New Music Artist Videos
1,856 Videos. Showing Newest from #1
The Rust Programming Language

Add to EJ Playlist  Google Tech Talk June 6, 2015 (click "show more" for more information) Presented by Alex Crichton ABSTRACT Rust is an exciting new programming language that combines fine-grained control over memory layout and allocation patterns with memory and thread safety. It reached 1.0 in May and is rapidly gaining acceptance in projects including Servo (Mozilla's next generation parallel layout engine), a block store being developed by Dropbox, and numerous startups. Its features include zero-cost abstractions, move semantics, guaranteed memory safety, threads without data races, trait-based generics, pattern matching, type inference, minimal runtime, and efficient C bindings. ‚Alex Crichton will present an introduction to the Rust programming language, explain how it can be used to build performant, reliable systems, and answer your questions. Speaker Info: Alex Crichton ‚About the speaker: Alex Crichton is a member of the Rust core team working at Mozilla Research. He focuses on the standard library, Cargo,, and Rust's tooling and infrastructure experience. He's been at Mozilla for almost two years and graduated from Carnegie Mellon in 2013 with a BS in both Computer Science and Electrical and Computer Engineering.

Google NYC Tech Talk: Google Ideas

Add to EJ Playlist  Google Tech Talk June 30, 2015 (click "show more" for more information) Presented by Yasmin Green, Salome Vakhtangadze, and Baris Yuksel ABSTRACT As long as people have expressed ideas, others have tried to silence them. Today one out of every three people lives in a society that is severely censored. Online barriers can include everything from filters that block content to targeted attacks designed to take down websites. For many people, these obstacles are more than an inconvenience—t hey represent full-scale repression. Engineers from Google Ideas will present the way technology is being used to enable free expression online, and protect every voice in the digital debate. Google Presenters: Yasmin Green Yasmin Green is the Head of Strategy and Operations for Google Ideas. She also oversees the team’s work on counter-radical ization and fragile states. At Google, Yasmin has assumed roles as Head of Sales Strategy and Operations for Southern Europe, Middle East, and Africa and Sub-Saharan Africa Operations Manager, and prior to joining Google, she consulted for Booz Allen Hamilton across Europe, the Middle East and Africa. Yasmin has extensive experience leading and project managing in some of the world’s toughest environments. She has spent time in Iran, Syria and Nigeria, where she has worked cross-functiona lly in sectors ranging from oil and gas to the Internet. She recently led a multi-partner coalition to launch Against Violent Extremism, the world's first online network of former violent extremists and survivors of terrorism. Yasmin is a Senior Advisor on Innovation to Oxford Analytica and Co-Chair of the European Commission's’ Working Group on Online Radicalization. Salome Vakhtangadze Salome is a Software Engineer. She joined Google Ideas after having worked on Google's Search Infrastructure team for 3 years. She currently works on uProxy, and is passionate about human rights. Baris Yuksel Baris Yuksel is a tech lead and software engineering manager. He is currently leading projects at Google Ideas, a think/do tank that explores how technology can enable people to confront threats in the face of conflict, instability and repression. Google Ideas was founded on the principle that the next five billion to come online will face far greater human challenges than the first two billion. Baris has been focusing on designing and developing next generation features and technologies, trying to create magic moments not for few but for all. In the past, Baris worked on Google Search with Craig Nevill-Manning, founder of Google New York. Together they worked on features bridging the gap between Search and Google Now. Baris has been a volunteer with ScriptEd teaching high school students with an experimental curriculum, which he has been developing under the guidance of ScriptEd. He is hopeful that it will contribute to the major teaching effort. Baris graduated from MIT with a BS in Electrical Engineering and Computer Science, minoring in Economics.

ZuriHac 2015 - Discrimination is Wrong: Improving Productivity

Add to EJ Playlist  Google Tech Talk May 30, 2015 ("show more" for more information) Presented by Edward Kmett Web: https://wiki.ha ac2015#Edward_K mett Slides are available here: https://github. com/meiersi/Has kellerZ/blob/ma ster/meetups/20 150530-ZuriHac2 015_Edward_Kmet t-Discriminatio n_is_Wrong_Impr oving_Productiv ity/Discriminat ion%20-%20Zurih ac.pdf ABSTRACT This talk is a case study in library design in Haskell. Fritz Henglein has shown through a number of excellent papers how to use "discrimination " to do lots of things in O(n): Sorting many more data types than you'd expect, table joins, etc. In the process of optimizing this approach and wrapping it up in a form that can be easily consumed, we'll take a lot of detours through the different ways you can think about code when optimizing Haskell. We'll need some category theory, from a deeper understanding of monoids to Day convolution. We'll need to consider final and initial encodings. We'll need to drift down to low level system concerns from building a custom foreign prim to nesting unsafePerformIO within unsafePerformIO . We'll need properties of laziness from productivity to IVars. Along the way we'll find and fix a small problem with the initial discrimination paper, which opens the door to streaming results, rather than having to wait until all the input is ready.

