Singularity Pulse  

AI to Provide On-Demand Portfolio Evaluation and Fundamental Research

Analytixinsight, a fintech company, has leveraged developments in the field of AI to create a portal, “”, which will perform company analysis, fundamental research and portfolio evaluation on 50,000 global equities and North American ETFs. It will work on a “freemium” model, meaning that it will provide free basic financial information to each of its clients while subscribers will receive in-depth analysis and predictive analytics results. Company has also partnered with Italy's largest retail bank, to create “MarketWall”, a real time stock trading app which will have CapitalCube's company analysis data automatically integrated to it.



AI Offers Threat Detection Solutions Against Malicious Data Samples

Fortinet, a global leader in integrated and automated cybersecurity solutions announced the launch of its intelligence services platform named “FortiGuard AI”. It delivers automated threat analysis and detection, thus keeping company’s customer security solutions updated to protect against the latest threats. FortiGuard uses machine learning and continuous training to analyze threats. It then classifies threats into different categories with a high accuracy and speed. The team of over 200 researchers and engineers in 30+ countries feed the machine with data from a global network of more than three million security sensors. Using supervised learning techniques, data is analyzed by over five billion processing nodes and identifies the malicious and clean features of each sample. It then determines if a sample poses any threat and accordingly flag those samples.



AI Techniques Reconstruct Mysteries of Quantum Systems

Physicists have now demonstrated that machine learning can reconstruct a quantum system using fewer experimental measurements. This will allow scientists to probe systems of particles exponentially faster, as compared with traditional brute-force methods. In quantum physics, systems of particles exist in lots of different configurations. But when measured, it collapses into just one configuration, meaning one can never observe the entire complexity of a system in a single experiment. Conventional methods aren't feasible for analyzing each possible configuration. Physicists fed experimental measurements to a software tool based on artificial neural networks. It learns over time and then mimics the system's behavior. Over time, it can accurately reconstruct the complete quantum system for further analysis.



AI Algorithms Are Generating Videos Out of Thin Air

Just like images, scientists at Duke and Princeton are now working to get AI to generate videos on its own. The algorithm simply reads a small phrase and creates a video out of it. In video AI attempts to predict what actions come next in a video. The algorithm studied multiple videos, mainly focused on sports, from Google's Kinetics Human Action Video Dataset. It then learned to identify each motion. Researchers used a two-step process to create the video; firstly "generate the gist”, where the gist is an image that gives the background color and layout. And secondly, "discriminator", where it judges the gist's work.

Information & Technology  

  Chemical Engineering

AI Methods Used for Developing Precision Cancer Medicine

Scientists have identified that with the use of artificial intelligence methods like network modeling, patient's own molecular data can be used to identify the best therapy option. Network modeling is used in integrating genome-scale patient data into detailed interaction networks. This can be analyzed by algorithms to identify combinations of drugs and inhibitors that are likely to be effective. The research is being done at Computational Biomodeling (Combio) Laboratory of Åbo Akademi and Turku Centre for Computer Science (TUCS), Finland. Currently, it focuses on individual patients to adapt their therapeutical strategies.



Machine Learning to Teach How to Draw

Sketch AR is a new app powered by AI algorithms which does the task of drawing for its user. The way the app works is simple. User must draw a few plus-signs on a piece of paper or wall (subject they want to draw) and then click a picture of it with their smartphones. For larger pieces, one can cycle through point-by-point thus associating free space with the next layer of drawing. The program is not based on markers, but the accuracy is better if some pre-installed markers held the virtual object in place. Method includes detection and tracking of key points to guide the user. It used machine learning and neural networks to analyze and train data from each user’s drawing.

Information & Technology  


AI to Check Society’s Gender Equality Problems

Currently women are being under-represented in many spheres of economic life. Technology, in its current form can do more harm than good as the biased data will only make algorithm derive predictions that are even more biased. However, machine learning algorithms can be tested for biases via stress testing the system. Anupam Datt, computer scientist, designed a programme to test whether AI showed bias in hiring new employees. In a candidate selection algorithm for a company, Datta’s program randomly changed the gender and weight data fields. This was then put to test and number of female employees selected for the position were to be recalculated. A no change in the number of women selected for interviews means that the algorithm is unbiased.



AI in Policing

New Orleans Police Department has partnered with national security company Palantir to predict crimes. The company provided software program that traced people’s ties to other gang members, outlined criminal histories, analyzed social media. It then predicts the likelihood that individuals would commit violence or become victim to one. The program began in 2012 as a partnership between New Orleans Police and Palantir Technologies as “pro bono” and philanthropic in nature. Since then it has been extended three times. Civil authorities are raising their concern over the accuracy as they believe it is still in prototype mode and could potentially flag out innocent people.



Machine Learning to Improve End-Of-Life Care for Critical Patients

KenSci develops machine learning risk prediction platform for healthcare industry. Recently they came out with research on predicting end-of-life mortality and improving care of patient till then. In US alone, there is over $200 billion being spent on patient’s end of life care. The company developed a ML algorithm by training it over data from two sets with broad demographics. It used Microsoft Azure cloud to keep its system cloud-based and therefore connects to new data sources as they become available. The model is taught “assistive intelligence”. At each step, human input is a critical factor to proceed to the next level of program. KenSci machine learning (ML) platform delivers models that can be interpreted for correctness and validation by physicians and clinicians. Besides, the algorithm also accepts corrections from humans.



Google To Make Machine Learning Education Available for All

Google recently introduced "Learn with Google AI", a set of educational resources for people. The course features videos from ML experts at Google, interactive visualisations illustrating ML concepts, coding exercises using cutting-edge TensorFlow APIs and a focus that teaches how practitioners implement ML in the real world. It also features a new, free course called Machine Learning Crash Course (MLCC) which provides exercises, interactive visualizations, and instructional videos that anyone can use to learn and practice ML concepts. So far, more than 18,000 Googlers have enrolled in MLCC.



Intel’s New AI Chip to Broadcast Real Time Drone Footage of Racing Cars

Intel has unveiled a new chip with deep learning capabilities called “Nervana Neural Network Processor (NNP)” and is deploying them in Ferrari cars to bring real-time, dynamic statistics and information to both the driver and spectators in minimum time possible. The new chip will also be able to analyze live video taken by drones to compare the exit angles taken by a driver during laps. NNP uses human brain-like features to collect stats on cars’ engine performance, called telemetric data. It stores similar data in multiple locations to make it easier and quicker to access. It downgrades the level of its precision minutely in favor of quicker and more effective processing.



AI’s Deep Dream and Neural Style Transfer Software Created Rock Music on Its Own

The recent scientific developments in artificial-intelligent image processing have been put to musical work by Hardcore Anal Hydrogen’s in developing their track titled as “Jean-Pierre”. The band used software like Google’s Deep Dream, which is a computer image processing, and Neural Style Transfer to create their track. In Deep Dream, AI is trained to recognize an image, using thousands of similar images. Once ready, the software processes a completely different image using the knowledge acquired during training. In Neural Style Transfer, AI is trained on the art style and colors of an image. Software tries to fill in the newly created image with this software.



Contact us:

5A, B-Wing, Trade Star Building, J. B. Nagar, Andheri-Kurla Road, Andheri (East), Mumbai - 400 059, Maharashtra, India
E:   T: +91 22300 15200