Octopus Energy to spin off $8.65bn tech arm Kraken Octopus Energy is set to spin off its Kraken Technologies arm as a standalone company after a deal to sell a stake in the platform valued it at $8.65bn (£6.4bn). The energy giant, Britain's biggest gas and electricity supplier, has sold a $1bn stake in the AI-based division to a group of investors led by New York-based D1 Capital Partners. The move paves the way for Kraken to be demerged from Octopus, and for a potential stock market flotation for the business in the future. Octopus founder and chief executive Greg Jackson told the BBC there was every chance Kraken would list its shares in the medium term, with the location of the flotation between London and the US. Kraken uses AI to automate customer service and billing for energy companies and can manage when customers use energy, rewarding them for reducing consumption at peak times. It was initially built for use by Octopus but has since picked up a raft of other utilities clients, including EDF, E.On Next, TalkTalk and National Grid US.

Large Transformer networks are increasingly used in settings where low inference latency is necessary to enable new applications and improve the end-user experience.However, autoregressive inference is resource intensive and requires parallelism for efficiency.Parallelism introduces collective communication that is both expensive and represents a phase when hardware resources are underutilized.Towards mitigating this, Kraken is an evolution of the standard Transformer architecture that is designed to complement existing tensor parallelism schemes for efficient inference on multi-device systems.By introducing a fixed degree of intra-layer model parallelism, the architecture allows collective operations to be overlapped with compute, decreasing latency and increasing hardware utilization.When trained on OpenWebText, Kraken models reach a similar perplexity as standard Transformers while also preserving their language modeling capabilities as evaluated on the SuperGLUE benchmark.Importantly, when tested on multi-GPU systems using TensorRT-LLM engines, Kraken speeds up Time To First Token by a mean of 35.6% across a range of model sizes, context lengths, and degrees of tensor parallelism.

Large Transformer networks are increasingly used in settings where low inference latency is necessary to enable new applications and improve the end-user experience.However, autoregressive inference is resource intensive and requires parallelism for efficiency.Parallelism introduces collective communication that is both expensive and represents a phase when hardware resources are underutilized.Towards mitigating this, Kraken is an evolution of the standard Transformer architecture that is designed to complement existing tensor parallelism schemes for efficient inference on multi-device systems.By introducing a fixed degree of intra-layer model parallelism, the architecture allows collective operations to be overlapped with compute, decreasing latency and increasing hardware utilization.When trained on OpenWebText, Kraken models reach a similar perplexity as standard Transformers while also preserving their language modeling capabilities as evaluated on the SuperGLUE benchmark.Importantly, when tested on multi-GPU systems using TensorRT-LLM engines, Kraken speeds up Time To First Token by a mean of 35.6% across a range of model sizes, context lengths, and degrees of tensor parallelism.

Accurate and reliable motion prediction is essential for safe urban autonomy. The most prominent motion prediction approaches are based on modeling the distribution of possible future trajectories of each actor in autonomous system's vicinity. These "independent" marginal predictions might be accurate enough to properly describe casual driving situations where the prediction target is not likely to interact with other actors. They are, however, inadequate for modeling interactive situations where the actors' future trajectories are likely to intersect. To mitigate this issue we propose Kraken -- a real-time trajectory prediction model capable of approximating pairwise interactions between the actors as well as producing accurate marginal predictions. Kraken relies on a simple Greedy Mode Processing technique allowing it to convert a factorized prediction for a pair of agents into a physically-plausible joint prediction. It also utilizes the Mode Transformer module to increase the diversity of predicted trajectories and make the joint prediction more informative. We evaluate Kraken on Waymo Motion Prediction challenge where it held the first place in the Interaction leaderboard and the second place in the Motion leaderboard in October 2021.

The interest in dynamic vision sensor (DVS)-powered unmanned aerial vehicles (UAV) is raising, especially due to the microsecond-level reaction time of the bio-inspired event sensor, which increases robustness and reduces latency of the perception tasks compared to a RGB camera. This work presents ColibriUAV, a UAV platform with both frame-based and event-based cameras interfaces for efficient perception and near-sensor processing. The proposed platform is designed around Kraken, a novel low-power RISC-V System on Chip with two hardware accelerators targeting spiking neural networks and deep ternary neural networks.Kraken is capable of efficiently processing both event data from a DVS camera and frame data from an RGB camera. A key feature of Kraken is its integrated, dedicated interface with a DVS camera. This paper benchmarks the end-to-end latency and power efficiency of the neuromorphic and event-based UAV subsystem, demonstrating state-of-the-art event data with a throughput of 7200 frames of events per second and a power consumption of 10.7 \si{\milli\watt}, which is over 6.6 times faster and a hundred times less power-consuming than the widely-used data reading approach through the USB interface. The overall sensing and processing power consumption is below 50 mW, achieving latency in the milliseconds range, making the platform suitable for low-latency autonomous nano-drones as well.

