News

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.  

NLP team helps a computer win the 2021 American Crossword Puzzle Tournament

A team at the Berkeley Natural Language Processing Group (NLP) helped augment an AI system named "Dr. Fill" that has won the 2021 American Crossword Puzzle Tournament (ACPT).  This is the first time in the contest's history that an AI has trumped its human competitors.  The team, which included CS Prof. Dan Klein, graduate students Nicholas Tomlin, Eric Wallace, and Kevin Yang, and undergraduate students Albert Xu and Eshaan Pathak, approached Matthew Ginsberg, who created the Dr. Fill algorithm in 2012, and offered to join forces by contributing their machine learning system  called the Berkeley Crossword Solver (BCS).  BCS employs a neural network model to combine general language understanding with more "creative" crossword puzzle clues, then applies its knowledge to practice puzzles, improving as it learns.  “We had a state-of-the-art natural language understanding and question-answering component but a pretty basic crossword handler, while Matt had the best crossword system around and a bunch of domain expertise, so it was natural to join forces,” said Klein. “As we talked, we realized that our systems were designed in a way that made it very easy to interoperate because they both speak the language of probabilities.”  ACPT is the oldest and biggest tournament of its kind, consisting of seven qualifying puzzles and a final playoff puzzle; solvers are ranked using a formula that balances accuracy and speed. Although Dr. Fill made three errors, it completed most puzzles in well under a minute, and ultimately outscored its top human competitor, who made zero errors, by 15 points.  The contest was held online this year and attracted more than 1,100 contestants vying for the $3K grand prize. 

EECS Faculty votes to drop GRE requirement indefinitely

After intensive debate spanning 2020 and 2021, and careful analysis of a trial cycle of GRE-free admissions for Fall 2021, the EECS Department has voted to drop the GRE requirement for graduate admissions indefinitely. Effective immediately, and beginning with the Fall 2022 cohort, whose application window opens in September 2021, the application requirements for all graduate research degree programs in EECS will neither require, nor accept, GRE scores.

In 2020, at the onset of the pandemic, the EECS faculty temporarily suspended the GRE requirement for graduate admissions for the 2020-21 cycle, i.e., for those admitted for Fall 2021, primarily due to the challenges posed by COVID. The department subsequently observed a 30% increase in applications from groups historically underrepresented in EECS, a 47% increase in admittance of those applicants, and a 150% increase in yield from those populations. Not only did we attract and admit more high-performing underrepresented students, but a higher percentage of those admitted decided to attend UC Berkeley to study EECS.

The graduate admissions process in EECS is a holistic review involving the following factors: transcripts, letters of recommendation, personal statements, statements about intended research, publications (if any), and for applicants evaluated favorably on these factors, one or more phone conversations with EECS faculty.  Since applicants come from a wide range of socioeconomic and educational backgrounds, we also consider the applicant's demonstrated ability and motivation taken in the context of the opportunities they had available. Given this thorough, multifaceted review, the majority of EECS faculty concluded, after extensive discussion, that the GRE does not add much value, relative to the harm it does to diversity and equity. 

Diversity in science, technology, engineering and mathematics (STEM) fields is a longstanding challenge. For example, nationally, fewer than 22% of computer science PhD degrees are awarded to women students, and only 4% to Black students. GRE scores show significant gender and race-based differences, but these differences do not correlate with later success in graduate school, much less with undergraduate grade point average (GPA) in many cases. Therefore, using GRE scores as a “cutoff” disadvantages women and underrepresented minorities applying to graduate programs. The UC Regents recently voted to drop ACT/SAT scores from undergraduate admissions for the UC system for similar reasons.

For these reasons, along with the financial burden GRE testing fees place on economically disadvantaged applicants across the globe, the EECS Department has concluded that the GRE score has limited benefit in evaluating PhD and masters degree applicants, and that the exam itself, as well as the administration of it, harms diversity and equity.

