Singularity Pulse  
     
  Finance

Morgan Stanley To Use Robots to Scour Its Clients’ Social Media Profiles

During major market sell-offs, financial advisers must make sure their clients remain calm and don't sell their stocks in a panic. Morgan Stanley is using AI to write emails to its wealthy clients when the markets go haywire. Robots are being used to support human advisers with these types of tasks. It has generated customized emails for its 16,000 financial advisers to send to clients. Gleaning information from a client's social media and other public sources, these emails are customized to each user & includes their personal details.

 
  Image  
 

Samsung SDS Unveils Blockchain-Based Finance Platform

Samsung SDS has unveiled a new blockchain-based finance platform called Nexfinance. It will offer services like digital identity, finance concierge, AI virtual assistant, and automatic insurance payment. The digital identity service is blockchain-based and will help protect customer information, while finance concierge uses AI and big data analytics to help clients manage their assets. The platform will also let clients to apply third-party solutions.

General  
  Image  
  Finance

An AI System to Defeat Facial Recognition Systems

Researchers in Toronto developed a shield against encroaching AI-powered surveillance systems. The researchers have created an algorithm that can disrupt facial recognition systems dynamically, on the fly, by applying light transformations to images.Their algorithm is trained on dataset of 600 faces. It has a real-time filter that can be applied to any picture. It targets highly specific, individual pixels in an image. Such high detailed oriented approach makes it imperceptible to the human eye. They used adversarial training comprising of two neural networks. It claims to reduce the proportion of detected faces in facial recognition systems to 0.5%.

 
  Image  
 

AI Solution Capable of Detecting 15 Types Of Chest X-Ray Abnormalities

Qure.ai, a Mumbai based AI health care start-up has developed a chest x-ray product that can identify 15 chest x-ray abnormalities. It is based on neural networks which takes milliseconds to process them. The team sourced anonymized x-ray scans of tuberculosis patients from about 15 institutions to gather images. In its preliminary study, qXR was 90% accurate in correctly diagnosing tuberculosis. The company will receive funding of up to $30 million from Fractal Analytics in the coming few years.

Digital Imaging  
  Image  
  Healthcare

Robots Getting Better Than Doctors at Diagnosing Cancer

An international study involving machines that were trained to detect signs of skin cancer were tested against 58 dermatologists. The computer was trained on 100,000 images of malignant melanomas and benign moles marked with diagnosis. It was then tested against doctors to diagnose 100 of the most difficult lesions. Machine was correct 95% of times as compared to 87% score of doctors. Misdiagnosis (false positive) of benign moles was also lower in cases of machines. Convolutional Neural Network (CNN) was used as AI technique to train machines.

 
  Image  
 

AI Software to Assist Design of Material for Solar Cells

Researchers at Osaka University have developed a program to automate matching of solar materials. Traditionally, chemists experiment with combinations of organic and polymer layers to reach highest power conversion efficiencies (PCEs). PCE is the ability to convert light into electricity. Team gathered data on 1200 Organic photovoltaics (OPVs) from 500+ studies. Model was built on Random Forest ML technique. The model was built combining band gap, molecular weight and chemical structure of OPVs to predict efficiency of new devices.

Pharma  
  Image  
  Information Technology

Robot Predicts Consumer Satisfaction Based on Beer Head

Researchers at the University of Melbourne have developed technology to assess consumer perceptions about beers based on the foam it produces by gaseous bubbles. It used RoboBEER prototype to decant beers from bottle to glasses. Videos were shown to participants. Neural Network and Eye-tracking technology were used to measure pupil dilation, heart rate, and emotional expression. The algorithm compared that data to person’s conscious ratings to have a relative score. The system was able to predict whether someone liked a beer’s foam height with 81% accuracy.

 
  Image  
 

Instagram Reveals How Its ML Algorithm Works

Instagram’s relevancy-sorting algorithm has been criticized but the company claims that users see 90% of their friends’ posts. Before the algorithm change, its users missed 70% of all posts and 50% of their friend’s posts. The company also said that it has seen increase in the time users spend on the app. To customize the feed, it uses ML on past behavior. Two people will receive different feeds of the same page. It uses three main factors — how much you’ll care about a post (based on past behavior), how recently the post was shared, and how close you are with the person who shared it

Healthcare  
  Image  
  General

Deep Learning Algorithm to Increase Depth Of 3D Hologram Images

UCLA researchers applied deep learning technology to extend the imaging depth of a hologram. In holography, “autofocusing” and “phase recovery” are required for image reconstruction. In general, these processes are time consuming. The newly developed technology called, “HIDEF”, can simultaneously perform the two tasks. This increases the reconstruction speed in the hologram along with increasing the image depth. HIDEF is based on convolutional neural network. The approach was validated successfully by reconstructing holograms of aerosols and human tissue samples.

 
  Image  
 

Researchers are Teaching AI to do Daily Household Chores

Researchers at MIT & University of Toronto are teaching AI to do household chores using a Sims inspired virtual simulator. Researchers collected common verbal descriptions of household chores, then translated into robot friendly code. They showcased the AI house assistants in a 3D VirtualHome stimulator. The robot was successful in doing tasks like making coffee, turning the toaster on etc. This also helped researchers create a database in natural language format, which could help Amazon Alexa to carry out more complicated tasks.

General  
  Image  
 
 
 
     
  Banner  
 
Contact us:

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