Computer scientists have dreamt of large-scale quantum computation since at least 1994 -- the hope is that quantum computers will be able to process certain calculations much more quickly than any classical computer, helping to solve problems ranging from complicated physics or chemistry simulations to solving optimization problems to accelerating machine learning tasks.

One of the primary challenges is that quantum memory elements (“qubits”) have always been too prone to errors. They’re fragile and easily disturbed -- any fluctuation or noise from their environment can introduce memory errors, rendering the computations useless. As it turns out, getting even just a small number of qubits together to repeatedly perform the required quantum logic operations and still be nearly error-free is just plain hard. But our team has been developing the quantum logic operations and qubit architectures to do just that.

In our paper “State preservation by repetitive error detection in a superconducting quantum circuit”, published in the journal Nature, we describe a superconducting quantum circuit with nine qubits where, for the first time, the qubits are able to detect and effectively protect each other from bit errors. This quantum error correction (QEC) can overcome memory errors by applying a carefully choreographed series of logic operations on the qubits to detect where errors have occurred.
Photograph of the device containing nine quantum bits (qubits). Each qubit interacts with its neighbors to protect them from error.

So how does QEC work? In a classical computer, we can monitor bits directly to detect errors. However, qubits are much more fickle -- measuring a qubit directly will collapse entanglement and superposition states, removing the quantum elements that make it useful for computation.

To get around this, we introduce additional ‘measurement’ qubits, and perform a series of quantum logic operations that look at the 'measurement' and 'data' qubits in combination. By looking at the state of these pairwise combinations (using quantum XOR gates), and performing some careful cross-checking, we can pull out just enough information to detect errors without altering the information in any individual qubit.
The basics of error correction. ‘Measurement’ qubits can detect errors on ‘data’ qubits through the use of quantum XOR gates.

We’ve also shown that storing information in five qubits works better than just storing it in one, and that with nine qubits the error correction works even better. That’s a key result -- it shows that the quantum logic operations are trustworthy enough that by adding more qubits, we can detect more complex errors that otherwise may cause algorithmic failure.

While the basic physical processes behind quantum error correction are feasible, many challenges remain, such as improving the logic operations behind error correction and testing protection from phase-flip errors. We’re excited to tackle these challenges on the way towards making real computations possible.


Discovering new treatments for human diseases is an immensely complicated challenge; Even after extensive research to develop a biological understanding of a disease, an effective therapeutic that can improve the quality of life must still be found. This process often takes years of research, requiring the creation and testing of millions of drug-like compounds in an effort to find a just a few viable drug treatment candidates. These high-throughput screens are often automated in sophisticated labs and are expensive to perform.

Recently, deep learning with neural networks has been applied in virtual drug screening1,2,3, which attempts to replace or augment the high-throughput screening process with the use of computational methods in order to improve its speed and success rate.4 Traditionally, virtual drug screening has used only the experimental data from the particular disease being studied. However, as the volume of experimental drug screening data across many diseases continues to grow, several research groups have demonstrated that data from multiple diseases can be leveraged with multitask neural networks to improve the virtual screening effectiveness.

In collaboration with the Pande Lab at Stanford University, we’ve released a paper titled "Massively Multitask Networks for Drug Discovery", investigating how data from a variety of sources can be used to improve the accuracy of determining which chemical compounds would be effective drug treatments for a variety of diseases. In particular, we carefully quantified how the amount and diversity of screening data from a variety of diseases with very different biological processes can be used to improve the virtual drug screening predictions.

Using our large-scale neural network training system, we trained at a scale 18x larger than previous work with a total of 37.8M data points across more than 200 distinct biological processes. Because of our large scale, we were able to carefully probe the sensitivity of these models to a variety of changes in model structure and input data. In the paper, we examine not just the performance of the model but why it performs well and what we can expect for similar models in the future. The data in the paper represents more than 50M total CPU hours.
This graph shows a measure of prediction accuracy (ROC AUC is the area under the receiver operating characteristic curve) for virtual screening on a fixed set of 10 biological processes as more datasets are added.