Wearable Cognitive Assistance

Add to EJ Playlist  Google Tech Talk June 15, 2015 ("show more" for more information) Presented by Mahadev Satyanarayanan, School of Computer Science, Carnegie Mellon University ABSTRACT GPS navigation systems have transformed our driving experience. They guide you step-by-step to your destination, offering you helpful just-in-time voice guidance about upcoming actions. Can we generalize this metaphor? Imagine a wearable cognitive assistance system that combines a device like Google Glass with cloud-based processing to guide you through a complex task. You hear a synthesized voice telling you what to do next, and you see visual cues in the Glass display. When you make an error, the system catches it immediately and corrects you before the error cascades. Fortunately, this is not science fiction. To support this genre of applications, we have created a multi-tiered system architecture called Gabriel. This architecture preserves tight end-to-end latency bounds on compute-intensi ve operations, while addressing concerns such as limited battery capacity and limited processing capability of wearable devices. We have gained initial experience on building Gabriel applications that provide user assistance for narrow and well-defined tasks that require specialized knowledge and/or skills. Specifically, we have built proof-of-concep t implementations for four tasks: assembling 2D Lego models, freehand sketching, playing ping-pong, and recommending context-relevan t YouTube tutorials. This talk will examine the potential and challenges of wearable cognitive assistance, present the Gabriel architecture, and describe the early proof-of-concep t applications that we have built on it. Credits Bio at http://cs.cmu.e du/~satya

ZuriHac 2015 - Performance

Add to EJ Playlist  Google Tech Talk May 31, 2015 ("show more" for more information) Presented by Johan Tibell Note: We are sorry for missing first few minutes of the presentation. https://wiki.ha ac2015#Johan_Ti bell ABSTRACT Production quality code not only needs to compute the correct result, but needs to do so within some given resource constraints (i.e. time and memory). We'll look at the most common optimizations you can use to make your Haskell code perform better and some rules of thumb you can use in your daily Haskell programming to avoid some performance problems "by design". Slides and other resources for this talk are available here: https://github. com/meiersi/Has kellerZ/tree/ma ster/meetups/20 150531-ZuriHac2 015_Johan_Tibel l-Performance

ZuriHac 2015 - Distributed Programming in Haskell

Add to EJ Playlist  Google Tech Talk May 30, 2015 ("show more" for more information) Presented by Mathieu Boespflug https://wiki.ha ac2015#Mathieu_ Boespflug ABSTRACT Shifting from building compilers and desktop applications to online services and mobile apps means that Haskell programming is increasingly about building systems, not standalone data-in data-out contraptions. Systems that are distributed across anything from a handful to millions of machines. We'll look at what Haskell has to offer for building robust and scalable distributed systems, starting from the paradigms popularized in other languages to more idiomatic, safe and composable models. Slides for this talk are available here: https://github. com/meiersi/Has kellerZ/blob/ma ster/meetups/20 150530-ZuriHac2 015_Mathieu_Boe spflug-Distribu ted_Programming /ZuriHac15%20ta lk-%20distribut ed%20haskell.pd f

ZuriHac 2015 - Better Faster Binary Serialization

Add to EJ Playlist  Google Tech Talk May 29, 2015 ("show more" for more information) Presented by Duncan Coutts https://wiki.ha ac2015#Duncan_C outts ABSTRACT This talk is a case study in low level optimization in Haskell. We have existing libraries for binary serialization but the mainstream ones use bad formats, and could always be faster. We will look at new work to improve on the existing libraries, to use a new better format, and to dramatically improve performance. We will go into the details of a number of standard and new techniques for getting a good combination of performance with features and flexibility. - High+low level approach to eliminating unnecessary allocations in deserialization - Surprising use of deep embeddings and interpreters - Moving work from compiled code to an interpreter - Rewrite rules to improve interpreted code - Deep embedding of composition to give statically allocated descriptions of branching programs - Types to safely use unsafe low level data structures - Looking at STG code to see allocations and other problems - Fast path / slow path tricks - Explicit control stacks - ABIs / calling conventions for unknown functions There'll be plenty of code, types and benchmarks. Slides for this talk are available here: https://github. com/meiersi/Has kellerZ/blob/ma ster/meetups/20 150529-ZuriHac2 015_Duncan_Cout ts-Better_Faste r_Binary_Serial ization/binary. pdf

ZuriHac 2015 - Beginning Web Programming in Haskell

Add to EJ Playlist  Google Tech Talk May 29, 2015 ("show more" for more information) Presented by Oliver Charles https://wiki.ha ac2015#Oliver_C harles ABSTRACT For many people, Haskell is not their first programming language - a lot of us learnt to program in other languages like Python or Java, and would like to be able to use Haskell to solve practical problems. However, what seem to be simple web applications in other languages can be daunting to even begin using Haskell. In this talk, we'll look at how to build a small web application to allow ZuriHac participants to share what they have been working on. We'll use the Spock web framework along with postgresql-simp le to build a basic CRUD application, and see how we can start prototyping a simple web application within minutes. Slides for this talk are available here: https://github. com/meiersi/Has kellerZ/blob/ma ster/meetups/20 150529-ZuriHac2 015_Oliver_Char les-Web_Program ming/talk.pdf

