Last week, Ben Goertzel and his company, Aidyia, turned on a hedge fund that makes all stock trades using artificial intelligence--no human intervention required. "If we all die," says Goertzel, a longtime AI guru and the company's chief scientist, "it would keep trading." Goertzel and other humans built the system, of course, and they'll continue to modify it as needed. But their creation identifies and executes trades entirely on its own, drawing on multiple forms of AI, including one inspired by genetic evolution and another based on probabilistic logic. Each day, after analyzing everything from market prices and volumes to macroeconomic data and corporate accounting documents, these AI engines make their own market predictions and then "vote" on the best course of action. If we all die, it would keep trading.

Researchers at Shenzhen University have recently devised a method to detect images generated by deep neural networks. Their study, pre-published on arXiv, identified a set of features to capture color image statistics that can detect images generated using current artificial intelligence tools. "Our research was inspired by the rapid development of image generative models and the spread of generated fake images," Bin Li, one of the researchers who carried out the study, told Tech Xplore. "With the rise of advanced image generative models, such as generative adversarial networks (GAN) and variational autoencoders, images generated by deep networks become more and more photorealistic, and it is no longer easy to identify them with human eyes, which entails serious security risks." Recently, several researchers and global media platforms have expressed their concern with the risks posed by artificial neural networks trained to generate images. For instance, deep learning algorithms such as generative adversarial networks (GAN) and variational autoencoders could be used to generate realistic images and videos for fake news or could facilitate online frauds and the counterfeit of personal information on social media.

The security industry has experienced its share of rapid changes that have been well documented in recent years. The question is how soon will the next wave of potential disruption really take hold? In this case, consider buzzwords that are becoming more common to hear at industry trade shows and in market research reports but perhaps remain foreign to many in terms of the everyday lexicon: artificial intelligence (AI), machine learning and deep learning. There is much work to be done as these still-futuristic-sounding technologies make their way into the security industry. Since many sensor manufacturers do not have the expertise in these areas, go-to market strategies have been mostly by partnering with video analytics solution providers taking their first steps into AI.

CAMLIS is a venue for discussing applied research on machine learning, deep learning and data science in information security. This is an intimate, two day, single tracked conference with 2 keynotes and many excellent talks. Find more information here: https://www.camlis.org/

Even though we like the idea that we never make mistakes, every software can contain bugs. Assuming that we may use Machine Learning models to make decisions in the real world, a bug in our code can be very dangerous. Relying only on the prediction accuracy might not be a good idea, because if we get a good accuracy score we might not even consider that there is a bug in our data pipeline. Most Machine Learning algorithms are black boxes, but LIME has a bold value proposition: explain the results of any predictive model. The tool can explain models trained with text, categorical, or continuous data.

The system uses deep learning and image recognition to monitor video footage in real time and identify any frames in which dangerous behavior is occurring. Right now the new AI system is only being trialed at two gas stations, one in Thailand and one in Singapore. But if all goes well, it will roll out to other gas stations in the future. And if you need to see how dangerous smoking at the pump really is, check out the video below.

Editor's note: You can read the previous post in this series, Power Laws in Deep Learning, here. In a previous post, we saw that the Fully Connected (FC) layers of the most common pre-trained Deep Learning display power law behavior. Remarkably, the FC matrices all lie within the Universality Class of Fat Tailed Random Matrices! We define a random matrix by defining a matrix of size, and drawing the matrix elements from a random distribution. In either case, Random Matrix Theory tells us what the asymptotic form of ESD should look like. But first, let's see what model works best.

R Data science includes data analysis. It is an important component of the skill set required for many jobs in this area. Generally, there are 3 types of learning algorithms: a. Supervised Machine Learning Algorithms To make predictions we use this machine learning algorithm. That was assigned to data points. Unsupervised Machine Learning Algorithms No labels are associated with data points.

Random Matrix Theory (RMT) and Randomized Numerical Linear Algebra (RandNLA) are applied to analyze the weight matrices of Deep Neural Networks (DNNs), including both production quality, pre-trained models and smaller models trained from scratch. Empirical and theoretical results clearly indicate that the DNN training process itself implicitly implements a form of self-regularization, implicitly sculpting a more regularized energy or penalty landscape. Building on relatively recent results in RMT, most notably its extension to Universality classes of Heavy-Tailed matrices, and applying them to these empirical results, we develop a theory to identify 5 1 Phases of Training, corresponding to increasing amounts of implicit self-regularization. For smaller and/or older DNNs, this implicit self-regularization is like traditional Tikhonov regularization, in that there appears to be a size scale'' separating signal from noise. For state-of-the-art DNNs, however, we identify a novel form of heavy-tailed self-regularization, similar to the self-organization seen in the statistical physics of disordered systems.

There's a cognitive quirk humans have that seems deceptively elementary. For example: every morning, you see a man in his 30s walking a boisterous collie. Then one day, a white-haired lady with striking resemblance comes down the street with the same dog. Subconsciously we immediately make a series of deductions: the man and woman might be from the same household. The lady may be the man's mother, or some other close relative.