One encouraging conclusion from this work is that our models are able to utilize data from many different experiments to increase prediction accuracy across many diseases. To our knowledge, this is the first time the effect of adding additional data has been quantified in this domain, and our results suggest that even more data could improve performance even further.

Machine learning at scale has significant potential to accelerate drug discovery and improve human health. We look forward to continued improvement in virtual drug screening and its increasing impact in the discovery process for future drugs.

Thank you to our other collaborators David Konerding (Google), Steven Kearnes (Stanford), and Vijay Pande (Stanford).


1. Thomas Unterthiner, Andreas Mayr, Günter Klambauer, Marvin Steijaert, Jörg Kurt Wegner, Hugo Ceulemans, Sepp Hochreiter. Deep Learning as an Opportunity in Virtual Screening. Deep Learning and Representation Learning Workshop: NIPS 2014

2. Dahl, George E, Jaitly, Navdeep, and Salakhutdinov, Ruslan. Multi-task neural networks for QSAR predictions. arXiv preprint arXiv:1406.1231, 2014.

3. Ma, Junshui, Sheridan, Robert P, Liaw, Andy, Dahl, George, and Svetnik, Vladimir. Deep neural nets as a method for quantitative structure-activity relationships. Journal of Chemical Information and Modeling, 2015.

4. Peter Ripphausen, Britta Nisius, Lisa Peltason, and Jürgen Bajorath. Quo Vadis, Virtual Screening? A Comprehensive Survey of Prospective Applications. Journal of Medicinal Chemistry 2010 53 (24), 8461-8467


Remember the classic videogame Breakout on the Atari 2600? When you first sat down to try it, you probably learned to play well pretty quickly, because you already knew how to bounce a ball off a wall in real life. You may have even worked up a strategy to maximise your overall score at the expense of more immediate rewards. But what if you didn't possess that real-world knowledge — and only had the pixels on the screen, the control paddle in your hand, and the score to go on? How would you, or equally any intelligent agent faced with this situation, learn this task totally from scratch?

This is exactly the question that we set out to answer in our paper “Human-level control through deep reinforcement learning”, published in Nature this week. We demonstrate that a novel algorithm called a deep Q-network (DQN) is up to this challenge, excelling not only at Breakout but also a wide variety of classic videogames: everything from side-scrolling shooters (River Raid) to boxing (Boxing) and 3D car racing (Enduro). Strikingly, DQN was able to work straight “out of the box” across all these games – using the same network architecture and tuning parameters throughout and provided only with the raw screen pixels, set of available actions and game score as input.

The results: DQN outperformed previous machine learning methods in 43 of the 49 games. In fact, in more than half the games, it performed at more than 75% of the level of a professional human player. In certain games, DQN even came up with surprisingly far-sighted strategies that allowed it to achieve the maximum attainable score—for example, in Breakout, it learned to first dig a tunnel at one end of the brick wall so the ball could bounce around the back and knock out bricks from behind.


So how does it work? DQN incorporated several key features that for the first time enabled the power of Deep Neural Networks (DNN) to be combined in a scalable fashion with Reinforcement Learning (RL)—a machine learning framework that prescribes how agents should act in an environment in order to maximize future cumulative reward (e.g., a game score). Foremost among these was a neurobiologically inspired mechanism, termed “experience replay,” whereby during the learning phase DQN was trained on samples drawn from a pool of stored episodes—a process physically realized in a brain structure called the hippocampus through the ultra-fast reactivation of recent experiences during rest periods (e.g., sleep). Indeed, the incorporation of experience replay was critical to the success of DQN: disabling this function caused a severe deterioration in performance.
Comparison of the DQN agent with the best reinforcement learning methods in the literature. The performance of DQN is normalized with respect to a professional human games tester (100% level) and random play (0% level). Note that the normalized performance of DQN, expressed as a percentage, is calculated as: 100 X (DQN score - random play score)/(human score - random play score). Error bars indicate s.d. across the 30 evaluation episodes, starting with different initial conditions. Figure courtesy of Mnih et al. “Human-level control through deep reinforcement learning”, Nature 26 Feb. 2015.
This work offers the first demonstration of a general purpose learning agent that can be trained end-to-end to handle a wide variety of challenging tasks, taking in only raw pixels as inputs and transforming these into actions that can be executed in real-time. This kind of technology should help us build more useful products—imagine if you could ask the Google app to complete any kind of complex task (“Okay Google, plan me a great backpacking trip through Europe!”).

