Lifeboat Foundation

MIT researchers demonstrate a way to sharply reduce errors in two-qubit gates, a significant advance toward fully realizing quantum computation.

MIT researchers have made a significant advance on the road toward the full realization of quantum computation, demonstrating a technique that eliminates common errors in the most essential operation of quantum algorithms, the two-qubit operation or “gate.”

“Despite tremendous progress toward being able to perform computations with low error rates with superconducting quantum bits (qubits), errors in two-qubit gates, one of the building blocks of quantum computation, persist,” says Youngkyu Sung, an MIT graduate student in electrical engineering and computer science who is the lead author of a paper on this topic published on June 16, 2021, in Physical Review X. “We have demonstrated a way to sharply reduce those errors.”

Advanced uses of time in image rendering and reconstruction have been the focus of much scientific research in recent years. The motivation comes from the equivalence between space and time given by the finite speed of light c. This equivalence leads to correlations between the time evolution of electromagnetic fields at different points in space. Applications exploiting such correlations, known as time-of-flight (ToF)¹ and light-in-flight (LiF)² cameras, operate at various regimes from radio^3,4 to optical⁵ frequencies. Time-of-flight imaging focuses on reconstructing a scene by measuring delayed stimulus responses via continuous wave, impulses or pseudo-random binary sequence (PRBS) codes¹. Light-in-flight imaging, also known as transient imaging⁶, explores light transport and detection^2,7. The combination of ToF and LiF has recently yielded higher accuracy and detail to the reconstruction process, especially in non-line-of-sight images with the inclusion of higher-order scattering and physical processes such as Rayleigh–Sommerfeld diffraction⁸ in the modeling. However, these methods require experimental characterization of the scene followed by large computational overheads that produce images at low frame rates in the optical regime. In the radio-frequency (RF) regime, 3D images at frame rates of 30 Hz have been produced with an array of 256 wide-band transceivers³. Microwave imaging has the additional capability of sensing through optically opaque media such as walls. Nonetheless, synthetic aperture radar reconstruction algorithms such as the one proposed in ref. ³ required each transceiver in the array to operate individually thus leaving room for improvements in image frame rates from continuous transmit-receive captures. Constructions using beamforming have similar challenges⁹ where a narrow focused beam scans a scene using an array of antennas and frequency modulated continuous wave (FMCW) techniques.

In this article, we develop an inverse light transport model¹⁰ for microwave signals. The model uses a spatiotemporal mask generated by multiple sources, each emitting different PRBS codes, and a single detector, all operating in continuous synchronous transmit-receive mode. This model allows image reconstructions with capture times of the order of microseconds and no prior scene knowledge. For first-order reflections, the algorithm reduces to a single dot product between the reconstruction matrix and captured signal, and can be executed in a few milliseconds. We demonstrate this algorithm through simulations and measurements performed using realistic scenes in a laboratory setting. We then use the second-order terms of the light transport model to reconstruct scene details not captured by the first-order terms.

We start by estimating the information capacity of the scene and develop the light transport equation for the transient imaging model with arguments borrowed from basic information and electromagnetic field theory. Next, we describe the image reconstruction algorithm as a series of approximations corresponding to multiple scatterings of the spatiotemporal illumination matrix. Specifically, we show that in the first-order approximation, the value of each pixel is the dot product between the captured time series and a unique time signature generated by the spatiotemporal electromagnetic field mask. Next, we show how the second-order approximation generates hidden features not accessible in the first-order image. Finally, we apply the reconstruction algorithm to simulated and experimental data and discuss the performance, strengths, and limitations of this technique.

And they say computers can’t create art.

In 1642, famous Dutch painter Rembrandt van Rijn completed a large painting called Militia Company of District II under the Command of Captain Frans Banninck Cocq — today, the painting is commonly referred to as The Night Watch. It was the height of the Dutch Golden Age, and The Night Watch brilliantly showcased that.

