Toggle light / dark theme

IBM Neuromorphic chip hits DARPA milestone and has been used to implement deep learning

IBM delivered on the DARPA SyNAPSE project with a one million neuron brain-inspired processor. The chip consumes merely 70 milliwatts, and is capable of 46 billion synaptic operations per second, per watt–literally a synaptic supercomputer in your palm.

Along the way—progressing through Phase 0, Phase 1, Phase 2, and Phase 3—we have journeyed from neuroscience to supercomputing, to a new computer architecture, to a new programming language, to algorithms, applications, and now to a new chip—TrueNorth.

Fabricated in Samsung’s 28nm process, with 5.4 billion transistors, TrueNorth is IBM’s largest chip to date in transistor count. While simulating complex recurrent neural networks, TrueNorth consumes less than 100mW of power and has a power density of 20mW / cm2.

[AI Lab] Pepper robot learning “ball in a cup”

This video realized by the AI Lab of SoftBank Robotics shows how Pepper robot learns to play the ball-in-a-cup game (“bilboquet” in French). The movement is first demonstrated to the robot by guiding its arm.

From there, Pepper has to improve its performance through trial-and-error learning. Even though the initial demonstration does not land the ball in the cup, Pepper can still learn to play the game successfully.

The movement is represented as a so-called dynamic movement primitive and optimized using an evolutionary algorithm. Our implementation uses the freely available software library dmpbbo: https://github.com/stulp/dmpbbo.

After 100 trials, Pepper has successfully optimized its behavior and is able to repeatedly land the ball in the cup.

Scientists hail new era of ‘motherless babies’

For 200 years, our knowledge of reproduction has been clear: sperm + egg = baby. But scientists say they may have found a way to create babies with two biological dads. Should we celebrate?

Which came first: the chicken or the egg? It is a question pondered since the time of Ancient Greece, when Aristotle decided that the answer must be both.

Now, scientists say it could be possible to remove the egg from the equation all together. Dr Tony Perry and his team announced this week that they have successfully bred mice without using a normal egg cell. Instead, they used sperm to fertilise a kind of non-viable embryo called a parthenogenote, which multiplies more like a normal cell. Then they ‘tricked’ it into developing into an embryo using special chemicals, planted it into a surrogate, and a new mouse was born. It survived, and has even gone on to have offspring of its own.

Mach Effect Propulsion Theory Updates

Theory of a mach effect thruster I

The Mach Effect Thruster (MET) is a propellant—less space drive which uses Mach’s principle to produce thrust in an accelerating material which is undergoing mass—energy fluctuations. Mach’s principle is a statement that the inertia of a body is the result of the gravitational interaction of the body with the rest of the mass-energy in the universe. The MET device uses electric power of 100 — 200 Watts to operate. The thrust produced by these devices, at the present time, are small on the order of a few micro-Newtons. Researchers give a physical description of the MET device and apparatus for measuring thrusts. Next they explain the basic theory behind the device which involves gravitation and advanced waves to incorporate instantaneous action at a distance. The advanced wave concept is a means to conserve momentum of the system with the universe. There is no momentun violation in this theory. We briefly review absorber theory by summarizing Dirac, Wheeler-Feynman and Hoyle-Narlikar (HN). They show how Woodward’s mass fluctuation formula can be derived from first principles using the HN-theory which is a fully Machian version of Einstein’s relativity. HN-theory reduces to Einstein’s field equations in the limit of smooth fluid distribution of matter and a simple coordinate transformation.

It is shown that if Mach’s Principle is taken seriously, and the inertia of a body can be described as the interaction of the body with the rest of the universe, then the advanced and retarded fields transmitted between the particle and the universe can be used to explain the thrust observed in the Mach Effect drive experiments. This idea was originally put forward by one of the authors, James Woodward. The idea of inertia being a gravitational effect was first postulated by Einstein. In fact Mach’s principle was the foundation on which Einstein’s general relativity was based.

Physicists Made a ‘Black Hole’ in a Lab That May Finally Prove Hawking Radiation Exists

Scientists may have found signs that phonons, the very small packets of energy that make up sound waves, were leaking out of sonic black holes, just as Hawking’s equations predicted.

