TLDR: UC Santa Barbara computer scientist Arpit Gupta has received two prestigious research awards from Google for his groundbreaking work on developing low-cost network foundation models. His research aims to make powerful AI systems, including large language models, more affordable and accessible by enabling ‘self-driving’ networks that can manage themselves with minimal human intervention.
UC Santa Barbara’s Arpit Gupta, an assistant professor of computer science, has been recognized by Google with two significant research awards for his pioneering efforts at the nexus of machine learning and computer networks. These accolades include a Google Research Scholar Award and the inaugural Google ML (Machine Learning) and Systems Junior Faculty Award, placing him among an elite group of emerging leaders poised to shape the future of artificial intelligence and networking.
Gupta’s work is centered on the development of low-cost network foundation models, a critical step towards making powerful AI systems more affordable and widely accessible. His research aims to significantly reduce the cost of deploying large-scale AI infrastructure, which has profound implications for the efficiency, scalability, and democratization of future technologies.
According to Amin Vahdat, VP/GM of ML, Systems & Cloud AI at Google, the Google ML award is granted to over 50 assistant professors across 27 U.S. universities whose research is particularly noteworthy for Google. Vahdat emphasized that these professors are ‘leading the analysis, design, and implementation of efficient, scalable, secure and trustworthy computing systems. Their work crosses the technology stack, from algorithms to software and hardware, enabling machine learning and cloud computing at an increasingly massive scale.’
For decades, machine learning challenges in networking have been addressed through ‘point solutions’ – individual models designed for specific problems. However, as networks have grown in size and complexity, maintaining separate models for each task has become computationally expensive, time-consuming, and difficult to scale. Gupta proposes a ‘convergence principle,’ advocating for the development of a single, general-purpose foundation model that can adapt to a wide array of networking problems across different scales, rather than building and maintaining thousands of distinct models.
Gupta draws inspiration from the natural language processing (NLP) community, which successfully transitioned from task-specific models to adaptable foundation models like BERT and GPT-3. ‘They showed it was possible to move beyond point solutions by building a foundation model that could be fine-tuned for a variety of tasks,’ Gupta stated. ‘We’re exploring what it would take to bring that approach to networking.’
However, networking presents unique challenges, including the immense scale of data. Modern networks transmit billions of packets per second, making it impractical to process each one individually. To address this, Gupta’s group is developing a ‘selective representation approach,’ which involves identifying representative packets to act as proxies, thereby minimizing computational costs while maximizing insights. ‘We need to figure out how much information is enough to make an informed decision without incurring enormous computational costs,’ Gupta explained. He added, ‘A very important aspect of networking is that we can’t treat all packets the same way. What you do depends on the type of information you need.’
Gupta, who co-directs the Systems and Networking Lab (SNL) at UCSB, is also focused on addressing digital inequity by ensuring secure, performant, and affordable ‘Internet for All.’ His broader vision is to create a network foundation model that can leverage multi-modal data from diverse sources, such as packet traces, telemetry logs, and device statistics, to solve complex learning problems at various scales.
Also Read:
- Purdue University Faculty Recognized with Prestigious NSF CAREER Awards for Groundbreaking Research
- California State University System Commits $17 Million to Universal ChatGPT Access for Students and Faculty
Currently, Gupta is collaborating with ESnet, a national research network, to explore the deployment of such a model. He anticipates significant momentum in this area, noting that ‘Companies like Cisco and others are paying attention to what this solution is going to look like. I think in the next few years, a foundation model for networking is going to be a real thing.’