Layout Analysis (the identification of zones and their classification) and line segmentation are the first steps in Optical Character Recognition and similar tasks. The ability of identifying the main body of text from marginal text or running titles makes the difference between extracting the full text of a digitized book and noisy outputs. We show that most segmenters focus on pixel classification and that polygonization of this output has not been used as a target for the latest competitions on historical documents (ICDAR 2017 and onwards), despite being the focus in the early 2010s. We suggest that transitioning the task from pixel classification-based polygonization to object detection using isothetic rectangles might improve results in terms of speed and accuracy. We compare the output of Kraken and YOLOv5 in terms of segmentation and show that the latter severely outperforms the first on small datasets (1110 samples and below). We release two datasets for training and evaluation on historical documents as well as a new package, YALTAi, which injects YOLOv5 in the segmentation pipeline of Kraken 4.1. I INTRODUCTION In recent years, automatic text extraction has become an important activity in digital philology and, in general, in corpus creation for historical documents.

Kindly give this article a like or a comment, so I know that you are still reading. I'm a big believer in A.I.'s ability to enable human civilization to become a multi-planetary species. I think artificial intelligence will be critical in enabling us to make this jump in the brief window afforded to us by time and history since the risks of human extinction will become greater in the decades and centuries ahead. I'm always searching and on the hunt for big stories in how A.I. is shaping our understanding of the world and in terms of business innovation. Sometimes however you have to look up.

Alán Aspuru-Guzik's research group at the University of Toronto has launched an open-access tool that promises to accelerate the discovery of new chemical reactions that underpin the development of everything from smartphones to life-saving drugs. The free tool, called Kraken, is a library of virtual, machine-learning calculated organic compounds – roughly 300,000 of them, with 190 descriptors each. It was created through a collaboration between Aspuru-Guzik's Matter Lab, the Sigman Research Group at the University of Utah, Technische Universität Berlin, Karlsruhe Institute of Technology, Vector Institute for Artificial Intelligence, the Center for Computer Assisted Synthesis at the University of Notre Dame, IBM Research and AstraZeneca "The world has no time for science as usual," says Aspuru-Guzik, a professor in U of T's departments of chemistry and computer science in the Faculty of Arts & Science. "Neither for science done in a silo. "This is a collaborative effort to accelerate catalysis science that involves a very exciting team from academia and industry." When developing a transition-metal catalyzed chemical reaction, a chemist must find a suitable combination of metal and ligand. Despite the innovations in computer-optimized ligand design led by the Sigman group, ligands would typically be identified by trial and error in the lab. With Kraken, however, chemists will eventually have a vast data-rich collection at their fingertips, reducing the number of trials necessary to achieve optimal results. "It takes a long time, a lot of money, and a whole lot of human resources to discover, develop and understand new catalysts and chemical reactions." "These are some of the tools that allow molecular scientists to precisely develop materials and drugs, from the plastics in your smartphone to the probes that allowed for humanity to achieve the COVID-19 vaccines at an unforeseen pace.

Artificial intelligence (AI) has helped in creating the first complete family tree of Earth's home galaxy – the Milky Way. An international team of researchers, led by astrophysicists Diederik Kruijssen of the University of Heidelberg and Joel Pfeffer of Liverpool John Moores University, published their work in Monthly Notices of the Royal Astronomical Society. The researchers used AI to analyse large groups of stars with as many as million stars, orbiting the Milky Way. "The Milky Way hosts over 150 such clusters, many of which formed in the smaller galaxies that merged to form the galaxy that we live in today," a Royal Astronomical Society (RAS) release noted. With the help of the latest models and observations, the researchers managed to use the clusters as "fossils" to generate the history of galaxies, it added.

The Milky Way has had a long and eventful life. Throughout its history, our galaxy has collided and merged with multiple other galaxies, events that are hard to disentangle and make sense of. With the aid of Artificial Intelligence, a team of astronomers took on this painstaking task, piecing together the most complex history of our galaxy -- and the main attraction is something called The Kraken. Just like geologists look for fossils to see how ancient life might have looked like, astronomers also look for fossils of their own -- but instead of trilobites or dinosaurs, astronomers are preoccupied with very old cosmic structures called globular clusters. Globular clusters are spherical-shaped, densely-packed collections of ancient stars.