similar approach can lower the barriers to effectively embracing gen AI.
Source: redhat.com