ZuriHac 2015 - Monads by Example

Add to EJ Playlist  Google Tech Talk May 29, 2015 ("show more" for more information) Presented by Jannis Limperg https://wiki.ha ac2015#Jannis_L imperg ABSTRACT The Monad type class is used extensively in both the Haskell standard library and numerous third-party libraries for all sorts of domains. Yet, it is also one of the most infamous roadblocks for Haskell beginners, which this talk aims to address. In order to build an intuition for what a Monad 'is' and especially for how it's useful in everyday programming, we will take a look at several important instances of the Monad class from the standard library (and the de-facto standard transformers package): Maybe, Either, [], State and IO. The examples will focus for the most part on how these very different types are all instances of the abstract concept of a Monad, and how using this abstraction and the associated syntactic sugar (do notation) can lead to cleaner, more readable code. Afterwards, we will take a look at how the Monad class is defined by a set of simple laws that every Monad instance must obey. Time permitting, we may also touch upon slightly more advanced topics such as Monad transformers and how Monads are related to Functors and Applicative Functors. Prerequisites: Learn Yourself A Haskell up to and including chapter 8 (or of course anything else that covers the same topics). In particular, a basic understanding of algebraic data types and type classes is essential. Slides and code examples for this talk are available here: https://github. com/meiersi/Has kellerZ/tree/ma ster/meetups/20 150529-ZuriHac2 015_Jannis_Limp erg-Monads_by_E xample The code shown in this talk is available at the above link so you can follow along if desired.

Liquid Democracy with Google Votes

Add to EJ Playlist  Google Tech Talk March 12, 2014 (show more for more details) Presented by Steve Hardt ABSTRACT Google Votes is an experiment in liquid democracy built on Google's internal corporate Google+ social network. A Liquid Democracy system gives all the control of Direct Democracy with the scalability of Representative Democracy. Users can vote directly or delegate power through their social networks. This talk covers user experience aspects of delegated voting and three graph algorithms for flowing votes through a social graph called Tally, Coverage, and Power. View part one of this two part series at /7UqV3eN9Pa8 Read the paper at 'Google Votes: A Liquid Democracy Experiment on a Corporate Social Network' Axj9e

Voting Methods with Google Votes

Add to EJ Playlist  Google Tech Talk March 3, 2014 (show more for more details) Presented by Steve Hardt ABSTRACT Learn the ins-and-outs of common voting methods and algorithms for group-decision- making with illustrated examples in Google Votes. Google Votes is a Google-internal voting platform used for decisions like food selections, cafe names, t-shirt designs, and Halloween contests. This talk covers Approval, Ranked, Range/Score, and Plurality voting along with the Borda and Schulze ranking algorithms and their relation to Condorcet's Paradox. View part two of this two part series at /F4lkCECSBFw Read the paper at 'Google Votes: A Liquid Democracy Experiment on a Corporate Social Network' Axj9e

Clasp: Common Lisp using LLVM and C++ for Molecular Metaprogramming

Add to EJ Playlist  Google Tech Talk June 9, 2015 (more info - click "show more") Presented by Christian Schafmeister. ABSTRACT This talk describes our unique approach to constructing large, atomically precise molecules (called "Molecular Lego" or "spiroligomers" ) that could act as new therapeutics, new catalysts (molecules that make new chemical reactions happen faster) and ultimately to construct atomically precise molecular devices. Then I describe Clasp and CANDO, a new implementation of the powerful language Common Lisp. Clasp is a Common Lisp compiler that uses LLVM to generate fast machine code and it interoperates with C++. CANDO is a molecular design tool that uses Clasp as its programming language. Together I believe that these are the hardware (molecules) and the software (the CANDO/Clasp compiler) that will enable the development of sophisticated molecular nanotechnology. Clasp is available at: https://github. com/drmeister/c lasp For more info see: https://chem.cs rectory/faculty /schafmeister/ More about Clasp Clasp is an implementation of Common Lisp that interoperates with C++ and uses LLVM as its backend. It is available at https://github. com/drmeister/c lasp. The goal of Clasp is to become a performant Common Lisp that can use C++ libraries and interoperate with LLVM-based tools and languages. The first sophisticated C++ library with which Clasp interoperates is the Clang C/C++ compiler front end. Using the Clang library, Common Lisp programs can be written that parse and carry out static analysis and automatic refactoring of C/C++ code. This facility is used to automatically analyze the Clasp C++ source code and construct an interface to the Memory Pool System compacting garbage collector. The primary purpose of Clasp is to act as a performant language for scientific computing that will be used to design sophisticated new molecular devices, catalysts and therapeutic molecules based on our "Molecular Lego" technology. Clasp is a general programming language that will support many other applications. About the Speaker: Christian Schafmeister visited Google’s Cambridge, MA office to deliver the talk "Clasp: Common Lisp using LLVM and C++ for Molecular Metaprogramming ”. Christian is Associate Professor of Chemistry at Temple University. Research projects within his group will use the tools of synthetic chemistry, molecular biology, and X-ray crystallography to develop a universal molecular scaffold that would allow the systematic design, construction, and investigation of macromolecules that display chemical functionality in three-dimension al space. A scaffold like this will allow the design and synthesis of new catalysts, molecular sensors, and ultimately molecular machines. His group will use synthetic chemistry to synthesize molecular building blocks that they will couple to each other through pairs of bonds to construct rigid macromolecules with diverse and programmable shapes. His group has developed computer software that allows the computer-aided design of these molecules to carry out specific functions.