We also hope this kind of domain general learning algorithm will give researchers new ways to make sense of complex large-scale data creating the potential for exciting discoveries in fields such as climate science, physics, medicine and genomics. And it may even help scientists better understand the process by which humans learn. After all, as the great physicist Richard Feynman famously said: “What I cannot create, I do not understand.”


We have just completed another round of the Google Faculty Research Awards, our biannual open call for research proposals on Computer Science and related topics, including systems, machine perception, structured data, robotics, and mobile. Our grants cover tuition for a graduate student and provide both faculty and students the opportunity to work directly with Google researchers and engineers.

This round we received 808 proposals, an increase of 12% over last round, covering 55 countries on 6 continents. After expert reviews and committee discussions, we decided to fund 122 projects, with 20% of the funding awarded to universities outside the U.S. The subject areas that received the highest level of support were systems, human-computer interaction, and machine perception.

The Faculty Research Award program enables us to build strong relationships with faculty around the world who are pursuing innovative research, and plays an important role for Google’s Research organization by fostering an exchange of ideas that advances the state of the art. Each round, we receive proposals from faculty who may be just starting their careers, or who might be experimenting in new areas that help us look forward and innovate on what's emerging in the CS community.

Congratulations to the well-deserving recipients of this round’s awards. If you are interested in applying for the next round (deadline is April 15), please visit our website for more information.


(Cross-posted from the Google for Education Blog)

Science is about observing and experimenting. It’s about exploring unanswered questions, solving problems through curiosity, learning as you go and always trying again.

That’s the spirit behind the fifth annual Google Science Fair, kicking off today. Together with LEGO Education, National Geographic, Scientific American and Virgin Galactic, we’re calling on all young researchers, explorers, builders, technologists and inventors to try something ambitious. Something imaginative, or maybe even unimaginable. Something that might just change the world around us.

From now through May 18, students around the world ages 13-18 can submit projects online across all scientific fields, from biology to computer science to anthropology and everything in between. Prizes include $100,000 in scholarships and classroom grants from Scientific American and Google, a National Geographic Expedition to the Galapagos, an opportunity to visit LEGO designers at their Denmark headquarters, and the chance to tour Virgin Galactic’s new spaceship at their Mojave Air and Spaceport. This year we’re also introducing an award to recognize an Inspiring Educator, as well as a Community Impact Award honoring a project that addresses an environmental or health challenge.

It’s only through trying something that we can get somewhere. Flashlights required batteries, then Ann Makosinski tried the heat of her hand. His grandfather would wander out of bed at night, until Kenneth Shinozuka tried a wearable sensor. The power supply was constantly unstable in her Indian village, so Harine Ravichandran tried to build a different kind of regulator. Previous Science Fair winners have blown us away with their ideas. Now it’s your turn.

Big ideas that have the potential to make a big impact often start from something small. Something that makes you curious. Something you love, you’re good at, and want to try.

So, what will you try?


In 2009, Google created the PhD Fellowship program to recognize and support outstanding graduate students doing exceptional work in Computer Science (CS) and related disciplines. In that time we’ve seen past recipients add depth and breadth to CS by developing new ideas and research directions, from building new intelligence models to changing the way in which we interact with computers to advancing into faculty positions, where they go on to train the next generation of researchers.