For more information about Berkeley EECS graduate admissions, please visit our website: 

https://eecs.berkeley.edu/academics/graduate/research-programs/admissions

Gloria Tumushabe cultivates women coders in Africa

EECS alumna and current Master's student Gloria Tumushabe (B.S. ’20) is the subject of an article in the Spring 2021 Berkeley Engineer titled "Cultivating female coders in Africa."  During the COVID pandemic shutdown, Tumushabe developed a program called Afro Fem Coders to allow her to remotely teach computer programming to girls in Uganda from her home in Walnut Creek.  Two weeks after reaching out by word-of-mouth and social media, she had heard back from more than 40 girls who were eager to participate.  She sent them money to pay for laptops and internet service, and formed an international network of women professionals to provide one-on-one mentoring.  In the year since the program began, it has grown to 120 girls from Uganda, Kenya, Tanzania, Namibia, Botswana and Ethiopia. “The more of us women in this space, the better,” she said.  Tumushabe is leading the EECS Anti-Racism Committee meetings this semester, and was awarded the 2021 EECS Eugene L. Lawler Prize for her "amazing work and dedication to diversity, equity and inclusion, and improving the EECS Department for students who come after her."

Leyla Kabuli wins 2021 University Medal

Senior undergraduate and future graduate EECS student Leyla Kabuli has won the University Medal, UC Berkeley's highest honor.  She is the daughter of EECS alumna A. Nazli Gündes (Ph.D. ’88, advisor: Charles Desoer), now an ECE professor at UC Davis.  Kabuli, who will graduate with a 4.0 GPA, attended Berkeley on a prestigious Regents' and Chancellor's scholarship, and earned simultaneous degrees in EECS and Music.  Her research interests lie in diagnostic imaging, vision and perception, and are focused on super-resolution microscopy and magnetic particle imaging.  Her other honors include a Jacobs Institute Innovation Catalysts Ignite Grant, an Outstanding Graduate Student Instructor Award, a Samuel Silver Memorial Scholarship Award, an Edward Frank Kraft Award for Freshmen, and a California Seal of Biliteracy in French and Turkish. The University Medal recognizes a graduating student’s outstanding research, public service and strength of character.  She will be funding her graduate education with a Berkeley Fellowship for Graduate Study, as well as a National Science Foundation fellowship for outstanding graduate students in STEM fields.  Kabuli was offered full graduate fellowships to attend Stanford and MIT but chose Berkeley because “I might be biased, but Berkeley has the best electrical engineering program in the country,” she said.

Jiaheng Zhang wins 2021 Facebook Fellowship for Security & Privacy

Third-year EECS graduate student Jiaheng Zhang (advisor: Dawn Song) has won a 2021 Facebook Fellowship for Security & Privacy.   He is the only student from Berkeley this year to win one of these coveted fellowships, which are designed to support emerging scholars who are engaged in innovative research.  Zhang's focus is on computer security and cryptography, especially zero-knowledge proofs and their applications on blockchain and machine learning models.  He is a member of the RISE Lab, the Initiative for Cryptocurrencies & Contracts Lab (IC3), and the Berkeley AI Research (BAIR). 

Berkeley Blue team takes silver medal at ACM programming championship

The Berkeley Blue team, which includes EECS undergraduates Ethan Guo and James Shi, and CS/Math undergraduate Justin Yokota, has won a silver medal at the 2020 ACM International Collegiate Programming Contest (ICPC) North America West Division Championship.  If the team does well in the North American Division (NADC) Championship this August, they will be eligible to compete in the the world’s most prestigious competition of young talents in the field of IT, the 2022 ICPC World Finals, which will be held in Moscow in 2022.   UCSD placed first, followed by Berkeley Blue, and teams from UCLA, UWash, Stanford, UBC, and the Berkeley Gold team, which includes students Ajit Kadaveru,  Samuel Lee, and Jonathan Guo.

