New AI strategy enables robots to rapidly adapt to real world environments

Delivery services may be able to overcome snow, rain, heat and the gloom of night, but a new class of legged robots is not far behind. Artificial intelligence algorithms developed by a team of researchers from UC Berkeley, Facebook and Carnegie Mellon University are equipping legged robots with an enhanced ability to adapt to and navigate unfamiliar terrain in real time.

‘Eye in the sky’ sensors reduce stress of in-home dementia care

People caring around the clock for a parent, spouse or other loved one with dementia are at a high risk for clinical anxiety, depression, social isolation and even suicidal ideation, according to longitudinal research from UC Berkeley. But thanks to a blend of artificial intelligence (AI) and behavioral science, relief may be on the horizon.

Envisioning Safer Cities With AI

A team of researchers from the NSF NHERI SimCenter, a computational modeling and simulation center for the natural hazards engineering community based at the University of California, Berkeley, have developed a suite of tools called BRAILS — Building Recognition using AI at Large-Scale — that can automatically identify characteristics of buildings in a city and even detect the risks that a city's structures would face in an earthquake, hurricane, or tsunami. ,b>Charles (Chaofeng) Wang, a postdoctoral researcher at the University of California, Berkeley, and the lead developer of BRAILS, says the project grew out of a need to quickly and reliably characterize the structures in a city.

The Robot Surgeon Will See You Now

UC Berkeley postdoctoral researcher Dr. Danyal Fer and his fellow researchers hope to advance the state of the art of robotic surgery. Automated robots in testing situations can meet or exceed a human in dexterity, precision and speed, according to a new research paper from the Berkeley team. The Berkeley team decided to build a new neural network that analyzed the robot's mistakes and learned how much precision it was losing with each passing day. "It learns how the robot's joints evolve over time," said Brijen Thananjeyan, a doctoral student on the team.

Light unbound: Data limits could vanish with new optical antennas

Researchers at the University of California, Berkeley, have found a new way to harness properties of light waves that can radically increase the amount of data they carry. They demonstrated the emission of discrete twisting laser beams from antennas made up of concentric rings roughly equal to the diameter of a human hair, small enough to be placed on computer chips.

New Algorithms Could Reduce Racial Disparity

Researchers trying to improve health care with artificial intelligence usually subject their algorithms to a form of machine med school. Software learns from doctors by digesting thousands or millions of x-rays or other data labeled by expert humans until it can accurately flag suspect moles or lungs showing signs of COVID-19 by itself. A study published this month took a different approach—training algorithms to read knee x-rays for arthritis by using patients as the AI arbiters of truth instead of doctors. The results revealed that radiologists may have literal blind spots when it comes to reading Black patients' x-rays. Ziad Obermeyer, an author of the study and a professor at the University of California Berkeley's School of Public Health, was inspired to use AI to probe what radiologists weren't seeing by a medical puzzle.

Obermeyer Says Government Regulation of AI Wouldn't Stop Creative or Dangerous Uses

In this video piece from the Washington Post, Ziad Obermeyer, professor of health policy and management at the UC Berkeley School of Public Health, said government regulation of artificial intelligence can have a positive impact, but it can't get ahead of the "many creative and potentially dangerous uses that people are going to put algorithms toward...In a lot of our work what we've found is there is a substantial amount of racial bias in algorithms that are fairly widespread...that's the kind of thing that certainly suggests a role for regulation."

Cal Professors Lead Poverty Study

Three University of California, Berkeley professors will lead the work to create new datasets to better understand poverty and mobility with the help of a $2 million grant. "By constructing these datasets and making them available to a wide array of researchers, we will unlock a generation's worth of new policy research," said professor Jesse Rothstein, co-lead researcher on the project, in a statement. Also leading the project are professors Hilary Hoynes and Steven Raphael. Researchers throughout the University of California system and others will take part in the work.

Spread of Fake News Demands Tough New Rules, Says German AI Leader

The chair of Germany's artificial intelligence committee has called for tougher measures to curb the spread of fake news. Her concern came as the CEP published a report called "The Creation, Weaponization and Detection of Deep Fakes". The report's author, Professor Hany Farid from the University of California, Berkeley, explained that deep fakes, which see famous people such as U.S. President Donald Trump having words placed over official video footage, are growing in sophistication. "I think everyone can see the damage that this type of content can create for misinformation campaigns," he said. "Now we have videos of someone saying whatever you want them to say to bolster that misinformation." For more on this, see our press release at Berkeley News.

To climb like a gecko, robots need toes

Robots with toes? Experiments suggest that climbing robots could benefit from having flexible, hairy toes, like those of geckos, that can adjust quickly to accommodate shifting weight and slippery surfaces.

