News

alt=""

New institute combines machine learning and chemistry to tackle climate change

The Bakar Institute of Digital Materials for the Planet (BIDMaP), led by Chemistry Prof. Omar M. Yaghi, brings together CS Profs. Christian Borgs, Joseph Gonzalez, Jennifer Listgarten, Jennifer Chayes, and Kathy Yelick, along with faculty from the Department of Chemistry and Statistics, respectively, to affect climate change by combining machine learning and chemistry. The institute aims to develop a new field of machine learning for experimental science, creating algorithms and designing platforms to optimize the discovery, development, and deployment of technology. “This is what we need to accelerate discovery at a rate that will save us from the worst effects of climate change,” said Jennifer Chayes, EECS prof., associate provost for CDSS and dean of the School of Information. “BIDMaP will bring together the founder of an important new field in chemistry and the best artificial intelligence and machine learning group in the world to imagine and create a better future.”

alt=""

Pieter Abbeel interviewed as Featured ACM Member

CS Prof. Pieter Abbeel has been interviewed as a Featured ACM Member. As part of the “People of ACM” bulletin, Abbeel details the groundbreaking work that led to his 2021 ACM Prize in Computing, and the direction of the field of AI and robotics in the warehousing industry and beyond. Given the different specializations required to pursue AI, he gives the following advice to the next generation of AI researchers: “In terms of foundations, basic mathematics such as calculus, probability, linear algebra are very important, and also optimization,” said Abbeel. “Taking physics classes can be very helpful, as it teaches you the skill of abstracting real world problem settings into equations." Prof. Abbeel is the director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, in addition to Co-Founder, President, and Chief Scientist of Covariant, a Berkeley-based AI robotics company.

BAIR Climate Initiative creates partnerships to fight climate change

Berkeley Artificial Intelligence researchers are joining forces with climate experts, government agencies, and industry, as part of the new Berkeley AI Research (BAIR) Climate Initiative, a multi-disciplinary student-led hub dedicated to fighting climate change.  The effort is being led by co-founding director CS Prof. Trevor Darrel and organized by three of his graduate students, Colorado Reed (co-advised by Kurt Keutzer), who will help lead the initiative, Medhini Narasimhan, and Ritwik Gupta (co-advised by Shankar Sastry).  Their objective is to develop AI techniques that address problems with data processing, particularly involving massive data sets. To maximize the benefit to other researchers studying the same problems around the world, all work done by the initiative will be openly published and available without exclusive or proprietary licensing.  One of their first projects, “The Fate of Snow,” will be a collaboration between BAIR Climate Initiative researchers and other scientists and policy experts on the Berkeley campus, Berkeley Lab (LBNL), Meta AI (which belongs to Meta Platforms, Inc.) and the Center for Western Weather and Water Extremes. The researchers plan to apply AI methods to a multitude of openly available weather and satellite data sources to estimate how much water is in the Sierra Nevada snowpack and forecast what that will mean for streadmflow in the region.

New Sky Computing Lab aims to revolutionize the cloud industry

Sky Computing Lab, the latest 5-year collaborative research lab launched out of Berkeley EECS, aims to build a new backbone for interconnected cloud computing, a milestone that would revolutionize the industry. The lab will leverage distributed systems, programming languages, security, and machine learning to decouple the services that companies want to implement from the choice of a specific cloud, with the goal of transforming the cloud into an undifferentiated commodity, much like the Internet. Google, IBM, Intel, Samsung SDS, and VMware are among the founding sponsors of the lab. The lab's team is comprised of over 60 members, including students, staff, and EECS faculty like Alvin Cheung, Natacha Crooks, Ken Goldberg, Joseph Gonzalez, Joe Hellerstein, Mike Jordan,  Anthony Joseph, Raluca Ada Popa, Koushik Sen, Scott Shenker, and Dawn Song. CS Prof. Ion Stoica, who will lead the lab, says “Sky will knock out current barriers and accelerate the transition to the cloud, which will accelerate the progress across different fields.”

 

Pravin Varaiya wins 2022 IEEE Simon Ramo Medal

EECS Prof. Emeritus and alumnus Pravin Varaiya (Ph.D. 1966, advisor: Lotfi Zadeh), who is currently a Professor in the Graduate School, has won the 2022 IEEE Simon Ramo Medal.  This major IEEE Corporate Award recognizes "exceptional achievement in systems engineering and systems science." Varaiya, who is known for his contributions to stochastic control, hybrid systems and the unification of theories of control and computation, was cited “for seminal contributions to the engineering, analysis, and design of complex energy, transportation, and communication systems.”

New AI system allows legged robots to navigate unfamiliar terrain in real time

A new AI system, Rapid Motor Adaptation (RMA), enhances the ability of legged robots, without prior experience or calibration, to adapt to, and traverse, unfamiliar terrain in real time.  A test robot figured out how to walk on sand, mud, and tall grass, as well as piles of dirt, pebbles, and cement, in fractions of a second.  The project is part of an industry-academic collaboration with the Facebook AI Research (FAIR) group and the Berkeley AI Research (BAIR) lab that includes CS Prof. Jitendra Malik as Principal Investigator, his grad student Ashish Kumar as lead author, and alumnus Deepak Pathak (Ph.D. 2019, advisors: Trevor Darrell and Alexei Efros), now an assistant professor at Carnegie Mellon, among others.  RMA combines a base policy algorithm that uses reinforcement learning to teach the robot how to control its body, with an adaptation module that teaches the robot how to react based on how its body moves when it interacts with a new environment.  “Computer simulations are unlikely to capture everything,” said Kumar. “Our RMA-enabled robot shows strong adaptation performance to previously unseen environments and learns this adaptation entirely by interacting with its surroundings and learning from experience. That is new.”  RMA's base policy and adaptation module run asynchronously and at different frequencies so that it can operate reliably on a small onboard computer.  

