Stories
Slash Boxes
Comments

SoylentNews is people

Submission Preview

Link to Story

UC Berkeley ML Pioneer Wins Top Computing Gong

Accepted submission by upstart at 2022-04-07 02:54:19
News

upstart [soylentnews.org] writes:

████ # This file was generated bot-o-matically! Edit at your own risk. ████

UC Berkeley ML pioneer wins top computing gong [theregister.com]:

2021 ACM computing prize goes to UC Berkeley machine-learning pioneer Pieter Abbeel may not be a name you know, but you've probably heard of his work Copy

This year's ACM Prize in Computing is going toward a machine learning specialist whose work, even if you haven't heard of him, is likely to be familiar.

Pieter Abbeel, UC Berkeley professor and co-founder of AI robotics company Covariant [theregister.com], was awarded the prize [acm.org] and its $250,000 bounty, which is given to those in the machine learning field "whose research contributions have fundamental impact and broad implications."

Abbeel is a professor of computer science and electrical engineering whose work has already received some recognition. Along with this new award, he was named a top young innovator under 25 by the MIT Technology Review and won a prize given out to the best US PhD thesis in robotics and automation.

ACM said Abbeel was a trailblazer in apprenticeship and reinforcement learning [theregister.com], and highlighted a clothes-folding robot he designed that was better able to manipulate deformable objects. He also contributed to the development of machines that could tie sutures and detect object trajectories.

In addition, Abbeel contributed to the development of Trust Region Policy Optimization [theregister.com], one-shot imitation learning [theregister.com], domain randomization [theregister.com], hindsight experience replay [theregister.com], generalized advantage estimation, and the reinforcement learning algorithm known as soft-actor critic.

Covariant, Abbeel's company, has also made breakthroughs, including in 2020 when it developed a robotic arm able to sort equipment faster than humans [theregister.com] can.

ACM said that one of Abbeel's most important contributions to the machine learning world was his work with deep reinforcement learning, which combines reinforcement learning with deep neural networks. "While early reinforcement learning programs were effective, they could only perform simple tasks… deep reinforcement learning can solve far more complex problems than computer programs developed with reinforcement learning alone," ACM said.

Deep reinforcement learning enables AI to learn more quickly with less prior knowledge because it's able to learn from abstract, unstructured data more effectively. The approach was used in high-profile applications like learning to beat humans at Go, Chess, and Poker, and others involve improving social media notifications and training self-driving cars.

Deep reinforcement learning [theregister.com] is alternatively held up by some as a paradigm of machine learning and an impractical process unable to accurately duplicate reality.

Google software engineer Alex Irpan wrote [alexirpan.com] in 2018 that deep RL suffered from its characteristic need for a very large amount of data to generate results, something that it may not have the advantage of in real-world applications. ®

Get ourTech Resources [theregister.com]ShareCopySimilar topics

  • AI
  • Machine Learning

Broader topics

  • Self-driving Car

Narrower topics

  • Deep Learning
  • Google AI
  • GPT-3
  • MCubed
  • Tensorflow
  • Tensor Processing Unit

Corrections [theregister.com] Send us news [theregister.com] Other stories you might like

  • First Light says it's hit nuclear fusion breakthrough with no fancy lasers, magnets We talk to CEO about projectile-based implosion design [theregister.com]

    British outfit First Light Fusion claims it has achieved nuclear fusion with an approach that could provide cheap, clean power.

    Rather than rely on expensive lasers, complicated optical gear, and magnetic fields, as some fusion reactor designs do, First Light's equipment instead shoots a tungsten projectile out of a gas-powered gun at a target dropped into a chamber.

    We're told that, in a fully working reactor, the projectile will reach a velocity of 6.5 km per second (14,540 miles per hour) before it hits the moving target, which contains a small deuterium fuel capsule that implodes in the impact. This rapid implosion causes the fuel's atoms to fuse, which releases a pulse of energy. This fusion energy can be absorbed by lithium flowing through the chamber, which runs through a heat exchange to boil water into steam that spins a turbine to turn a generator that produces electricity.

    Continue reading [theregister.com]

  • IBM deliberately misclassified mainframe sales to enrich execs, lawsuit claims Lawsuit accuses Big Blue of cheating investors by shifting systems revenue to trendy cloud, mobile tech [theregister.com]

    Special report IBM has been sued by investors who claim the company under former CEO Ginni Rometty propped up its stock price and deceived shareholders by misclassifying revenues from its non-strategic mainframe business – and moving said sales to its strategic business segments – in violation of securities regulations.

    The investors' securities fraud lawsuit [PDF [regmedia.co.uk]] was filed on Tuesday, April 5 in a southern New York federal court. It names as defendants not only IBM but current and former executives including Rometty, former CFO Martin J. Schroeter (now CEO of IBM spin-off Kyndryl), current CFO James J. Kavanaugh, and current CEO Arvind Krishna.

    IBM "improperly and in violation of Generally Accepted Accounting Principles ('GAAP') embarked on a fraudulent scheme to shift billions of dollars in revenues from its mainframe line of business to its Strategic Imperatives and CAMSS line of business," the complaint reads.

