Toggle light / dark theme

Is deep learning a necessary ingredient for artificial intelligence?

The earliest artificial neural network, the Perceptron, was introduced approximately 65 years ago and consisted of just one layer. However, to address solutions for more complex classification tasks, more advanced neural network architectures consisting of numerous feedforward (consecutive) layers were later introduced. This is the essential component of the current implementation of deep learning algorithms. It improves the performance of analytical and physical tasks without human intervention, and lies behind everyday automation products such as the emerging technologies for self-driving cars and autonomous chat bots.

The key question driving new research published today in Scientific Reports is whether efficient learning of non-trivial classification tasks can be achieved using brain-inspired shallow feedforward networks, while potentially requiring less .

“A positive answer questions the need for deep learning architectures, and might direct the development of unique hardware for the efficient and fast implementation of shallow learning,” said Prof. Ido Kanter, of Bar-Ilan’s Department of Physics and Gonda (Goldschmied) Multidisciplinary Brain Research Center, who led the research. “Additionally, it would demonstrate how brain-inspired shallow learning has advanced computational capability with reduced complexity and energy consumption.”

A defence of human uniqueness against AI encroachment, with Kenn Cukier

Despite the impressive recent progress in AI capabilities, there are reasons why AI may be incapable of possessing a full “general intelligence”. And although AI will continue to transform the workplace, some important jobs will remain outside the reach of AI. In other words, the Economic Singularity may not happen, and AGI may be impossible.

These are views defended by our guest in this episode, Kenneth Cukier, the Deputy Executive Editor of The Economist newspaper.

For the past decade, Kenn was the host of its weekly tech podcast Babbage. He is co-author of the 2013 book “Big Data”, a New York Times best-seller that has been translated into over 20 languages. He is a regular commentator in the media, and a popular keynote speaker, from TED to the World Economic Forum.

A neuromorphic visual sensor can recognize moving objects and predict their path

A new bio-inspired sensor can recognize moving objects in a single frame from a video and successfully predict where they will move to. This smart sensor, described in a Nature Communications paper, will be a valuable tool in a range of fields, including dynamic vision sensing, automatic inspection, industrial process control, robotic guidance, and autonomous driving technology.

Current motion detection systems need many components and complex algorithms doing frame-by-frame analyses, which makes them inefficient and energy-intensive. Inspired by the human visual system, researchers at Aalto University have developed a new neuromorphic vision technology that integrates sensing, memory, and processing in a single device that can detect motion and predict trajectories.

At the core of their technology is an array of photomemristors, that produce in response to light. The current doesn’t immediately stop when the light is switched off. Instead, it decays gradually, which means that photomemristors can effectively “remember” whether they’ve been exposed to light recently. As a result, a sensor made from an array of photomemristors doesn’t just record instantaneous information about a scene, like a camera does, but also includes a dynamic memory of the preceding instants.

Room-temperature superfluidity in a polariton condensate Physics

face_with_colon_three year 2017.


