Lifeboat Foundation

A DeepMind research group conducted a comprehensive generalization study on neural network architectures in the paper ‘Neural Networks and the Chomsky Hierarchy’, which investigates whether insights from the theory of computation and the Chomsky hierarchy can predict the actual limitations of neural network generalization.

While we understand that developing powerful machine learning models requires an accurate generalization to out-of-distribution inputs. However, how and why neural networks can generalize on algorithmic sequence prediction tasks is unclear.

The research group performed a thorough generalization study on more than 2000 individual models spread across 16 tasks of cutting-edge neural network architectures and memory-augmented neural networks on a battery of sequence-prediction tasks encompassing all tiers of the Chomsky hierarchy that can be evaluated practically with finite-time computation.

Using reinforcement learning (RL) to train robots directly in real-world environments has been considered impractical due to the huge amount of trial and error operations typically required before the agent finally gets it right. The use of deep RL in simulated environments has thus become the go-to alternative, but this approach is far from ideal, as it requires designing simulated tasks and collecting expert demonstrations. Moreover, simulations can fail to capture the complexities of real-world environments, are prone to inaccuracies, and the resulting robot behaviours will not adapt to real-world environmental changes.

The Dreamer algorithm proposed by Hafner et al. at ICLR 2020 introduced an RL agent capable of solving long-horizon tasks purely via latent imagination. Although Dreamer has demonstrated its potential for learning from small amounts of interaction in the compact state space of a learned world model, learning accurate real-world models remains challenging, and it was unknown whether Dreamer could enable faster learning on physical robots.

In the new paper DayDreamer: World Models for Physical Robot Learning, Hafner and a research team from the University of California, Berkeley leverage recent advances in the Dreamer world model to enable online RL for robot training without simulators or demonstrations. The novel approach achieves promising results and establishes a strong baseline for efficient real-world robot training.

In today’s business world, machine-learning algorithms are increasingly being applied to decision-making processes, which affects employment, education, and access to credit. But firms usually keep algorithms secret, citing concerns over gaming by users that can harm the predictive power of algorithms. Amid growing calls to require firms to make their algorithms transparent, a new study developed an analytical model to compare the profit of firms with and without such transparency. The study concluded that there are benefits but also risks in algorithmic transparency.

Conducted by researchers at Carnegie Mellon University (CMU) and the University of Michigan, the study appears in Management Science.

“As managers face calls to boost transparency, our findings can help them make decisions to benefit their firms,” says Param Vir Singh, Professor of Business Technologies and Marketing at CMU’s Tepper School of Business, who coauthored the study.

A computer algorithm can use a technique called “ghost imaging” to reconstruct objects from a person’s brainwaves that the person themselves can’t see.

Text-to-image generation is the hot algorithmic process right now, with OpenAI’s Craiyon (formerly DALL-E mini) and Google’s Imagen AIs unleashing tidal waves of wonderfully weird procedurally generated art synthesized from human and computer imaginations. On Tuesday, Meta revealed that it too has developed an AI image generation engine, one that it hopes will help to build immersive worlds in the Metaverse and create high digital art.

A lot of work into creating an image based on just the phrase, “there’s a horse in the hospital,” when using a generation AI. First the phrase itself is fed through a transformer model, a neural network that parses the words of the sentence and develops a contextual understanding of their relationship to one another. Once it gets the gist of what the user is describing, the AI will synthesize a new image using a set of GANs (generative adversarial networks).

Continue reading “Meta’s ‘Make-A-Scene’ AI blends human and computer imagination into algorithmic art” »

Deep learning techniques have recently proved to be highly promising for detecting cybersecurity attacks and determining their nature. Concurrently, many cybercriminals have been devising new attacks aimed at interfering with the functioning of various deep learning tools, including those for image classification and natural language processing.

Perhaps the most common among these attacks are adversarial attacks, which are designed to “fool” deep learning algorithms using data that has been modified, prompting them to classify it incorrectly. This can lead to the malfunctioning of many applications, biometric systems, and other technologies that operate through deep learning algorithms.

Several past studies have shown the effectiveness of different adversarial attacks in prompting deep neural networks (DNNs) to make unreliable and false predictions. These attacks include the Carlini & Wagner attack, the Deepfool attack, the fast gradient sign method (FGSM) and the Elastic-Net attack (ENA).

A new computer algorithm developed by the University of Toronto’s Parham Aarabi can store and recall information strategically—just like our brains.

The associate professor in the Edward S. Rogers Sr. department of electrical and computer engineering, in the Faculty of Applied Science & Engineering, has also created an experimental tool that leverages the new algorithm to help people with memory loss.

“Most people think of AI as more robot than human,” says Aarabi, whose framework is explored in a paper being presented this week at the IEEE Engineering in Medicine and Biology Society Conference in Glasgow. “I think that needs to change.”

EPFL researchers have used swarms of drones to measure city traffic with unprecedented accuracy and precision. Algorithms are then used to identify sources of traffic jams and recommend solutions to alleviate traffic problems.

Given the wealth of modern technology available—roadside cameras, big-data algorithms, Bluetooth and RFID connections, and smartphones in every pocket—transportation engineers should be able to accurately measure and forecast city traffic. However, current tools advance towards the direction of showing the symptom but systematically fail to find the root cause, let alone fix it. Researchers at EPFL utilize a monitoring tool that overcomes many problems using drones.

“They provide excellent visibility, can cover large areas and are relatively affordable. What’s more, they offer greater precision than GPS technology and eliminate the behavioral biases that occur when people know they’re being watched. And we use drones in a way that protects people’s identities,” says Manos Barmpounakis, a post-doc researcher at EPFL’s Urban Transport Systems Laboratory (LUTS).

NVIDIA introduces QODA, a new platform for hybrid quantum-classical computing, enabling easy programming of integrated CPU, GPU, and QPU systems.

The past decade has seen quantum computing leap out of academic labs into the mainstream. Efforts to build better quantum computers proliferate at both startups and large companies. And while it is still unclear how far we are away from using quantum advantage on common problems, it is clear that now is the time to build the tools needed to deliver valuable quantum applications.

To start, we need to make progress in our understanding of quantum algorithms. Last year, NVIDIA announced cuQuantum, a software development kit (SDK) for accelerating simulations of quantum computing. Simulating quantum circuits using cuQuantum on GPUs enables algorithms research with performance and scale far beyond what can be achieved on quantum processing units (QPUs) today. This is paving the way for breakthroughs in understanding how to make the most of quantum computers.

Continue reading “Introducing QODA: The Platform for Hybrid Quantum-Classical Computing” »