Lifeboat Foundation

Marking a new era of “diagnosis by software,” the US Food and Drug Administration on Wednesday gave permission to a company called IDx to market an AI-powered diagnostic device for ophthalmology.

What it does: The software is designed to detect greater than a mild level of diabetic retinopathy, which causes vision loss and affects 30 million people in the US. It occurs when high blood sugar damages blood vessels in the retina.

How it works: The program uses an AI algorithm to analyze images of the adult eye taken with a special retinal camera. A doctor uploads the images to a cloud server, and the software then delivers a positive or negative result.

Continue reading “FDA approves AI-powered diagnostic that doesn’t need a doctor’s help” »

Machines don’t actually have bias. AI doesn’t ‘want’ something to be true or false for reasons that can’t be explained through logic. Unfortunately human bias exists in machine learning from the creation of an algorithm to the interpretation of data – and until now hardly anyone has tried to solve this huge problem.

A team of scientists from Czech Republic and Germany recently conducted research to determine the effect human cognitive bias has on interpreting the output used to create machine learning rules.

The team’s white paper explains how 20 different cognitive biases could potentially alter the development of machine learning rules and proposes methods for “debiasing” them.

Continue reading “Human bias is a huge problem for AI. Here’s how we’re going to fix it” »

“If you went to bed last night as an industrial company, you’re going to wake up this morning as a software and analytics company.” Jeff Immelt, former CEO of General Electric

The second wave of digitization is set to disrupt all spheres of economic life. As venture capital investor Marc Andreesen pointed out, “software is eating the world.” Yet, despite the unprecedented scope and momentum of digitization, many decision makers remain unsure how to cope, and turn to scholars for guidance on how to approach disruption.

The first thing they should know is that not all technological change is “disruptive.” It’s important to distinguish between different types of innovation, and the responses they require by firms. In a recent publication in the Journal of Product Innovation, we undertook a systematic review of 40 years (1975 to 2016) of innovation research. Using a natural language processing approach, we analyzed and organized 1,078 articles published on the topics of disruptive, architectural, breakthrough, competence-destroying, discontinuous, and radical innovation. We used a topic-modeling algorithm that attempts to determine the topics in a set of text documents. We quantitatively compared different models, which led us to select the model that best described the underlying text data. This model clustered text into 84 distinct topics. It performs best at explaining the variability of the data in assigning words to topics and topics to documents, minimizing noise in the data.

Continue reading “What 40 Years of Research Reveals About the Difference Between Disruptive and Radical Innovation” »

Machine learning and Artificial Intelligence developments are happening at a break neck speed! At such pace, you need to understand the developments at multiple levels – you obviously need to understand the underlying tools and techniques, but you also need to develop an intuitive understanding of what is happening.

By end of this article, you will develop an intuitive understanding of RNNs, specially LSTM & GRU.

Ready?

Continue reading “Replicating human memory structures in neural networks to create precise NLU algorithms” »

An algorithm tailored to individual brain activity shows it can boost memory with electrical zaps.

Protein synthesis is a critical part of how our cells operate and keep us alive and when it goes wrong it drives the aging process. We take a look at how it works and what happens when things break down.

Suppose that your full-time job is to proofread machine-translated texts. The translation algorithm commits mistakes at a constant rate all day long; from this point of view, the quality of the translation stays the same. However, as a poor human proofreader, your ability to focus on this task will likely decline throughout the day; therefore, the number of missed errors, and therefore the number of translations that go out with mistakes, will likely go up with time, even though the machine doesn’t make any more errors at dusk than it did at dawn.

To an extent, this is pretty much what is going on with protein synthesis in your body.

Continue reading “Protein Synthesis in Aging” »

Researchers just overturned a 70-year-old fundamental understanding of how our brains learn – paving the way for faster, more advanced AI applications and a different approach to medical treatments for brain disorders. [This article first appeared on LongevityFacts. Author: Brady Hartman. ]

Researchers just overturned the way scientists thought our brains learn – a view that up until now has been widely accepted for almost 70 years.

This discovery-based upon new experimental evidence – paves the way for more modern artificial intelligence (AI) applications such as machine learning and deep learning algorithms that imitate our brain functions at a much faster speed with advanced features. Moreover, the research may change how doctors view disorders of the brain, such as Alzheimer’s and may alter treatments for other forms of dementia.

Continue reading “Our Brain Learns Completely Differently Than Experts Assumed For Nearly A Century” »

Engineering and construction is behind the curve in implementing artificial intelligence solutions. Based on extensive research, we survey applications and algorithms to help bridge the technology gap.

The engineering and construction (E&C) sector is worth more than $10 trillion a year. And while its customers are increasingly sophisticated, it remains severely underdigitized. To lay out the landscape of technology, we conducted a comprehensive study of current and potential use cases in every stage of E&C, from design to preconstruction to construction to operations and asset management. Our research revealed a growing focus on technological solutions that incorporate artificial intelligence (AI)-powered algorithms. These emerging technologies focus on helping players overcome some of the E&C industry’s greatest challenges, including cost and schedule overruns and safety concerns.

New developments require new materials. Until recently, these have been developed mostly by tedious experiments in the laboratory. Researchers at the Fraunhofer Institute for Algorithms and Scientific Computing SCAI in Sankt Augustin are now significantly shortening this time-consuming and cost-intensive process with their “Virtual Material Design” approach and the specially developed Tremolo-X software. By combining multi-scale models, data analysis and machine learning, it is possible to develop improved materials much more quickly. At the Hanover Trade Fair from April 23 to 27, 2018, Fraunhofer will be demonstrating how the virtual material design of the future looks.

In almost every industry, new materials are needed for new developments. Let’s take the automotive industry: while an automobile used to consist of just a handful of materials, modern cars are assembled from thousands of different materials – and demand is increasing. Whether it’s making a car lighter, getting better fuel economy or developing electric motor batteries, every new development requires finding or developing the material that has exactly the right properties. The search for the right material has often been like a guessing game, though. The candidates have usually been selected from huge material databases and then tested. Although these databases provide insight into specific performance characteristics, they usually do not go far enough into depth to allow meaningful judgments about whether a material has exactly the desired properties. To find that out, numerous laboratory tests have to be performed.