    Continue reading [theregister.com]

  • Canada wants Big Tech to share its riches with news publishers Who, what, when, how still to be worked out, ironically [theregister.com]

    The Canadian government this week introduced a law bill that would force the likes of Google and Facebook to pay Canadian news publishers for using their articles online.

    The Online News Act [canada.ca] was created to address what Canadian Heritage Minister Pablo Rodriguez described as a crisis in the country's media sector that has resulted in 451 outlets disappearing between 2008 and 2015. "We want to make sure that news outlets and journalists receive fair compensation for their work. We want to make sure that local independent news thrives in our country," Rodriguez said in a press statement [newswire.ca].

    Specifically, the proposed law seeks to ensure journalists and publishers get a fair cut of the revenues Big Tech banks from aggregating, distributing, sharing, or summarizing stories; the exact arrangements have yet to be hammered out.

    Continue reading [theregister.com]

  • Feds take down Kremlin-backed Cyclops Blink botnet Control systems scrubbed, hijacked network devices need to be patched and cleaned [theregister.com]

    The US Justice Department today revealed details of a court-authorized take-down of command-and-control systems the Sandworm cyber-crime ring used to direct network devices infected by its Cyclops Blink malware.

    The move follows a joint security alert [ncsc.gov.uk] in February from US and UK law enforcement that warned of WatchGuard firewalls and ASUS routers [theregister.com] being compromised to run Cyclops Blink. This botnet malware – technical breakdown here [ncsc.gov.uk] [PDF] – allows the equipment to be remote controlled to carry out attacks on behalf of its masterminds.

    Previously, Uncle Sam said the Sandworm crew worked for the Russian Federation's GRU espionage nerve-center, which handles foreign intel operations.

    Continue reading [theregister.com]

  • DARPA says US hypersonic missile is ready for real world A new arms race emerges [theregister.com]

    DARPA announced a second successful test of its hypersonic cruise missile, adding that the weapon, capable of speeds in excess of Mach 5, is ready for real-world use.

    DARPA has been testing hypersonic weapons for several years. The program in question, called the Hypersonic Air-breathing Weapon Concept (HAWC [darpa.mil]), began in 2016 [theregister.com] with Raytheon as a partner. Raytheon's HAWC first flew in 2021, and this latest test used a design manufactured by Lockheed Martin and Aerojet Rocketdyne.

    Hypersonic weapons are any missile that travels at Mach 5 or faster, but with an added caveat: they can be controlled. Traditional missiles often exceed Mach 5 as well, but can't be steered once they're launched.

    Continue reading [theregister.com]

  • Google Cloud previews new BigLake data lakehouse service Hoarding a bunch of data that 'may prove useful' someday? Data-wrangling product is aimed at you [theregister.com]

    Google has announced a preview on Google Cloud of BigLake, a data lake storage service that it claims can remove data limits by combining data lakes and data warehouses.

    BigLake [google.com] is designed to address the problems associated with the growing volumes of data and different types of data now being stored and retained by organizations of all sizes. The motivation for storing all of this data can often be summed up as "because it may prove useful", with the idea being that if it is analyzed using the right tools, it will yield valuable insights that will benefit the business.

    Unveiled to coincide with Google's Data Cloud Summit [google.com], BigLake allows organizations to unify their data warehouses and data lakes to analyze data without worrying about the underlying storage layer. This eliminates the need to duplicate or move data around from its source to another location for processing and reduces cost and inefficiencies, Google claimed.

    Continue reading [theregister.com]

  • CircleCI suffering 'degraded performance' for hours DevOps platform tells devs it is ‘continuing to work on a longer-term fix’ [theregister.com]

    Updated DevOps outfit CircleCI is suffering from performance issues that have been ongoing for most of the day, with developers left waiting for the platform to come back online.

    The service keeps an eye on the likes of GitHub and Bitbucket and fires off a build when commits are made. It then tests builds and pings out notifications if things go wrong. The platform also features a managed cloud service.

    However, today things went a bit awry. CircleCI described the issue [circleci.com] as "degraded performance," which had begun with "database read/write delays." Meanwhile, over on Twitter, the firm promoted a Lego raffle [twitter.com] as users became increasingly antsy.

    Continue reading [theregister.com]

  • Block claims ex-employee downloaded customer data after leaving firm Leak highlights worker offboarding policies as SaaS use grows [theregister.com]

    A former employee with Block used the digital financial services firm's Cash App products to access and download personal information about US customers in December 2021, the firm has claimed.

    In a filing [sec.gov] this week with the Securities and Exchange Commission (SEC), Block officials alleged the ex-employee on December 10 downloaded reports of the company's Cash App Investing subsidiary.

    The data, it said, included such information as the full name and brokerage account number – a unique identification number linked to a customer's stock activity – as well as brokerage portfolio value, holdings and stock trading activity.

    Continue reading [theregister.com]

ABOUT US

MORE CONTENT

SITUATION PUBLISHING

The Register - Independent news and views for the tech community. Part of Situation Publishing

Biting the hand that feeds IT © 1998–2022

Do not sell my personal information [theregister.com]Cookies [theregister.com]Privacy [theregister.com]Ts&Cs [theregister.com]

Original Submission