Some 42 years ago, renowned theoretical physicist Stephen Hawking proposed that not everything that comes in contact with a black hole succumbs to its unfathomable nothingness. Tiny particles of light (photons) are sometimes ejected back out, robbing the black hole of an infinitesimal amount of energy, and this gradual loss of mass over time means every black hole eventually evaporates out of existence.

Known as Hawking radiation, these escaping particles help us make sense of one of the greatest enigmas in the known Universe, but after more than four decades, no one’s been able to actually prove they exist, and Hawking’s proposal remained firmly in hypothesis territory.

Reinforcement Learning for Torch: Introducing torch-twrl

Advances in machine learning have been driven by innovations and ideas from many fields. Inspired by the way that humans learn, Reinforcement Learning (RL) is concerned with algorithms which improve with trial-and-error feedback to optimize future performance.

Board games and video games often have well-defined reward functions which allow for straightforward optimization with RL algorithms. Algorithmic advances have allowed for RL to be in real-world problems, such as high degree-of-freedom robotic manipulation and large-scale recommendation tasks, with more complex goals.

Twitter Cortex invests in novel state-of-the-art machine learning methods to improve the quality of our products. We are exploring RL as a learning paradigm, and to that end, Twitter Cortex built a framework for RL development. Today, Twitter is open sourcing torch-twrl to the world.

New technology may help read brain signals directly

Nice.


Researchers have developed a new technology that can help read brain signals directly and may also aid people with movement disabilities to better communicate their thoughts and emotions. The technology involves a multi-electrode array implanted in the brain to directly read signals from a region that ordinarily directs hand and arm movements used, for example, to move a computer mouse.

The algorithms translate those signals and help to make letter selections.

“Our results demonstrate that this interface may have great promise for use in people as it enables a typing rate sufficient for a meaningful conversation,” said Paul Nuyujukian, postdoctoral student at Stanford University in California, US.

Quantum Mechanics Revisited: Physicists Propose New Structure of Time

Read a little further into the paper, and things get really weird. If the equations of quantum mechanics must be altered in accordance with the new research, then it will give rise to a new and very curious definition of time.

Time is, essentially, a “crystal”—a highly organized lattice of discrete “particles,” or regularly repeating segments.

“The physical universe is really like a movie/motion picture, in which a series of still images shown on a screen creates the illusion of moving images,” said Mir Faizal of the University of Waterloo and the University of Lethbridge in Canada, and lead author of the paper.

8 Takes on the Rise of AI and Its Implications

In recent years, it’s been exciting watching advances in AI like IBM’s Watson smashing humans at Jeopardy and Google’s AlphaGo AI beating champions at the game of Go a decade earlier than expected. But the sophisticated algorithms under the hood are really the stars of the show.

These powerful computing systems are fundamentally changing industries and automating a growing number of day-to-day tasks. At the same time, AI still isn’t perfect, and we’ve seen hints of its potential dark side. Our algorithms are only as good as the data we feed them. And there’s been a spirited debate about existential dangers down the road.

Here’s a look into some of the topics leading the dialogue as AI technology evolves into its next generation.

Quantum Cosmology and the Evolution of Inflationary Spectra [CL]

We illustrate how it is possible to calculate the quantum gravitational effects on the spectra of primordial scalar/tensor perturbations starting from the canonical, Wheeler-De Witt, approach to quantum cosmology. The composite matter-gravity system is analysed through a Born-Oppenheimer approach in which gravitation is associated with the heavy degrees of freedom and matter (here represented by a scalar field) with the light ones. Once the independent degrees of freedom are identified the system is canonically quantised. The differential equation governing the dynamics of the primordial spectra with its quantum-gravitational corrections is then obtained and is applied to diverse inflationary evolutions. Finally, the analytical results are compared to observations through a Monte Carlo Markov Chain technique and an estimate of the free parameters of our approach is finally presented and the results obtained are compared with previous ones.

Read this paper on arXiv…

A. Kamenshchik, A. Tronconi and G. Venturi Tue, 13 Sep 16 11/91.