First observed in liquid helium below the lambda point, superfluidity manifests itself in a number of fascinating ways. In the superfluid phase, helium can creep up along the walls of a container, boil without bubbles, or even flow without friction around obstacles. As early as 1938, Fritz London suggested a link between superfluidity and Bose–Einstein condensation (BEC)3. Indeed, superfluidity is now known to be related to the finite amount of energy needed to create collective excitations in the quantum liquid4,5,6,7, and the link proposed by London was further evidenced by the observation of superfluidity in ultracold atomic BECs1,8. A quantitative description is given by the Gross–Pitaevskii (GP) equation9,10 (see Methods) and the perturbation theory for elementary excitations developed by Bogoliubov11. First derived for atomic condensates, this theory has since been successfully applied to a variety of systems, and the mathematical framework of the GP equation naturally leads to important analogies between BEC and nonlinear optics12,13,14. Recently, it has been extended to include condensates out of thermal equilibrium, like those composed of interacting photons or bosonic quasiparticles such as microcavity exciton-polaritons and magnons14,15. In particular, for exciton-polaritons, the observation of many-body effects related to condensation and superfluidity such as the excitation of quantized vortices, the formation of metastable currents and the suppression of scattering from potential barriers2,16,17,18,19,20 have shown the rich phenomenology that exists within non-equilibrium condensates. Polaritons are confined to two dimensions and the reduced dimensionality introduces an additional element of interest for the topological ordering mechanism leading to condensation, as recently evidenced in ref. 21. However, until now, such phenomena have mainly been observed in microcavities embedding quantum wells of III–V or II–VI semiconductors. As a result, experiments must be performed at low temperatures (below ∼ 20 K), beyond which excitons autoionize. This is a consequence of the low binding energy typical of Wannier–Mott excitons. Frenkel excitons, which are characteristic of organic semiconductors, possess large binding energies that readily allow for strong light–matter coupling and the formation of polaritons at room temperature. Remarkably, in spite of weaker interactions as compared to inorganic polaritons22, condensation and the spontaneous formation of vortices have also been observed in organic microcavities23,24,25. However, the small polariton–polariton interaction constants, structural inhomogeneity and short lifetimes in these structures have until now prevented the observation of behaviour directly related to the quantum fluid dynamics (such as superfluidity). In this work, we show that superfluidity can indeed be achieved at room temperature and this is, in part, a result of the much larger polariton densities attainable in organic microcavities, which compensate for their weaker nonlinearities.

Our sample consists of an optical microcavity composed of two dielectric mirrors surrounding a thin film of 2,7-Bis[9,9-di(4-methylphenyl)-fluoren-2-yl]-9,9-di(4-methylphenyl)fluorene (TDAF) organic molecules. Light–matter interaction in this system is so strong that it leads to the formation of hybrid light–matter modes (polaritons), with a Rabi energy 2 ΩR ∼ 0.6 eV. A similar structure has been used previously to demonstrate polariton condensation under high-energy non-resonant excitation24. Upon resonant excitation, it allows for the injection and flow of polaritons with a well-defined density, polarization and group velocity.

The experimental configuration is shown in Fig. 1a. The sample is positioned between two microscope objectives to allow for measurements in a transmission geometry while maintaining high spatial resolution. A polariton wavepacket with a chosen wavevector is created by exciting the sample with a linearly polarized 35 fs laser pulse resonant with the lower polariton branch (see Methods). By detecting the reflected or transmitted light using a spectrometer and a charge-coupled device (CCD) camera, energy-resolved space and momentum maps can be acquired. An example of the experimental polariton dispersion under white light illumination is shown in Fig. 1b. The parabolic TE-and TM-polarized lower polariton branches appear as dips in the reflectance spectra. The figure also shows an example of how the laser energy, momentum and polarization can be precisely tuned to excite, in this case, the TE lower polariton branch at a given angle.

Algorithms auditing algorithms: GPT-4 a reminder that responsible AI is moving beyond human scale

Join top executives in San Francisco on July 11–12, to hear how leaders are integrating and optimizing AI investments for success. Learn More

Artificial intelligence (AI) is revolutionizing industries, streamlining processes, and, hopefully, on its way to improving the quality of life for people around the world — all very exciting news. That said, with the increasing influence of AI systems, it’s crucial to ensure that these technologies are developed and implemented responsibly.

Responsible AI is not just about adhering to regulations and ethical guidelines; it is the key to creating more accurate and effective AI models.

Mixed Reality Tracking System For Future Pilot Training

Vrgineers and Advanced Realtime Tracking demonstrate the combination of XTAL 3 headset and SMARTTRACK3/M in a mixed reality pilot trainer. The partnership between these two technological companies started in 2018. At IT2EC 2023 in Rotterdam, the integrated SMARTTRACK3/M into an F-35-like Classroom Trainer manufactured and delivered to USAF and RAF will be for display. This unique combination of the latest ART infrared all-in-one hardware and Vrgineers algorithms for cockpit motion compensation creates an unseen immersion for every mixed reality training. One of the challenges in next-generation pilot training using virtual technology and motion platforms is the alignment of the pilot’s position in the cockpit. By overcoming this issue, the simulator industry is moving forward to eliminate the disadvantages of simulated training.