Mark Gold: "Water Water Everywhere Nor Any Drop to Drink"

Add to EJ Playlist  Professor Mark Gold visited Google LA on May 1, 2015. He is the former director of Heal the Bay and made a presentation on the current water crisis. California is in an unparalleled water crisis. Not only are we in the 4th year of a drought, but the April 1st Sierra snowpack was a record 5% of normal. In response to the crisis, cities have been mandated to cut their water use by an average of 25% and agriculture will soon receive massive cutbacks in their water allocations. How will the drought affect you and can you do anything to make a difference?

Alán Aspuru-Guzik: "Billions and Billions of Molecules"

Add to EJ Playlist  Alán Aspuru-Guzik visited the Quantum AI Lab at Google LA on May 12, 2015 and gave this talk: "Billions and Billions of Molecules: Molecular Materials Discovery in the Age of Machine Learning" Abstract: Many of the challenges of the twenty-first century are related to molecular processes such as the generation, transmission, and storage of clean energy, water purification and desalination. These transformations require a next generation of more efficient and ecologically-fr iendly materials. In the life sciences, we face similar challenges, for example drug-resistant bacterial strains require novel antibiotics. One of the paradigm shifts that the theoretical and experimental chemists needs to embrace is that of accelerated molecular discovery: The design cycles need to be sped up by the constant interaction of theoreticians and experimentalist s, the use of high-throughput computational techniques, tools from machine learning and big data, and the development of public materials databases. I will describe three projects from my research group that aim to operate in this accelerated design cycle. First, I will describe our efforts on the Harvard Clean Energy Project (http://cleanen ergy.harvard.ed u), a search for materials for organic solar cells. I will continue by talking about our work on developing organic molecules for energy storage in flow batteries. Finally, I will describe our work towards the discovery of novel molecules for organic light-emitting diodes. If time permits, I will talk about molecular networks related to the origins of life. Bio: Professor Alán Aspuru-Guzik is currently Professor of Chemistry and Chemical Biology at Harvard University. He began at Harvard in​ 2006 and ​was promoted to Full Professor in 2013. Alán received his B.Sc.​ Chemistry from the National Autonomous University of Mexico (UNAM) in 1999. He received the Gabino Barreda Medal from UNAM. ​He obtained a PhD in Physical Chemistry from the University of California, Berkeley in 2004, under Professor William A. Lester, Jr., he was a postdoctoral scholar in the group of Martin Head-Gordon at UC Berkeley from 2005-2006. In 2009, Professor Aspuru-Guzik received the DARPA Young Faculty Award, the Camille and Henry Dreyfus Teacher-Scholar award and the Sloan Research Fellowship. In 2010, he received the Everett-Mendels son Graduate Mentoring Award and received the HP Outstanding Junior Faculty award by the Computers in Chemistry division of the American Chemical Society. In the same year, he was selected as a Top Innovator Under 35 by the Massachusetts Institute of Technology Review magazine. In 2012, he was elected as a fellow of the American Physical Society, and in 2013, he received the ACS Early Career Award in Theoretical Chemistry. He is associate editor of the journal Chemical Science. Professor Aspuru-Guzik carries out research at the interface of quantum information and chemistry. In particular, he is interested in the use of quantum computers and dedicated quantum simulators for chemical systems. He has proposed quantum algorithms for the simulation of molecular electronic structure, dynamics and the calculation of molecular properties. He recently has proposed two new approaches for quantum simulation: the variational quantum eigensolver and the adiabatic quantum chemistry approach. He also proposed the demon-like algorithmic cooling algorithm. He has studied the role of quantum coherence in excitonic energy transfer in photosynthetic complexes. Alán has been involved as a theoretician in several experimental demonstrations of quantum simulators using quantum optics, nuclear magnetic resonance, nitrogen vacancy centers and recently superconducting qubits. ​Alán develops methodology for the high-throughput search of organic materials, especially organic materials. This has led to his discovery of candidate molecules for high mobility organic semiconductors, organic flow battery molecules and high-performanc e molecules for organic light-emitting diodes. Alan is very interested in the interface of machine learning and material discovery and has carried out the largest set of quantum chemistry calculations to date.

Christopher Monroe: "Modular Ion Trap Quantum Networks: Going Big"

