Berkeley launches RISELab, enabling computers to make intelligent real-time decisions
UC Berkeley today launched the RISELab, the successor of AMPLab, and the latest in its series of five-year intensive research labs in computer science, with the goal of improving how machines make intelligent decisions based on real-time input.
The new Berkeley lab focuses on the development of data-intensive systems that provide Real-Time Intelligence with Secure Execution (RISE). The lab kicks off this January with support from founding sponsors Amazon Web Services, Ant Financial, Capital One, Ericsson, GE Digital, Google, Huawei, Intel, IBM, Microsoft, and VMWare.
RISELab principal investigator and director Prof. Ion Stoica describes the lab’s mission as “tackling a long-standing grand challenge in computing: to enable machines to rapidly take intelligent actions based on real-time data and context from the world around them. This technology has applications wherever computing decisions need to interact with the world in real time, ranging from earthquake warning systems, to coordinating fleets of self-driving cars and drones, to cybersecurity and real-time financial services.”
The RISELab follows on from the success of Berkeley’s AMPLab, a pioneering Big Data effort, which launched widely-used open source projects including Apache Spark, Apache Mesos and Alluxio. “Data science is a critical area of expertise for the future of organizations across all industries. From its work with SQL to Apache Spark, Berkeley’s AMPLab has been driving innovation in the area of data science,” said Dinesh Nirmal, Vice President, Analytics Development, IBM. “We are looking forward to working with UC Berkeley RISELab to further advance data science technology for the next generation workforce.”
Like much of the Big Data movement, the AMPLab focused mostly on offline data analysis problems, where minutes and hours could be devoted to extracting value from data. By contrast, the RISELab researchers are looking to make real-time decisions in milliseconds. “The RISE challenge is to allow machines to make decisions in a tight loop with the real world,” said lab principal investigator Prof. Joseph Gonzalez, an expert in Machine Learning systems. “So we need systems that can both understand the big context, and continuously adapt their beliefs and make robust decisions in real time.” Jeff Dean, Senior Fellow at Google, adds that “the ability to train and deploy large-scale machine learning systems that can learn and react to observations about their environment in real time, and to ensure that they do so in a safe and predictable manner is important, which is why we are very excited about the research charter of the RISELab.”
Some of the most impressive practical advancements in AI have been thanks to the access to massive data. However, today more and more data is collected by disparate organizations that have to manage complex relationships that are both cooperative and competitive. New techniques are needed to enable this “coopetition” in a data-rich way. “As part of RISELab, we will develop new Machine Learning algorithms that incentivize organizations to share their data by getting better service in exchange, and by guaranteeing their data confidentiality” said Michael Jordan, world-renowned Machine Learning expert, and a leader in the lab. Raghu Ramakrishnan, CTO for Data at Microsoft, agrees “Enabling learning on shared data is key to enable powerful ML based services to enterprises that cannot share their data due to regulations or competitive concerns, and holds the potential to open entirely new business models.”
Prof. Joseph Hellerstein, a database systems veteran and lab principal investigator, elaborated on the shift in technology trends. “Ten years ago, a huge acceleration in data growth ushered in the era of Big Data and practical machine learning. The next commoditization shock is underway with the proliferation of data-centric devices. Billions of networked sensors—in cellphones, cameras, cars and buildings—instrument the world around us. We’re seeing rapid growth in programmable devices that can take action as well. The next big challenge in data-centric computing is to make it easy to close the loop between sensing and acting, via new platforms for real-time intelligent decision-making.”
With the shift from offline data analysis to real-time decision-making, security issues become even more critical. “RISE systems have to be architected with security in mind from the grounds up,” said lab principal investigator Prof. Raluca Ada Popa, an expert in computer security. “We need to make it possible for data and code to be protected everywhere they are used, whether the computation runs on an edge device, a corporate data center, or in public cloud computing infrastructure.” Hardware capabilities such as Intel® Software Guard Extensions (Intel® SGX) are key to enabling these security properties. “We are looking forward to collaborating with RISELab to understand the security demands of real-time intelligent applications making decisions on live data”, said Sridhar Iyengar, vice president in Intel Labs and a founding sponsor of RISELab. “We are interested in how our Intel® SGX technology can better support such applications.”
Stoica described why the RISELab plans to build systems to power a wide range of applications: “What’s missing from the ecosystem today is a general-purpose platform that can foster widespread innovation. In the same way that Hadoop and Spark opened up the Big Data space to developers, we want the systems we build in the RISELab to help a wide range of software developers build innovative real-time applications.”
This broad outlook resonates across the industries represented by the founding sponsors. David Tennenhouse, VMware’s Chief Research Officer, sees relevance for a broad range of businesses: “Real-time predictions and decisions open new possibilities for optimizing processes and pipelines in enterprise computing, and we look forward to collaborating with RISELab to make this promise a reality”. Diane Lye, Senior Vice President, Enterprise Data Services, Capital One, notes that “in the financial services industry, we are increasingly moving toward a future where banking is more real-time, digital and better able to anticipate customer needs. Advances in machine learning and the ability to harness data and context in real time to help our customers with their financial lives is core to this future state, and we’re looking forward to working with the UC Berkeley RISELab to research flexible, open-source platforms that will lead to breakthrough customer experiences in the financial services industry and beyond.” Hu Xi, Chief Head of Ant Financial Infrastructure amplified on the impact for financial services: “Ant Financial’s core mission is, billions of normal people around the globe have equal and easy access to financial services. The ability of secure intelligent decision-making in real-time with low cost is extremely valuable to customer experience and business efficiency. We will work together with RISELab on several fundamental research challenges, required by innovative and smart financial services.”
The RISELab agenda spans both consumer and industrial applications. “GE is leading the digital revolution for the Industrial Internet of Things, and RISELab’s mission is nicely aligned with our vision,” said Darren Haas, Head of Predix Cloud Engineering, GE Digital. “Addressing the need for secure, real-time inference at scale will ultimately help our industrial customers capitalize on the latest advances in computing.” Martin Frojd, Head of Cloud and Data Platforms of Ericsson, elaborated on the potential for RISELab technologies to interact with advances in telecommunications: “The key workloads in the 5G era will be distributed data collection and control systems. We are excited and honored to be working with the RISELab on an intelligent, open source platform to enable a world of connected devices.”
In addition to professors Joseph Gonzalez, Joseph Hellerstein, Michael Jordan, Raluca Ada Popa, and Ion Stoica, the RISELab includes Berkeley professors Krste Asanović, Ali Ghodsi, Ken Goldberg, Randy Katz, Michael Mahoney, David Patterson, and Dawn Song.
Intel, the Intel logo and Intel® Software Guard Extensions (Intel® SGX) are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.