UC Berkeley researchers open-source RAD to improve any reinforcement learning algorithm

A group of Berkeley researchers, including electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, has released a state-of-the-art open-sourced Reinforcement Learning with Augmented Data (RAD) module. The reinforcement learning algorithm, used in machine learning for robotics, beats prior modules, including Google AI's PlaNet, DeepMind's Dreamer, and SLAC from Berkeley and DeepMind.

Learning to learn

When children play with toys, they learn about the world around them — and today’s robots aren’t all that different. At UC Berkeley’s Robot Learning Lab, groups of robots are working to master the same kinds of tasks that kids do: placing wood blocks in the correct slot of a shape-sorting cube, connecting one plastic Lego brick to another, attaching stray parts to a toy airplane.

Denim, as a Crime-Solving Tool, Has Holes

After a 1997 criminal case was largely decided by an unusual dark-and-light pattern in denim jeans captured by a security camera, a study of the case was used to set legal precedent for how patterns in photographs could be used as evidence. But now a study co-authored by postdoctoral information scholar Sophie Nightingale and information and electrical engineering and computer sciences professor Hany Farid, an expert on deepfake images, questions the reliability of such evidence. "Even under ideal conditions, trying to get an exact match is difficult," Professor Farid says. "This technique should be used with extreme caution, if at all." To conduct the study, they bought 100 pairs of jeans from local thrift stores, and they had 111 workers send in pictures of their own jeans. They then compared images of similar seams. According to the reporter: "The data showed that two images of the same seam often looked quite different -- so much so that it was often impossible to tell whether a pair of images were of the same seam or different ones. Much of the problem, the researchers concluded, comes down to the fact that cloth is flexible: it stretches, folds and drapes in complicated ways, which changes how it looks in photos. ... The lack of distinctiveness in images of seams significantly limits the accuracy of jeans identification, according to the study. The algorithm made a significant number of false matches between different pairs of jeans." Their study is available at PNAS.org.

Creating informed responses: Berkeley’s computing and data science in action

In a live webcast on Tuesday, April 7, an interdisciplinary cast of Berkeley faculty members joined Nobel laureate Saul Perlmutter, director of the Berkeley Institute for Data Science, and Michael Lu, dean of Berkeley’s School of Public Health, to discuss how data is guiding our society’s response to the pandemic and how more and better data is needed to help us emerge from the crisis.

Energy-Saving AI Is Coming for Your Office Thermostat

Electrical engineering and computer sciences professor Costas Spanos is profiled in an article about his mission to use artificial intelligence to cut office energy use by half, as one way of addressing climate change. The story talks about how Singapore turned to him to help the heavily built tropical nation reduce its air conditioning use to ease climate risks. The author writes: "Recently, the Singaporean government offered Spanos a floor in an office building to renovate. After he finished in January, workers returned to an unassuming new interior evoking the aesthetics of a hip budget airline. The room had been packed with tiny sensors detecting humidity, light, temperature, and CO2 concentration; Spanos had also devised a way to use Wi-Fi to triangulate people's locations by detecting their phones as they move through space. The theory: Armed with that anonymized data, the system would learn the workers' movements, schedules, and preferences and tweak their environment to suit. ... If the workers got too hot or too cold, they could tap an app to say so. The AI would adapt, creating microclimates to reflect their feedback. But in time, Spanos expected, the workers wouldn't bother. His goal is to make the system forgettable."

UC Berkeley robotics lab wants to fully automate a polyculture garden

AlphaGarden, a new project at Berkeley's AUTOLAB robotics laboratory, aims to develop AI systems for polyculture gardening. The AUTOLAB is best known for its DexNet system for robotic grasping. This project will see if humans can train robotic systems to fully automate the tending of a garden with multiple species of plants. Whether or not they can succeed is an "open question," says industrial engineering and operations research professor Ken Goldberg, AUTOLAB's director. "It certainly might be interesting to be able to have a fully automated garden. In my own view it's probably unlikely to be viable as a real functioning productive garden. I think that it's going to be very hard to learn, and that's the art side of the lesson, which is that nature is very complex, and that we can put some very complex machinery on it, but it's not going to necessarily open up and be controllable." Due to the COVID-19 pandemic, students working on the project will now use simulations and models, instead of a campus greenhouse. "For every real garden, we have 100,000 or millions of gardens that can be generated," Professor Goldberg says. "This runs at 100,000 times faster than nature so you can accelerate time dramatically, and for each one you can say, 'Well, if I tweak these parameters in my control policy, here's what the outcome will be in terms of how often you water, in what conditions you water, etc.'"