Add to EJ Playlist  Christopher Monroe visited the Google LA Quantum AI Lab on May 4, 2015. Abstract: Laser-cooled and trapped atomic ions are standards for quantum information science, acting as qubits with unsurpassed levels of quantum coherence while also allowing near-perfect measurement. Trapped ions can be entangled locally with external laser beams that map the internal atomic qubits through their Coulomb-coupled motion; they can be entangled remotely through optical photons traveling through fibers. This quantum hardware platform scales favorably when compared to any other platform, because (a) each trapped ion is an atomic clock that is identical to the others, (b) the qubit wiring (interaction graph) is provided by external fields and is hence reconfigurable, and (c) the modular architecture of local and remote quantum gates provides a realistic path to building out huge systems with thousands of qubits. I will summarize the state-of-the art in ion trap quantum networks, which will soon involve systems of 50+ fully-interacti ng qubits, eclipsing the performance of classical computers for certain tasks. The remaining challenges in the ion trap architecture are primarily engineering, setting the stage for focused efforts to build a large scale quantum computer. Bio: Christopher Monroe is an quantum physicist who specializes in the isolation of individual atoms for applications in quantum information science. After graduating from MIT, Monroe studied with Carl Wieman and Eric Cornell at the University of Colorado, earning his PhD in Physics in 1992. His work paved the way toward the achievement of Bose-Einstein condensation in 1995 and led to the 2001 Nobel Prize for Wieman and Cornell. From 1992-2000 he was a postdoc then staff physicist at the National Institute of Standards and Technology, in the group of David Wineland. With Wineland, Monroe led the team that demonstrated the first quantum logic gate in 1995, and exploited the use of trapped atoms for the first controllable qubit demonstrations, eventually leading to Wineland's Nobel Prize in 2012. In 2000, Monroe became Professor of Physics and Electrical Engineering at the University of Michigan, where he pioneered the use of single photons to couple quantum information between atoms and also demonstrated the first electromagnetic atom trap integrated on a semiconductor chip. From 2006-2007 was the Director of the National Science Foundation Ultrafast Optics Center at the University of Michigan. In 2007 he became the Bice Zorn Professor of Physics at the University of Maryland and a Fellow of the Joint Quantum Institute. In 2008, Monroe's group succeeded in producing quantum entanglement between two widely separated atoms and for the first time teleported quantum information between matter separated by a large distance. Since 2009 his group has investigated the use of ultrafast laser pulses for speedy quantum entanglement operations, pioneered the use of trapped ions for quantum simulations of many-body models related to quantum magnetism, and with Jungsang Kim (Duke University) has proposed and made the first steps toward a scalable, reconfigurable, and modular quantum computer.

Toward Gender-Inclusive Software: A Tale of 5 Facets

Add to EJ Playlist  Google Tech Talk May 11, 2015 (click "show more" for more info) Presented by Professor Margaret Burnett ABSTRACT Gender inclusiveness in Silicon Valley is receiving a lot of attention these days, but it overlooks a potentially critical factor—software itself. Research into how individual differences cluster by gender shows that males and females tend to work differently with software that aims to help people solve problems (e.g., tools for debugging, for end-user programming, for game-based learning, and for visualizing information). However, many features of problem-solving software are (inadvertently) designed around the way males problem-solve. In this talk, I’ll explain 5 facets of these gender differences and how they tie to a large body of foundational work on gender differences from computer science, psychology, education, communications, and women’s studies. I’ll then present emerging work on our "GenderMag method”, an inspection method that encapsulates these 5 facets into practitioner-re ady form. Early empirical results suggest that GenderMag is effective at enabling practitioners to pinpoint gender inclusiveness issues in software features. My visit to Google is to reach out to any Googlers who might be interested in trying GenderMag, so as to identify gender inclusiveness issues with their own software projects. About the speaker: Margaret Burnett is currently a professor of Computer Science in the School of EECS at Oregon State University. Margaret Burnett's research on gender inclusiveness in software — especially in tools for programming by end users — spans over 10 years. Burnett and her team systematically debunked misconceptions of gender inclusiveness in a variety of software platforms, and then devised software features that help avert the identified problems. She has reported these results in over 30 publications, and has presented keynotes and invited talks on this topic in 7 countries. She serves on a variety of HCI and Software Engineering committees and editorial boards, and serves on the Advisory Board of the Academic branch of National Center for Women & Information Technology (NCWIT).

Rethinking Learning Systems with the Tin Can API

Add to EJ Playlist  Google Tech Talk January 16, 2014 (Click "Show More" for abstract and speaker info) Presented by Mike Rustici ABSTRACT We use learning systems to capture and quantify what people learn. Today these systems only have visibility into a tiny fraction of a person’s knowledge and learning experiences. The Tin Can API is changing that. This new API provides a common language for describing experience to talk about the things people are doing and what they are learning. Tin Can is enabling a wave of disruption and innovation in the online learning industry. Learning systems are transitioning from monolithic LMSs to a collection of best of breed services that are “tightly coupled, yet loosely integrated”. This Tech Talk will describe the transformative effect of the Tin Can API and how it is changing the way we learn. Speaker Info: Mike Rustici (mike.rustici@s Mike struck out on his own in 2002 with the intent of building web-based software for people as a contractor. Mike’s experience with SCORM, dating back to 2000, quickly led to high demand for his insight, solutions, and products. At Rustici Software, Mike’s collar sets him apart from the rest of the crew. Mike is regularly mocked in the office for being among the top five SCORM/Tin Can professionals in the world. This is only funny because it is probably true. Tim Martin (tim.martin@sco Tim joined Mike in 2003, adding the exciting LLC part of the Rustici Software LLC name. Mike and Tim had spent several years together at PureSafety collectively deciding what they believed to be good software architecture, great company leadership, and a fun place to work. While Tim used to be a software developer, he has fallen gracefully into the world of sales. If you’re looking to buy something from Rustici Software, play an inane game, or just get a plainspoken answer to a SCORM or Tin Can API question, Tim is the right person to talk to.

