Just as most organizations prefer multi-cloud options and multiple databases for different purposes, there's an artificial intelligence (AI) model for every purpose. A survey of over 1,000 IT decision makers found that the most advanced AI adopters are leveraging hundreds of models simultaneously.
We are in the era of “multi-model AI.” According to a study conducted by S&P Global Market Intelligence in partnership with Vultr, the average number of distinct AI models in operation today is 158, and within the next 12 months, this number is predicted to grow to 176.
Also, do AI tools make it easier to start a new business? 5 factors to consider
The most advanced users report using an average of 175 models, and expect that to grow 14% to 200 models within a year. Respondents with the second highest AI maturity expect their number of models to grow 18% year over year. Two-thirds (66%) of managers surveyed have built or developed their own models or are using open-source models.
There are practical reasons to deploy multiple models for different use cases. For example, the MIT report gives an example of a system that uses three models trained on language, vision, and behavioral data to help robots develop and execute plans for household, construction, and manufacturing tasks. “Each underlying model employed captures a different part of the decision-making process, working together when making decisions,” the MIT researchers say.
As Erica Dingman noted in a MovableInk post, what's emerging now is an “ensemble” approach to AI, where multiple models work simultaneously on any given output: “The difference between a single model and an ensemble model approach is like the difference between one violin and an entire orchestra,” she said.
“Each instrument brings value, but it's when multiple instruments work together that the real magic happens.” Additionally, employing diverse datasets and a “set of models that are constantly being updated and trained” can reduce or eliminate bias in AI output.
And when is the right time to invest in AI? 4 ways to help you decide
The widespread distribution and diversity of systems that support or are driven by AI models is also driving this proliferation: For example, research from S&P and Vultr shows that AI is increasingly moving to the edge.
“Distributed AI architectures, in which the edge is a key component of applications spanning an organization's infrastructure, are likely to become the new normal,” the survey authors stated. The majority of IT decision makers surveyed (85%) said it is likely or very likely that this shift will occur in their environment, with 32% considering the change “very likely.”
The study authors uncovered what they define as organizations leading the AI pack — those with “transformative AI practices.” Half of these transformers are performing “significantly better” than their operational-level peers. Nearly all transformers say they've seen year-over-year performance improvements from 2022 to 2023 in customer satisfaction (90%), revenue (91%), cost reduction/margin expansion (88%), risk (87%), marketing (89%) and market share (89%).
Plus: Agile development can unlock the power of generative AI – here’s how
According to the survey, AI spending is expected to outpace general IT spending for all organizations surveyed: Nearly nine in ten (88%) companies plan to increase their AI spending in 2025, with 49% anticipating a moderate to significant increase.
But behind the skyrocketing numbers are challenges related to demands on existing IT infrastructure. “When considering high-demand AI activities such as real-time inference, respondents expressed concern that their existing infrastructure would not be able to keep up,” the survey authors wrote. The top three concerns were insufficient CPU or GPU resources (65%), data locality issues (53%), and storage performance issues (50%).
“This is reflected in the qualitative data, with interviewees expressing concern about delays in scheduling large-capacity GPU instances in public clouds and the potential impact on data availability,” the authors wrote.
Plus: Beyond programming: AI will create a new generation of jobs
“We've also noticed growing concerns about the impact of infrastructure costs. Once a project is in production, cost often becomes a more pressing concern. Historically, organizations have had limited ability to forecast costs effectively.”