A.I. Versus the Coronavirus

Electrical engineering and computer sciences professor S. Shankar Sastry is going to be co-directing a new $367 million public-private research consortium with the artificial intelligence firm C3.ai and other elite universities, to work on finding A.I. solutions to global problems. The first goal of the C3.ai Digital Transformation Institute, managed jointly by Berkeley and the University of Illinois, Urbana-Champaign, will be targeting ways to slow the COVID-19 coronavirus pandemic. "I cannot imagine a more important use of A.I.," says Thomas M. Siebel, the founder and chief executive of C3.ai, who is backing the project along with Microsoft Corp. According to this reporter: "The new institute will seek new ways of slowing the pathogen's spread, speeding the development of medical treatments, designing and repurposing drugs, planning clinical trials, predicting the disease's evolution, judging the value of interventions, improving public health strategies and finding better ways in the future to fight infectious outbreaks."

COVID-19 first target of new AI research consortium

The University of California, Berkeley, and the University of Illinois at Urbana-Champaign (UIUC) are the headquarters of a bold new research consortium established by enterprise AI software company C3.ai to leverage the convergence of artificial intelligence (AI), machine learning and the internet of things (IoT) to transform societal-scale systems.

BADGR mobile robot learns to navigate on its own

Berkeley researchers have developed a mobile robot called BADGR, which learns to navigate independently. "Most mobile robots think purely in terms of geometry; they detect where obstacles are, and plan paths around these perceived obstacles in order to reach the goal," says doctoral electrical engineering and computer sciences student Gregory Kahn, one of the study's co-authors. "This purely geometric view of the world is insufficient for many navigation problems." Kahn worked on the robot with electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, and assistant electrical engineering and computer sciences professor Sergey Levine. With just 42 hours of autonomously collected data, BADGR outperformed Simultaneous Localization and Mapping (SLAM) approaches in a test, and it had less data to work with than other navigation methods, according to the researchers. Link to video.

Meet Berkeley’s new data science leader

As a Microsoft technical fellow, Jennifer Tour Chayes has made a name for herself as an expert in the field of network science and a leader of multidisciplinary labs that bring data science tools to bear on a wide range of problems. In January, she will leave her current position at Microsoft to become UC Berkeley’s first associate provost for the Division of Data Science and Information and Dean of the School of Information. 

Meet Blue, the low-cost, human-friendly robot designed for AI

Enter Blue, a new low-cost, human-friendly robot conceived and built by a team of researchers at the University of California, Berkeley. Blue was designed to use recent advances in artificial intelligence (AI) and deep reinforcement learning to master intricate human tasks, all while remaining affordable and safe enough that every artificial intelligence researcher — and eventually every home — could have one.

Literally Switching Strategies to Handle the Internet Data Flood

Cloud applications and the ever-increasing demand by large enterprises to transmit and analyze “big data” are stretching the capacity of even the largest data center servers as traditional switches become data flow bottlenecks. Ming Wu has invented a new optical, or photonic, switch capable of record-breaking speed that can be fabricated as integrated circuits so they can be mass-produced, keeping the cost per device low.

Kathy Yelick Testifies on 'Big Data Challenges and Advanced Computing Solutions'

NEWS & EVENTS News Archived News CS In the News InTheLoop Seminars & Events Kathy Yelick Testifies on 'Big Data Challenges and Advanced Computing Solutions' JULY 12, 2018 Contact: John German, jdgerman@lbl.gov, +1 510-486-6601 Kathy Yelick, Associate Laboratory Director for Computing Sciences at Berkeley Lab, was one of four witnesses testifying before the U.S. House of Representatives’ Committee on Science, Space, and Technology at 7 a.m. PDT / 10 a.m. EDT on Thursday, July 12. The discussion focused on big-data challenges and advanced computing solutions.

Megamovie project to crowdsource images of August solar eclipse

With only six months to go before one of the most anticipated solar eclipses in a lifetime, the University of California, Berkeley, and Google are looking for citizen scientists to document and memorialize the event in a “megamovie,” and help scientists learn about the sun in the process.

Three new Signatures Innovation Fellows announced

Three faculty members have been selected as 2016-17 Signatures Innovation Fellows, receiving as much as $100,000 per year each for up to two years to pursue commercially promising data science and software projects.

Big Thinking About Big Data

To Michael Jordan, the smart way to extract and analyze key information embedded in mountains of “Big Data” is to ignore most of it. Instead, zero in on collections of small amounts of data.

Seeing Through the Big Data Fog

Joe Hellerstein and his students developed a new programming model for distributed computing which MIT Technology Review named one of the 10 technologies “most likely to change our world”.

Subscribe to Data Science and AI