Learning Language from Other People

Add to EJ Playlist  Google Tech Talk March 25, 2012 (more info below) Presented by Mike Frank ABSTRACT How do children learn their first words? One possibility is that, even from early infancy, children's word learning is driven by a desire to share information with others - to communicate, in other words. This provocative hypothesis has an increasing amount of support in the empirical literature. My research has been devoted to understanding the empirical and formal consequences for early word learning of children's orientation towards communication. I'll show a formal framework for thinking about communicative word learning and then highlight data from an in-depth investigation of a single child's word learning and from the large-scale dynamics apparent in children's vocabularies at the scale of tens of thousands of kids. Bio: Michael C. Frank is Associate Professor of Psychology at Stanford University. He earned his BS from Stanford University in Symbolic Systems in 2005 and his PhD from MIT in Brain and Cognitive Sciences in 2010. He studies children's language learning and how it interacts with their developing understanding of the social world, using behavioral experiments, computational tools, and novel measurement methods including large-scale web-based studies, eye-tracking, and head-mounted cameras.

RobustIRC — IRC without Netsplits

Add to EJ Playlist  Google Tech Talk February 9, 2015 (more info below) Presented by Michael Stapelberg ABSTRACT In this tech talk, I will present RobustIRC, an IRC server built as a distributed system on top of Raft (a simpler alternative to Paxos), written in Go. Due to its distributed nature and a transport protocol that can take a failing TCP connection, netsplits are a thing of the past in RobustIRC. That all said, as the name implies, RobustIRC is usable with plain old IRC clients such as irssi, WeeChat, … I’ll also cover RobustIRC’s scalability properties and what the limitations are. Target audience is anyone who has ever used IRC (or is interested in distributed systems), probably especially SREs due to their inexplicable preference for IRC over other chat protocols ;).

Daniel Lidar: "Quantum Information Processing: Are We There Yet?"

Add to EJ Playlist  Daniel Lidar visited the Quantum AI Lab at Google LA to give a talk: "Quantum Information Processing: Are We There Yet?" This talk took place on January 22, 2015. Abstract: Quantum information processing holds great promise, yet large-scale, general purpose quantum computers capable of solving hard problems are not yet available despite 20+ years of immense effort. In this talk I will describe some of this promise and effort, as well as the obstacles and ideas for overcoming them using error correction techniques. I will focus on a special purpose quantum information processor called a quantum annealer, designed to speed up the solution to tough optimization problems. In October 2011 USC and Lockheed-Martin jointly founded a quantum computing center housing a commercial quantum annealer built by the Canadian company D-Wave Systems. A similar device is operated by NASA and Google. These processors use superconducting flux qubits to minimize the energy of classical spin-glass models with as many spins as qubits, an NP-hard problem with numerous applications. There has been much controversy surrounding the D-Wave processors, questioning whether they offer any advantage over classical computing. I will survey the recent work we have done to benchmark the processors against highly optimized classical algorithms, to test for quantum effects, and to perform error correction. Bio: Daniel Lidar has worked in quantum computing for nearly 20 years. He is a professor of electrical engineering, chemistry, and physics at USC, and hold a Ph.D. in physics from the Hebrew University of Jerusalem. His work revolves around various aspects of quantum information science, including quantum algorithms, quantum control, the theory of open quantum systems, and theoretical as well as experimental adiabatic quantum computation. He is a Fellow of the AAAS, APS, and IEEE. Lidar is the Director of the USC Center for Quantum Information Science and Technology, and is the Scientific Director of the USC-Lockheed Martin Center for Quantum Computing. Two of his former graduate students are now research scientists at Google’s quantum artificial intelligence lab.

Stochastic Optimization for x86 Binaries

Add to EJ Playlist  Google Tech Talks January 12, 2015 (more info below) ABSTRACT The optimization of short sequences of loop-free fixed-point x86_64 code sequences is an important problem in high-performanc e computing. Unfortunately, the competing constraints of transformation correctness and performance improvement often force even special purpose compilers to produce sub-optimal code. We show that by encoding these constraints as terms in a cost function, and using a Markov Chain Monte Carlo sampler to rapidly explore the space of all possible programs, we are able to generate aggressively optimized versions of a given target program. Beginning from binaries compiled by gcc -O0, we are able to produce provably correct code sequences that either match or outperform the code produced by gcc -O3, and in some cases expert hand-written assembly. Because most high-performanc e applications contain floating-point computations, we extend our technique to this domain and show a novel approach to trading full floating-point precision for further increases in performance. We demonstrate the ability to generate reduced precision implementations of Intel's handwritten C numerics library that are up to six times faster than the original code, and achieve end-to-end speedups of over 30% on a direct numeric simulation and a ray tracer. Because optimizations that contain floating-point computations are not amenable to formal verification using the state of the art, we present a technique for characterizing maximum error and providing strong evidence for correctness. Publication list: http://cs.stanf eschkufz/ Github: https://github. com/eschkufz/st oke-release About the speaker Eric Schkufza is a postdoctoral scholar at Stanford University working with professor Alex Aiken. He graduated from Stanford University with a PhD in computer science in June 2014. He is interested in applying stochastic search techniques to the design of optimizing compilers.

