Continual learning agents should incrementally learn a sequence of tasks while satisfying two main desiderata: accumulating on previous knowledge without forgetting and transferring previous relevant knowledge to help in future learning. Existing research largely focuses on alleviating the catastrophic forgetting problem. There, an agent is altered to prevent forgetting based solely on previous tasks. This hinders the balance between preventing forgetting and maximizing the forward transfer. In response to this, we investigate the stability-plasticity dilemma to determine which model components are eligible to be reused, added, fixed, or updated to achieve this balance. We address the class incremental learning scenario where the agent is prone to ambiguities between old and new classes. With our proposed Knowledge-Aware contiNual learner (KAN), we demonstrate that considering the semantic similarity between old and new classes helps in achieving this balance. We show that being aware of existing knowledge helps in: (1) increasing the forward transfer from similar knowledge, (2) reducing the required capacity by leveraging existing knowledge, (3) protecting dissimilar knowledge, and (4) increasing robustness to the class order in the sequence. We evaluated sequences of similar tasks, dissimilar tasks, and a mix of both constructed from the two commonly used benchmarks for class-incremental learning; CIFAR-10 and CIFAR-100. Continual learning (CL) aims to build intelligent agents based on deep neural networks that can learn a sequence of tasks, use previous knowledge in future learning, and accumulate on it without forgetting. The main challenge in this paradigm is the stability-plasticity dilemma (Mermillod et al., 2013). While fixing all weights, highest stability, hinders learning new tasks. Finding the right balance between stability and plasticity is challenging. This sharpens the community's focus on the forgetting problem.

The continual learning (CL) paradigm aims to enable neural networks to learn tasks continually in a sequential fashion. The fundamental challenge in this learning paradigm is catastrophic forgetting previously learned tasks when the model is optimized for a new task, especially when their data is not accessible. Current architectural-based methods aim at alleviating the catastrophic forgetting problem but at the expense of expanding the capacity of the model. Regularization-based methods maintain a fixed model capacity; however, previous studies showed the huge performance degradation of these methods when the task identity is not available during inference (e.g. class incremental learning scenario). In this work, we propose a novel architectural-based method referred as SpaceNet for class incremental learning scenario where we utilize the available fixed capacity of the model intelligently. SpaceNet trains sparse deep neural networks from scratch in an adaptive way that compresses the sparse connections of each task in a compact number of neurons. The adaptive training of the sparse connections results in sparse representations that reduce the interference between the tasks. Experimental results show the robustness of our proposed method against catastrophic forgetting old tasks and the efficiency of SpaceNet in utilizing the available capacity of the model, leaving space for more tasks to be learned. In particular, when SpaceNet is tested on the well-known benchmarks for CL: split MNIST, split Fashion-MNIST, and CIFAR-10/100, it outperforms regularization-based methods by a big performance gap. Moreover, it achieves better performance than architectural-based methods without model expansion and achieved comparable results with rehearsal-based methods, while offering a huge memory reduction.

AI is poised to deliver measurable value for a variety of public and private sector applications, and is just beginning to make inroads in the enterprise. This article offers best practices for picking an AI use case, points out barriers to AI success, and advises on how to find the best AI talent. Artificial intelligence (AI) technology dominates the headlines, but it's still not widely used. According to Gartner, between 2018 and 2019, the number of organizations deploying AI grew to just 14 percent. This may lead some enterprises to wonder, is the transition to AI necessary?

SAN FRANCISCO – Capella Space, a startup building a constellation of synthetic aperture radar (SAR) satellites, is joining SpaceNet, a nonprofit focused on geospatial applications for artificial intelligence. SpaceNet was established in 2016 by CosmiQ Works, the In-Q-Tel organization focused on commercial space, and DigitalGlobe, now part of Maxar Technologies. SpaceNet provides free high-resolution electro-optical satellite imagery, advertises challenges and awards cash prizes to encourage people to test and refine image analysis tools. SpaceNet has helped expand geospatial research and applications for electro-optical satellite imagery, said Ryan Lewis, SpaceNet general manager and senior vice president at In-Q-Tel, the venture capital arm of U.S. intelligence agencies. "Looking ahead, we want to see the same occur not just for imagery but also for SAR data," he told SpaceNews.

A CIA-linked firm has joined forces with Amazon in a bid to use "computer vision" to snoop on the Earth in unprecedented detail. CosmiQ Works, a firm closely associated with the US intelligence agency, is working with the online retail giant and the satellite mapping firm DigitalGlobe to train algorithms to work out what's happening on the surface of our planet. Satellites can already capture astonishingly detailed images from up in space, but the CIA-linked project wants to go one step further and use artificial intelligence to analyse these pictures. The partners hope to collect 60 million satellite images and store them in a database called SpaceNet which will be open and accessible by members of the public. Programmers will then design algorithms which can work out what's happening in the images or highlight the buildings, objects and natural features in the photos.

Private conversations and moments might not be so private anymore following CIA's latest plan. A CIA-linked firm is reportedly joining forces with Amazon. This is to spy on earth in an unprecedented detail. The firm known to be closely associated with the US Intelligence agency, CosmiQ Works, is working with Amazon and DigitalGlobe, a satellite mapping firm. The trio will be training artificial intelligence with an algorithm to find out what exactly is happening on the surface of the earth.

It sounds like something out of a sci-fi film - an AI that constantly monitors the Earth, looks for unusual activity. However, CosmiQ Works, a division of the CIA's venture arm, has revealed SpaceNet, a project with Amazon, satellite mapping firm DigitalGlobe and chip firm Nvidia to train algorithms to work out what's happening on our planet. The project will create a giant online database of hi-res images that AIs will be able to use to teach themselves - and started with images of Rio during the Olympics. SpaceNet will launch with an initial contribution of DigitalGlobe multi-spectral satellite imagery and 200,000 curated building footprints across the city of Rio de Janeiro, Brazil. 'Each minute something is happening in the world,' said said Tony Frazier, Senior Vice President at DigitalGlobe.