Ranade, Shrivastava, Monga, Yang, Rampure and Shen win Extraordinary Teaching in Extraordinary Times Awards

EECS alumna and Assistant Teaching Prof. Gireeja Ranade (M.S. '09/Ph.D. '14, advisor: Anant Sahai), and Graduate Student Instructors (GSIs) Ritika Shrivastava, Jay Monga, Maxson Yang, Suraj Rampure and Allen Shen have won UC Berkeley Extraordinary Teaching in Extraordinary Times awards.  They are among 59 people from of pool of over 500 nominees honored at Berkeley by the Academic Senate’s Committee on Teaching for embracing the challenges posed by the 2020 COVID-19 pandemic, and engaging in or supporting excellent teaching. "These instructors and staff used innovative methods and worked beyond their traditional roles to ensure that students remained engaged and supported, and were challenged to do meaningful work under extraordinary circumstances."

Shrivastava, a fall GSI for EECS C106A/206A Introduction to Robotics, provided a warm, supportive, and positive environment for her students, developed new materials, and used tools to promote inclusiveness and overcome technological differences.  Jay Monga, also a fall GSI and lab TA for EECS 106A/206A, helped students with their lab-focused robotics class by creating a video walkthrough and slides demonstrating procedures and assignments, recording a presentation to promote asynchronous instruction, helping to design a more accessible lab, and creating a Discord server for better virtual learning. Yang, who was a summer GSI for CS 10 The Beauty and Joy of Computing,  released a comprehensive student survey to guide course policy and focused on  reducing common stressors (like deadlines), implementing weekly check-ins, and creating ways to improve the students' virtual experience (like memes).  Rampure, who was a fall GSI and summer instructor for Data C100 Principals & Techniques of Data Science, and Shen, who was a fall GSI and summer instructor for CS 186 Introduction to Data Systems, won the award together for teaching two of Berkeley’s flagship undergraduate data science courses.  They introduced new applications of course material, prioritized accessibility in lectures, designed assessments, and used real-world examples to promote engagement. 

Wenshuo Guo wins 2021 Google PhD Fellowship

EECS graduate student Wenshuo Guo (advisor: Michael I. Jordan) has won a 2021 Google PhD Fellowship in Algorithms, Optimization and Markets.  This award acknowledges and supports exemplary PhD students in computer science and related fields who are making contributions to their areas of specialty.   Guo studies robustness guarantees in algorithms and machine learning foundations, as well as their impact on society.  She is also interested in the intersection of CS and economics, and is currently focused on mechanism design, causal inference, and statistical questions in reinforcement learning. The award, which will cover full tuition, fees, and a stipend for the 2021-22 school year, will be presented at the Global Fellowship Summit over the summer.
 

Maryann Simmons and Hayley Iben win Academy Awards

CS alumnae Maryann Simmons (B.A. / M.S./ Ph.D. '01, advisor: Carlo Séquin) and Hayley Iben (M.S. '05/Ph.D. '07, advisor: James O'Brien) have won 2020 Technical Achievement Awards (SciTech Oscars) from the Academy of Motion Picture Arts and Sciences for hair simulation systems.

Simmons is now a senior staff software engineer and the technical lead for Hair & Cloth at Walt Disney Animation Studios (WDAS).  She was part of the team responsible for the WDAS Hair Simulation System, which the citation describes as "a robust, predictable, fast and highly art-directable system built on the mathematics of discrete elastic rods. This has provided Disney artists the flexibility to manipulate hair in hyper-realistic ways to create the strong silhouettes required for character animation and has enabled a wide range of complex hairstyles in animated feature films." According to The Hollywood Reporter, the WDAS System was "used in animated features such as Tangled, to manage Rapunzel’s ultra-long waves."  While at Berkeley, Simmons was a member of Phi Beta Kappa and the Golden Key Honor Society.

Iben, who is now the director of engineering at Pixar Animation Studios, was part of the team responsible for the Taz Hair Simulation System.  The citation describes Taz as "a robust, predictable and efficient mass-spring hair simulation system with novel formulations of hair shape, bending springs and hair-to-hair collisions. It has enabled Pixar artists to bring to life animated digital characters with a wide variety of stylized hair, from straight to wavy to curly."  While at Berkeley, Iben was president of Women in Computer Science and Electrical Engineering (WiCSE) from 2004-2007, and a member CSGSA.