Cloud startup Databricks raises $1 billion in Series G funding

Databricks, a cloud startup founded by CS Adjunct Assistant Prof. Ali Ghodsi, CS Prof. Scott Shenker, CS Prof. Ion Stoica, and alumni Andrew Konwinski (M.S. '09/Ph.D. 12, advisor: Randy Katz), Reynold Xin (Ph.D. '13, advisor: Ion Stoica), Patrick Wendell (M.S. '13, advisor: Ion Stoica), and Matei Zaharia (Ph.D. '13, advisors: Scott Shenker & Ion Stoica), has received $1 billion in a Series G funding round.  Franklin Templeton led the round and now values the company at $28 billion.  Amazon Web Services, CapitalG, the growth equity arm of Google parent Alphabet, and Salesforce Ventures are backing Databricks for the first time, while Microsoft joins a group of existing investors including BlackRock, Coatue, T. Rowe Price and Tiger Global.  Ghodsi, who is CEO of the company, says Databricks plans to use the funds to accelerate its international presence. “This lets us really hit the gas and go aggressive in these big markets. It’s almost like starting the company all over again,” he says.  Databricks grew out of the AMPLab project and is built on top of Apache Spark, an open-source analytics tool developed at Berkeley.  The company provides data analytics and AI tools to businesses.  It has grown more than 75% year-over-year, with the majority of its revenue coming from enterprises like Comcast, Credit Suisse, Starbucks and T-Mobile, who use it as a "data lake house"--a place to store structured and unstructured data, then layer business intelligence or machine-learning tools easily on top.

Ambidextrous wins SVR 'Good Robot' Excellence Award

Ambidextrous, a company co-founded in 2018 by CS Prof. Ken Goldberg, his graduate student Jeffrey Mahler (CS Ph.D. '18), and AutoLab postdocs (and ME alumni) Stephen McKinley (M.S. '14/Ph.D. '16) and David Gealy (B.S. '15), has won the inaugural Silicon Valley Robotics (SVR) ‘Good Robot’ Innovation and Overall Excellence Industry Award.  Ambidextrous utilizes an AI-enhanced operating system, Dexterity Network (Dex-Net) 4.0, that empowers versatile robots for automated e-commerce order fulfillment by allowing them to learn to pick, scan, and pack a wide variety of items in just a few hours.  This universal picking (UP) technology has enabled new levels of robotic flexibility, reliability, and accuracy.

Deep learning helps robots grasp and move objects with ease

CS Prof. Ken Goldberg is the co-author of a study published in Science Robotics which describes the creation of a new artificial intelligence software that gives robots the speed and skill to grasp and smoothly move objects, making it feasible for them to soon assist humans in warehouse environments.  He and postdoc Jeffrey Ichnowski had previously created a Grasp-Optimized Motion Planner that could compute both how a robot should pick up an object and how it should move to transfer the object from one location to another, but the motions it generated were jerky.  Then they, along with EECS graduate student Yahav Avigal and undergraduate (3rd year MS) student Vishal Satish, integrated a deep learning neural network into the motion planner, cutting the average computation time from 29 seconds to 80 milliseconds, or less than one-tenth of a second.  Goldberg predicts that, with this and other advances in robotic technology, robots could be assisting in warehouse environments in the next few years.

Mike Stonebraker wins 2020 C&C Prize

EECS Prof. Emeritus Michael Stonebraker has won the prestigious NEC Computers and Communications (C&C) Prize "For Pioneering Contributions to Relational Database Systems." The prize is awarded "to distinguished persons in recognition of outstanding contributions to research and development and/or pioneering work in the fields of semiconductors, computers, and/or telecommunications and in their integrated technologies."  In the early 1970's, Stonebraker and Prof. Eugene Wong began researching Relational Database Management Systems (RDBMS), which culminated in the creation of the Interactive Graphics and Retrieval System (INGRES), a practical and efficient implementation of the relational model running on Unix-based DEC machines.  It included a number of key ideas still widely used today, including B-trees, primary-copy replication, the query rewrite approach to views and integrity constraints, and the idea of rules/triggers for integrity checking in an RDBMS.  Stonebraker, Wong, and Prof. Larry Rowe, founded a startup called Relational Technology, Inc. (renamed Ingres Corporation), which they sold to Computer Associates in the early 1990's for $311M.  Stonebraker's student, Robert Epstein (Ph.D. '80), founded the startup Sybase, which created the code used as a basis for the Microsoft SQL Server.  Stonebraker also created Postgres in the late 1980's, which made it easier for programmers to modify or add to the optimizer, query language, runtime, and indexing frameworks.  It broadened the commercial database market by improving both database programmability and performance, making it possible to push large portions of a number of applications inside the database, including geographic information systems and time series processing.  Stonebraker retired from Berkeley in 2000 to found more companies and become an adjunct professor at MIT.  His achievements have been recognized with an IEEE John von Neumann Medal in 2005, ACM A.M. Turing Award in 2014, and ACM SIGMOD Systems Award in 2015.