Matthias Troyer: "High Performance Quantum Computing"

Add to EJ Playlist  Matthias Troyer visited Google LA to speak about "High Performance Quantum Computing." This talk took place on December 2, 2014. Abstract: As the outlines of a roadmap to building powerful quantum devices becomes more concrete an important emerging question is that of important real-world applications of quantum computers. While there exist many quantum algorithms which asymptotically outperform classical algorithms, asymptotic superiority can be misleading. In order for a quantum computer to be competitive, it needs to not only be asymptotically competitive but be able to solve problems within a limited time (for example one year) that no post-exa-scale classical supercomputer can solve within the same time. This search for a quantum killer-app turns out to be a formidable challenge. Using quantum chemistry simulations as a typical example, it turns out that significant advances in quantum algorithms are needed to achieve this goal. I will review how substantial improvements and optimized massively parallel implementation strategies of quantum algorithms have brought the problem of quantum chemistry from the realm of science fiction closer to being realistic. Similar algorithmic improvements will be needed in other areas in order to identify more “killer apps” for quantum computing. I will end with a short detour to quantum annealers and present a summary of our recent results on simulated classical and quantum annealing. Bio: Matthias Troyer is professor of computational physics at ETH Zurich where he teaches advanced C++ programming, high performance computing, and simulations methods for quantum systems. He is a pioneer of cluster computing in Europe, having been responsible for the installation of the first Beowulf cluster in Europe with more than 500 CPUs in 1999, and the most energy efficient general purpose computer on the top-500 list in 2008. He is a Fellow of the American Physical Society and his activities range from quantum simulations and quantum computing to the development of novel simulation algorithms, high performance computing, and computational provenance. He is, the author of the Boost MPI C++ library for message passing on parallel computers, and the leader of the open-source ALPS library for the simulation of quantum many body systems.

Evolving JavaScript with TypeScript

Add to EJ Playlist  Google Tech Talk November 13, 2014 Presented by Anders Hejlsberg (more info below) ABSTRACT In this talk Anders will give an overview of TypeScript, a statically typed superset of JavaScript that compiles to plain JavaScript. TypeScript adds optional static types, classes, interfaces, and modules to JavaScript, enabling IDE productivity features such as statement completion, refactoring, and code navigation, and making it easier for teams to communicate requirements and build applications safely. The TypeScript language features an innovative structural type system that incorporates gradual typing and type inference. The TypeScript project is open-source and hosted on GitHub. Speaker Info: Anders Hejlsberg is a Technical Fellow in the Developer Division at Microsoft Corporation and works on Microsoft’s .NET and JavaScript development tools. He is the chief designer of the C# programming language and co-designer of the TypeScript programming language. Before joining Microsoft in 1996, Anders was a Principal Engineer at Borland International. As one of the first employees of Borland, he was the original author of Turbo Pascal and later worked as the Chief Architect of the Delphi product line. Anders studied Engineering at the Technical University of Denmark.

CheckInside: A Fine-grained Indoor Location-based Social Network

Add to EJ Playlist  Google Tech Talk November 19, 2014 Presented by Moustafa Youssef ABSTRACT Existing location-based social networks (LBSNs), e.g. Foursquare, depend mainly on GPS or network-based localization to infer users’ locations. However, GPS is unavailable indoors and network-based localization provides coarse-grained accuracy. This limits the accuracy of current LBSNs in indoor environments, where people spend 89% of their time. This in turn affects the user experience, in terms of the accuracy of the ranked list of venues, especially for the small screens of mobile devices; misses business opportunities; and leads to reduced venues coverage. In this talk, we present CheckInside: a system that can provide a fine-grained indoor location-based social network. CheckInside leverages the crowd-sensed data collected from users’ mobile devices during the check-in operation and knowledge extracted from current LBSNs to associate a place with its name and semantic fingerprint. This semantic fingerprint is used to obtain a more accurate list of nearby places as well as automatically detect new places with similar signatures. A novel algorithm for handling incorrect check-ins and inferring a semantically-en riched floorplan is proposed as well as an algorithm for enhancing the system performance based on the user implicit feedback. Evaluation of CheckInside in four malls over the course of six weeks with 20 participants shows that it can provide the actual user location within the top five venues 99% of the time. This is compared to 17% only in the case of current LBSNs. In addition, it can increase the coverage of current LBSNs by more than 25%. Speaker Info: Moustafa Youssef is an Associate Professor at Egypt-Japan University of Science and Technology (EJUST) and Director of the Wireless Research Center, Egypt. He received his Ph.D. degree in computer science from University of Maryland, USA in 2004 and a B.Sc. and M.Sc. in computer science and engineering from Alexandria University, Egypt in 1997 and 1999 respectively. His research interests include mobile wireless networks, mobile computing, location determination technologies, pervasive computing, and network security. He has twelve issued and pending patents. He is an associate editor for the ACM TSAS, an area editor of the ACM MC2R and served on the organizing and technical committees of numerous conferences. Dr. Youssef is the recipient of the 2003 University of Maryland Invention of the Year award, the 2010 joint TWAS-AAS-Micros oft Award for Young Scientists, the 2012 Egyptian State Award, the 2014 COMESA Innovation Award, among others. He is also an ACM Distinguished Speaker.

