The rapid development of open artificial intelligence models is creating a new challenge for major AI companies that have invested heavily in large computing systems and expensive proprietary models.
Large technology companies have spent enormous amounts of money building data centers and developing advanced AI systems. However, smaller models and open models are increasingly being used by businesses because they can provide useful AI capabilities at a lower computing cost.
Recent research from Stanford indicates that smaller language models can handle many basic queries while using significantly less computing power and energy than the largest AI systems. The research suggests that some smaller models can reduce computing and energy requirements by as much as 80 percent for certain tasks.
Open models are also becoming easier for companies to access. Platforms such as Hugging Face provide access to a large number of open models that businesses and developers can use and adapt for different applications.
This is changing the way companies think about AI deployment. Instead of sending every task to a large external AI service, some businesses are choosing smaller systems that can be operated internally.
Data privacy is another important factor behind this trend. Companies working with sensitive information may prefer to keep AI systems within their own infrastructure instead of sending data to an external service.
Businesses including AT&T, Siemens and DoorDash have been among companies exploring open models. Financial and legal organizations are also showing interest because internal AI systems can give them greater control over sensitive information.
The growing use of smaller and open models could have wider consequences for the technology industry. Large AI companies and cloud providers have based much of their growth expectations on rising demand for expensive computing infrastructure.
If businesses can complete more AI tasks using smaller models, they may not require as much computing power for every application. This could affect the economics of some AI services and influence how companies plan future data center investments.
However, large AI models are unlikely to disappear. Advanced models remain important for complicated tasks that require greater computing power. Cloud based open models also continue to use data center infrastructure.
The bigger change could be that the AI market becomes more diverse. Instead of relying on a small number of very large models, businesses may use different systems depending on the complexity, cost and privacy requirements of each task.
The development of efficient AI models is therefore becoming an important part of the technology industry. Companies are looking for ways to achieve useful AI performance without always depending on the most expensive computing infrastructure.
As open and smaller models continue to improve, businesses will have more choices in how they build and operate AI systems. This could increase competition in the AI industry and put pressure on companies that depend heavily on extremely large models and costly infrastructure.