The painting measured 363 cm × 437 cm (11.91 ft × 14.34 ft) — so big that the characters in it were almost life-sized, but that’s only the start of what makes it so special. Rembrandt made dramatic use of light and shadow and also created the perception of motion in what would normally be a stationary military group portrait. Unfortunately, though, the painting was trimmed in 1715 to fit between two doors at Amsterdam City Hall.

Continue reading “An AI algorithm just completed a famous Rembrandt painting” »

A team of researchers working at Johannes Kepler University has developed an autonomous drone with a new type of technology to improve search-and-rescue efforts. In their paper published in the journal Science Robotics, the group describes their drone modifications. Andreas Birk with Jacobs University Bremen has published a Focus piece in the same journal issue outlining the work by the team in Austria.

Finding people lost (or hiding) in the forest is difficult because of the tree cover. People in planes and helicopters have difficulty seeing through the canopy to the ground below, where people might be walking or even laying down. The same problem exists for thermal applications—heat sensors cannot pick up readings adequately through the canopy. Efforts have been made to add drones to search-and–rescue operations, but they suffer from the same problems because they are remotely controlled by pilots using them to search the ground below. In this new effort, the researchers have added new technology that both helps to see through the tree canopy and to highlight people that might be under it.

Continue reading “An autonomous drone for search and rescue in forests using optical sectioning algorithm” »

Without GPS, autonomous systems get lost easily. Now a new algorithm developed at Caltech allows autonomous systems to recognize where they are simply by looking at the terrain around them—and for the first time, the technology works regardless of seasonal changes to that terrain.

Details about the process were published on June 23 in the journal Science Robotics.

The general process, known as visual terrain-relative navigation (VTRN), was first developed in the 1960s. By comparing nearby terrain to high-resolution satellite images, autonomous systems can locate themselves.

If you walk down the street shouting out the names of every object you see — garbage truck! bicyclist! sycamore tree! — most people would not conclude you are smart. But if you go through an obstacle course, and you show them how to navigate a series of challenges to get to the end unscathed, they would.

Most machine learning algorithms are shouting names in the street. They perform perceptive tasks that a person can do in under a second. But another kind of AI — deep reinforcement learning — is strategic. It learns how to take a series of actions in order to reach a goal. That’s powerful and smart — and it’s going to change a lot of industries.

Two industries on the cusp of AI transformations are manufacturing and supply chain. The ways we make and ship stuff are heavily dependent on groups of machines working together, and the efficiency and resiliency of those machines are the foundation of our economy and society. Without them, we can’t buy the basics we need to live and work.

Summary: Combining deep learning algorithms with robotic engineering, researchers have developed a new robot able to combine vision and touch.

Source: EBRAINS / human brain project.

Continue reading “A Robot Has Learned to Combine Vision and Touch” »

Chipmaker patches nine high-severity bugs in its Jetson SoC framework tied to the way it handles low-level cryptographic algorithms.

Flaws impacting millions of internet of things (IoT) devices running NVIDIA’s Jetson chips open the door for a variety of hacks, including denial-of-service (DoS) attacks or the siphoning of data.

NVIDIA released patches addressing nine high-severity vulnerabilities including eight additional bugs of less severity. The patches fix a wide swath of NVIDIA’s chipsets typically used for embedded computing systems, machine-learning applications and autonomous devices such as robots and drones.
Impacted products include Jetson chipset series; AGX Xavier, Xavier NX/TX1, Jetson TX2 (including Jetson TX2 NX), and Jetson Nano devices (including Jetson Nano 2GB) found in the NVIDIA JetPack software developers kit. The patches were delivered as part of NVIDIA’s June security bulletin, released Friday.

Last week, I wrote an analysis of Reward Is Enough, a paper by scientists at DeepMind. As the title suggests, the researchers hypothesize that the right reward is all you need to create the abilities associated with intelligence, such as perception, motor functions, and language.

This is in contrast with AI systems that try to replicate specific functions of natural intelligence such as classifying images, navigating physical environments, or completing sentences.

The researchers go as far as suggesting that with well-defined reward, a complex environment, and the right reinforcement learning algorithm, we will be able to reach artificial general intelligence, the kind of problem-solving and cognitive abilities found in humans and, to a lesser degree, in animals.