Blowing up the (C++11) Atomic Barrier

Add to EJ Playlist  Google Tech Talk October 27, 2014 Presented by Robin Morisset ABSTRACT Atomics in C11 and C++11 let the programmer express the guarantees needed for racy accesses in lock-free code, in theory bringing a zero-cost abstraction for parallelism to the language. This talk will showcase how you can use atomics today and where the abstraction breaks down. We’ll focus on LLVM’s recent improvements for atomics that provide significant performance gains on ARMv7, Power and x86. Finally we’ll discuss some extremely non-intuitive behaviors of atomics, how atomics in C++ may evolve, and how it may impact LLVM. Slides can be found here: https://docs.go tation/d/1Ok3NH ezdnKLD4XSzI8WV wdMC07DR3w87n80 Mj5ydtGE Speaker Info: Robin Morisset interned in Chrome's NaCl team for the summer, and focused on tuning C11/C++11 atomics in LLVM for x86/ARM/Power. He is a PhD Student at the ENS Paris.

GTAC 2014: Going Green: Cleaning up the Toxic Mobile Environment

Add to EJ Playlist  Thomas Knych (Google), Stefan Ramsauer (Google), Valera Zakharov (Google) and Vishal Sethia (Google) We will present tools and techniques for creating fast, stable, hermetic test environments for executing Android tests in both interactive development and continuous integration modes. This builds on the higher level talk we presented at the last GTAC.

GTAC 2014: Beyond Coverage: What Lurks in Test Suites?

Add to EJ Playlist  Patrick Lam (University of Waterloo) We all want "better" test suites. But what makes for a good test suite? Certainly, test suites ought to aim for good coverage, at least at the statement coverage level. To be useful, test suites should run quickly enough to provide timely feedback. This talk will investigate a number of other dimensions on which to evaluate test suites. The talk claims that better test suites are more maintainable, more usable (for instance, because they run faster, or use fewer resources), and have fewer unjustified failures. In this talk, I'll present and synthesize facts about 10 open-source test suites (from 8,000 to 246,000 lines of code) and evaluate how they are doing.

GTAC 2014: The Challenge of Fairly Comparing Cloud Providers and What We're Doing About It

Add to EJ Playlist  Anthony Voellm (Google) This talk will cover the history of benchmarking from mainframe to Cloud. The goal is to lay a foundation around where benchmarks started and how they have gotten to where they are. Ideas will be laid out for the future of benchmarking Cloud and how we can do it practically.

GTAC 2014: Impact of Community Structure on SAT Solver Performance

Add to EJ Playlist  Zack Newsham (University of Waterloo) Modern CDCL SAT solvers routinely solve very large in- dustrial SAT instances in relatively short periods of time. It is clear that these solvers somehow exploit the structure of real-world instances. How- ever, to-date there have been few results that precisely characterise this structure. In this paper, we provide evidence that the community struc- ture of real-world SAT instances is correlated with the running time of CDCL SAT solvers. It has been known for some time that real-world SAT instances, viewed as graphs, have natural communities in them. A community is a sub-graph of the graph of a SAT instance, such that this sub-graph has more internal edges than outgoing to the rest of the graph. The community structure of a graph is often characterised by a quality metric called Q. Intuitively, a graph with high-quality community struc- ture (high Q) is easily separableinto smaller communities, while the one with low Q is not. We provide three results based on empirical data which show that community structure of real-world industrial instances is a bet- ter predictor of the running time of CDCL solvers than other commonly considered factors such as variables and clauses. First, we show that there is a strong correlation between the Q value and Literal Block Distance metric of quality of conflict clauses used in clause-deletion policies in Glucose-like solvers. Second, using regression analysis, we show that the the number of communities and the Q value of the graph of real-world SAT instances is more predictive of the running time of CDCL solvers than traditional metrics like number of variables or clauses. Finally, we show that randomly-genera ted SAT instances with 0.05 ≤ Q ≤ 0.13 are dramatically harder to solve for CDCL solvers than otherwise.

NEWS   Top Videos   Music   Classical   Listen   Funny   Fails   Artist List   Aww   Gaming   Minisode   Science   Technology   TED   TWiT   Trailers   M.M.   PBS   WSJ   AP   CSPAN   CNN   RT   TMZ   E!   ABC   CBS   Politics   Sports   ESPN   WTF   Conspiracy  

Related Uploaders:

GoogleTechTalks Playlists:
GTAC 2014
GTAC 2013
Google Faculty Summit 2012
Online Learning
Alan M. Turing Centennial Conference - Israel
NIPS 2011 Big Learning: Algorithms, System & Tools Workshop
NIPS 2011 Sparse Representation & Low-rank Approx Workshop
NIPS 2011 Music and Machine Learning Workshop
NIPS 2011 Learning Semantics Workshop
NIPS 2011 Domain Adaptation Workshop
Make the Web Faster
GTAC 2011
Google NYC Tech Talks
GTAC 2010
Perspectivas Speaker Series (beta)
GTAC 2009
Google Intenet Summit 2009
Web Exponents
Energy and the Environment