Dive Brief:
- Small AI models will continue to gain traction among enterprises in the next two years, Gartner analysts anticipate, according to a report published last week.
- Gartner predicts organizations will use small, task-specific models three times more than general-purpose large language models by 2027 as leaders look to minimize compute and reduce operational and maintenance costs.
- “The variety of tasks in business workflows and the need for greater accuracy are driving the shift towards specialized models fine-tuned on specific functions or domain data,” Sumit Agarwal, VP analyst at Gartner, said in a statement.
Dive Insight:
The generative AI conversation once revolved around large language models. Despite the versatility and broad domain knowledge that once lured enterprises in, leaders are now looking for less resource-intensive options, giving way to the rise of small models.
CIOs have thus far struggled to make progress on AI initiatives due to resource constraints. More than 1 in 3 technology leaders said they’ve delayed AI projects by a minimum of three months due to computing availability, budget and skills gaps, according to a Civo report.
These specialized models can also offer a greener alternative to LLMs due to the smaller amounts of computing needed, a welcomed change for businesses with sustainability targets. But just like any technology, small models have limitations.
Enterprises will still need to work to understand when to use smaller models and how to pair them with the correct use cases, analysts told CIO Dive.
Technology leaders will also need to wade through countless options for small models, from Google’s lightweight Gemma to Microsoft’s Phi and OpenAI’s mini models.
AI providers have jumped on the bandwagon with expanded customer options for fine-tuning and customization. In these instances, an enterprise’s data becomes the key differentiator, Gartner said. Most businesses grapple with a range of data dilemmas, however, leading to increased costs, eroded trust and poor performance.
CIOs have a crucial role to play in getting enterprise data strategies ready for AI. Technology leaders should help the business assess the resiliency, strength and sustainability of existing practices to better understand where to tweak or change the approach to produce the desired results.