“We are continuously working on removing the technological challenges of modern simulators, one of which is caused by front-facing camera position distance from users’ eyes. We are developing advanced algorithms for motion compensation to minimize the shift between virtual and physical scene, making experience realistic. The durability and compact size of SMARTTRACK3/M, which was optimized for using in cockpits, allows us as training device integrator to make it a comprehensive part of a simulation,” says Marek Polcak, CEO of Vrgineers.

“This is the application SMARTTRACK3/M was designed for., We have taken the proven hardware from the SMARTTRACK3 and adapted it to the limited space available. As a result, we have the precision and the reliability of a seasoned system in a form factor fitting to simulator cockpits” says Andreas Werner, business development manager for simulations at ART.

Cortico-Hippocampal Computational Modeling Using Quantum-Inspired Neural Networks

Many current computational models that aim to simulate cortical and hippocampal modules of the brain depend on artificial neural networks. However, such classical or even deep neural networks are very slow, sometimes taking thousands of trials to obtain the final response with a considerable amount of error. The need for a large number of trials at learning and the inaccurate output responses are due to the complexity of the input cue and the biological processes being simulated. This article proposes a computational model for an intact and a lesioned cortico-hippocampal system using quantum-inspired neural networks. This cortico-hippocampal computational quantum-inspired (CHCQI) model simulates cortical and hippocampal modules by using adaptively updated neural networks entangled with quantum circuits. The proposed model is used to simulate various classical conditioning tasks related to biological processes. The output of the simulated tasks yielded the desired responses quickly and efficiently compared with other computational models, including the recently published Green model.

Several researchers have proposed models that combine artificial neural networks (ANNs) or quantum neural networks (QNNs) with various other ingredients. For example, Haykin (1999) and Bishop (1995) developed multilevel activation function QNNs using the quantum linear superposition feature (Bonnell and Papini, 1997).

The prime factorization algorithm of Shor was used to illustrate the basic workings of QNNs (Shor, 1994). Shor’s algorithm uses quantum computations by quantum gates to provide the potential power for quantum computers (Bocharov et al., 2017; Dridi and Alghassi, 2017; Demirci et al., 2018; Jiang et al., 2018). Meanwhile, the work of Kak (1995) focused on the relationship between quantum mechanics principles and ANNs. Kak introduced the first quantum network based on the principles of neural networks, combining quantum computation with convolutional neural networks to produce quantum neural computation (Kak, 1995; Zhou, 2010). Since then, a myriad of QNN models have been proposed, such as those of Zhou (2010) and Schuld et al. (2014).

QuASeR: Quantum Accelerated de novo DNA sequence reconstruction

In this, we present QuASeR, a reference-free DNA sequence reconstruction implementation via de novo assembly on both gate-based and quantum annealing platforms. This is the first time this important application in bioinformatics is modeled using quantum computation. Each one of the four steps of the implementation (TSP, QUBO, Hamiltonians and QAOA) is explained with a proof-of-concept example to target both the genomics research community and quantum application developers in a self-contained manner. The implementation and results on executing the algorithm from a set of DNA reads to a reconstructed sequence, on a gate-based quantum simulator, the D-Wave quantum annealing simulator and hardware are detailed. We also highlight the limitations of current classical simulation and available quantum hardware systems. The implementation is open-source and can be found on https://github.com/QE-Lab/QuASeR.

Citation: Sarkar A, Al-Ars Z, Bertels K (2021) QuASeR: Quantum Accelerated de novo DNA sequence reconstruction. PLoS ONE 16: e0249850. https://doi.org/10.1371/journal.pone.

Editor: Archana Kamal, University of Massachusetts Lowell, UNITED STATES.