SAS, the global data and AI company, recently released its Artificial Intelligence (AI) Trends to 2025 report, which covers nine key trends in AI, including sustainability, data quality, ethical responsibility, and more.
Here are some of the report's key takeaways :
Speed and algorithmic efficiency cannot be ignored as critical levers to reduce cloud consumption. While energy-hungry AI will continue to fuel the drive toward sustainable energy sources including nuclear, it will also increase demand for more energy efficient models. Just like the home appliance industry and auto industry made huge advancements in energy efficiency, we must make AI models more efficient.
AI’s ability to personalize and operate at massive scale is reshaping how we interact with information, including the rise of misinformation and manipulation of social norms. AI attacks can happen on an individual, group or at the institutional level – threatening our ways of life. Democratic societies and their governments have a vested interest in protecting good faith civil discourse, elections and maintaining cultural norms. To help mitigate the threat, business leaders need to own the conversation on ethical use of AI within the organization by doubling down on organizational values and publishing AI principles, policies, standards and controls.
2025 will reveal some organizations are thriving with generative AI – outpacing the competition, creating specialized customer experiences, launching innovative products faster. But other organizations are falling behind in the generative AI race. They’re abandoning the wave of projects begun in 2023 because they overlooked a critical reality: AI needs good data. Poor data impedes AI performance, and organizations need to be brave enough to step back and fix their pervasive data issues.
Generative AI will never not be cool, but we’ve reached a point where we give a slight nod to the hype cycle, then get down to the business of delivering real business value. This happens by simplifying our approaches, rules and models, and complementing them with the targeted use of large language models (LLMs) and specialized small language models (SLMs). Keep a close eye on that Nvidia stock.
The rush to adopt AI is leading to inefficient models that consume vast amounts of cloud resources and contribute to a larger carbon footprint. It is not only up to hardware providers and hyperscalers to reduce environmental impact – it’s a shared responsibility with the AI users managing data and AI workloads. Greater efficiency in AI model development – made possible by cloud-optimized data and AI platforms – will help to reduce unnecessary duplication and waste and minimize energy consumption.