While AI models in large languages are still in the headlines, models in small languages are where the action is. At least Meta seems to be betting on it, according to a recently published paper by a group of researchers.
Large language models like ChatGPT, Gemini and Llama can use billions, even trillions of parameters to produce results. The size of these models makes them too large to use on mobile devices. Meta-researchers found in their study that there is a growing need for efficient large language models on mobile devices, a need driven by cloud costs and latency issues.
In their study, the researchers explained how they created high-quality, large language models with fewer than a billion parameters that they believe are suitable for mobile use.
Contrary to popular belief, which emphasizes the central role of data quantity and parameters in determining model quality, the researchers achieved results with their small language model that were in some areas comparable to Meta Llama’s LLM.
“The prevailing paradigm is ‘bigger is better,’ but this shows that it’s really about how the parameters are used,” said Nick DeGiacomo, CEO of AI-powered electronic supply chain platform Bucephalus. new York.
“This paves the way for wider adoption of artificial intelligence in devices,” he told TechNewsWorld.
Decisive Step
Meta’s research is significant because it challenges the current norm of cloud-based AI, which often sees data crunched in remote data centers, explains Darian Shimy, CEO and founder of FutureFund. venture capital in San Francisco.
“By bringing AI processing into the device itself, Meta is changing the script—potentially reducing the carbon footprint associated with data transmission and processing in massive, energy-guzzling data centers and making the tech ecosystem the primary vehicle for device-based AI,” he said. to TechNewsWorld.
“This research is the first comprehensive and publicly shared project of this scale,” added Yashin Manraj, CEO of Eagle Point, Ore.-based end-to-end security developer Pvotal Technologies.
“This is a critical first step in achieving a unified SLM-LLM approach where developers can find the right balance between cloud and on-device computing,” he told TechNewsWorld. “This sets the stage where the promise of AI-based applications can reach the levels of support, automation and assistance that have been marketed in recent years but lacked the technical capability to support these visions.”
Meta-researchers have also taken an important step to reduce the language model. “They’re coming up with a design that shrinks the size, making it easy to use with laptops, headsets and cell phones,” said Nishant Neekhra, senior director of mobile marketing at Skyworks Solutions, an Irvine, Calif.-based semiconductor company. .
“They introduce entirely new applications of artificial intelligence while providing new ways for AI to interact in the real world,” he told TechNewsWorld. “By getting smaller, they also solve a big growth problem for LLMs, which is their ability to deploy at the edges.”
Big impact on healthcare
One area where small language patterns can have a big impact is medicine.
“Research promises to unlock the potential of generative artificial intelligence in mobile applications that are ubiquitous in today’s healthcare environment for remote monitoring and biometric assessments,” Danielle Kelvas, MD, IT Medicine, Global Medical Advisor. software development company, TechNewsWorld said.
By showing that efficient SLM models can have fewer than a billion parameters and still perform comparably to larger models for certain tasks, researchers are opening the door to pervasive artificial intelligence in everyday health monitoring and personalized patient care.
Kelvas explained that the use of SLMs can also ensure that sensitive health information can be securely processed on the device, thereby improving patient privacy. They can also facilitate real-time health monitoring and intervention, which is critical for patients with chronic conditions or those who require ongoing care.
He added that the models could also reduce the technical and financial barriers to adopting AI in healthcare, potentially democratizing advanced health monitoring technologies to the wider population.
Reflects Industry Trends
Meta’s focus on small AI models for mobile devices reflects a broader industry trend to optimize AI for efficiency and accessibility, explains Caridad Muñoz, professor of new media technology at CUNY LaGuardia Community College. “This change is not just about practical challenges, but also aligns with growing concerns about the environmental impact of large-scale AI operations,” he told TechNewsWorld.
“By keeping models smaller and more efficient, Meta sets a precedent for sustainable and inclusive AI development,” Muñoz added.
Small language models also fit the extreme computing trend, which focuses on bringing AI capabilities closer to users. “Large language models from OpenAI, Anthropic and others are often overwhelming—if you just have a hammer, everything looks like a nail,” DeGiacomo said.
“Specialized, customized models can be more efficient and cost-effective for certain tasks,” he noted, “Many mobile apps don’t need cutting-edge AI. You don’t need a supercomputer to send a text message. .”
“This approach allows the device to focus on routing between SLM responders and special use cases, similar to a relationship GP specialist,” he added. .
Profound Impact on Global Connectivity
Shimy argued that the impact of SLMs on global connectivity is profound.
“As device AI becomes more powerful, the need for a constant Internet connection will decrease, which could dramatically change the technology landscape in places where Internet access is patchy or expensive,” he noted. “It could democratize access to advanced technologies and make cutting-edge AI tools available in many global markets.”
While Meta is leading the development of SLMs, Manraj noted that developing countries are aggressively monitoring the situation to keep AI development costs under control. “China, Russia and Iran seem to be strongly interested in the possibility of delay computing with local hardware, especially when advanced AI hardware chips are banned or not readily available,” he said.
“However, we do not expect this to be a sudden or drastic change,” he predicted, “as complex and multilingual surveys will continue to require cloud-based LLMs to deliver higher value to end users. However, this move to enable a “last mile” model in the facility could help reduce the burden on LLMs for smaller tasks, reduce feedback loops and enrich local knowledge, AI Language Models.
“Ultimately,” he continued, “the end user is clearly the winner, as this would allow a new generation of capabilities to appear on their devices and a more promising reconfiguration of user interface applications and how people interact with the world. .
“While the usual suspects are driving innovation in this area, with a promising potential impact on everyone’s daily life,” he added, “SLM can also be a Trojan horse that brings a new level of complexity to our daily lives. lives live with models capable of collecting data and metadata at an unprecedented level. We hope that with proper safeguards we can bring these efforts to a productive outcome..