Reflecting our continuing commitment to building strong relations with the global academic community, we are excited to announce the latest North American Google PhD Fellows. The following 15 fellowship recipients were chosen from a highly competitive group, and represent the outstanding quality of nominees provided by our university partners:

  • Justin Meza, Google US/Canada Fellowship in Systems Reliability (Carnegie Mellon University)
  • Waleed Ammar, Google US/Canada Fellowship in Natural Language Processing (Carnegie Mellon University)
  • Aaron Parks, Google US/Canada Fellowship in Mobile Networking (University of Washington)
  • Kyle Rector, Google US/Canada Fellowship in Human Computer Interaction (University of Washington)
  • Nick Arnosti, Google US/Canada Fellowship in Market Algorithms (Stanford University)
  • Osbert Bastani, Google US/Canada Fellowship in Programming Languages (Stanford University)
  • Carl Vondrick, Google US/Canada Fellowship in Machine Perception, (Massachusetts Institute of Technology)
  • Wojciech Zaremba, Google US/Canada Fellowship in Machine Learning (New York University)
  • Xiaolan Wang, Google US/Canada Fellowship in Structured Data (University of Massachusetts Amherst)
  • Muhammad Naveed, Google US/Canada Fellowship in Security (University of Illinois at Urbana-Champaign)
  • Masoud Moshref Javadi, Google US/Canada Fellowship in Computer Networking (University of Southern California)
  • Riley Spahn, Google US/CanadaFellowship in Privacy (Columbia University)
  • Saurabh Gupta, Google US/Canada Fellowship in Computer Vision (University of California, Berkeley)
  • Yun Teng, Google US/Canada Fellowship in Computer Graphics (University of California, Santa Barbara)
  • Tan Zhang, Google US/Canada Fellowship in Mobile Systems (University of Wisconsin-Madison)

This group of students represent the next generation of researchers who endeavor to solve some of the most interesting challenges in Computer Science. We offer our congratulations, and look forward to their future contributions to the research community with high expectations.


Nature reserves have a vital role for protecting biodiversity and its many functions. However, there is often insufficient information available to determine where to most effectively invest conservation efforts to prevent future extinctions, or which species may be left out of conservation actions entirely.

To help address these issues, Map of Life, in collaboration with Google Earth Engine, has now pre-released a new service to pinpoint at-risk species and where in the world that they occur. At the fingertips of regional naturalists, conservation groups, resource managers and global threat assessors, the tool has the potential to help identify and close key information gaps and highlight species of greatest concern.

Take the Tamaulipas Pygmy Owl, one of the smallest owls in the world that is restricted to highland forests in Mexico. The consensus range map for the species indicates a broad distribution of over 50,000 km2:
Left: Tamaulipas Pygmy Owl (Glaucidium sanchezi, photo credit: Adam Kent). Right: Map of Life consensus range map showing the potentially habitable range of this species.

But accounting for available habitat in the area using remotely sensed information presents a different picture: less than 10% of this range are forested and at the suitable elevation.
Users can change the habitat association settings and explore on-the-fly how this affects the distribution and map quality. This refined range map now allows a much improved evaluation of the owl’s potential protection. Furthermore, the sensitivity of conservation assessments to various assumptions can be directly explored in this tool.
The owl’s potential protection is likely to occur in only around 1,000 km2 that are under formal protection, representing seven reserves of which only two have greater than 100 km2 area. This is much less than would be desirable for a species with this small a global range.

Another species example, the Hildegard’s Tomb Bat, is similarly concerning: less than 6,000 km2 of suitable range remains for this forest specialist in East Africa, with less than half currently under protection.

A demonstration of this tool for 15 example species was pre-released at the decadal World Parks Congress in Sydney Australia last November to the global community of conservation scientists and practitioners. In the coming months this interactive evaluation will be expanded to thousands more species, providing a valuable resource to aid in global conservation efforts. For more information and updates, follow Map of Life.