#artificialintelligence


MIT is turning AI into a pizza chef

#artificialintelligence

The result is a system that isn't perfect (it's better at ordering synethetic pizza images than real ones), but it's still reasonably accurate. The scientists found that PizzaGAN could determine the right order 88 percent of the time, albeit using pizzas with just two toppings. It would likely have a harder time with a fully decked-out pie. You won't see pizza bots in the near future, since it'd take a while to teach robots to prepare and cook pizza all on their own. The lessons learned here could be valuable going forward, though, and not just in the cooking realm.


Self-driving car startup Argo AI is giving researchers free access to its HD maps – TechCrunch

#artificialintelligence

Argo AI is releasing curated data along with high-definition maps to researchers for free, the latest company in the autonomous vehicle industry to open-source some of the information it has captured while developing and testing self-driving cars. The aim, the Ford Motor-backed company says, is to give academic researchers the ability to study the impact that HD maps have on perception and forecasting, such as identifying and tracking objects on the road, and predicting where those objects will move seconds into the future. In short, Argo sees this as a way to encourage more research and hopefully breakthroughs in autonomous vehicle technology. Argo has branded this collection of data and maps Argoverse, which is being released for free. Argo isn't releasing everything it has.


Python Could Rule the Machine Learning/A.I. World

#artificialintelligence

What do developers actually use Python for? According to a developer survey by JetBrains (which also introduced Kotlin, the up-and-coming language for Android development), some 49 percent say they use Python for data analytics, ahead of web development (46 percent), machine learning (42 percent), and system administration (37 percent). Significant numbers of developers also use the language for software testing (25 percent), software prototyping (22 percent), and "educational purposes" (20 percent). Far fewer chose it for graphics, embedded development, or games/mobile development. This data just reinforces the general idea that Python is swallowing the data-analytics space whole.


Verily creates machine learning tool to aid diagnostic development

#artificialintelligence

Verily has published details of a machine learning approach that may support development of new diagnostic tools. The approach, DeepMass, is designed to improve characterization of disease-relevant protein profiles by tackling a limitation on the use of mass spectrometry. In early tests, DeepMass was more accurate than an existing prediction model and expanded the coverage of known biomarkers when applied to clinical data. Verily, the Alphabet unit formerly known as Google Life Sciences, measures protein profiles to find new disease biomarkers in a range of its programs, including The Project Baseline Health Study that has attracted the support of Pfizer and other drugmakers. These searches for biomarkers use a form of protein mass spectrometry designed to increase the accuracy of protein identification and quantification.


From one brain scan, more information for medical artificial intelligence: System helps machine-learning models glean training information for diagnosing and treating brain conditions

#artificialintelligence

An active new area in medicine involves training deep-learning models to detect structural patterns in brain scans associated with neurological diseases and disorders, such as Alzheimer's disease and multiple sclerosis. But collecting the training data is laborious: All anatomical structures in each scan must be separately outlined or hand-labeled by neurological experts. And, in some cases, such as for rare brain conditions in children, only a few scans may be available in the first place. In a paper presented at the recent Conference on Computer Vision and Pattern Recognition, the MIT researchers describe a system that uses a single labeled scan, along with unlabeled scans, to automatically synthesize a massive dataset of distinct training examples. The dataset can be used to better train machine-learning models to find anatomical structures in new scans -- the more training data, the better those predictions.


How to initialize a Neural Network

#artificialintelligence

Training a neural net is far from being a straightforward task, as the slightest mistake leads to non-optimal results without any warning. Training depends on many factors and parameters and thus require a thoughtful approach. It is known that the beginning of training (i.e., the first few iterations) is very important. When done improperly, you get bad results -- sometimes, the network won't even learn anything at all! For this reason, the way you initialize the weights of the neural network is one of the key factors to good training.


San Francisco DA Office Attempts Novel Feat: Using AI to Combat Bias Legaltech News

#artificialintelligence

Algorithms are coming to the San Francisco District Attorney's Office. In a bid to combat implicit bias, artificial intelligence-backed software will redact race information from police incident reports before prosecutors make their initial charging decision.


This Colorado hospital is using Qventus' AI to improve operations - MedCity News

#artificialintelligence

Wheat Ridge, Colorado-based Lutheran Medical Center, which is part of Broomfield, Colorado-based SCL Health, wanted to improve its operations. "We determined a few years ago that for a hospital like ours that has a very challenging payer mix, … running an extremely cost-efficient operation was necessary for stability," said Lutheran Medical Center president and CEO Grant Wicklund in a phone interview. "One of the ways we identified we could become even more cost-efficient was to be absolutely world-class at having the appropriate length of stay." Noomi Hirsch, the medical center's vice president of operations, took the lead on the effort. In a phone interview, she explained that the organization was able to hit low-hanging fruit areas, but eventually started looking at options in the technology world to tackle the problem.


What's new in Core ML

#artificialintelligence

To say this year has been massive for API and framework updates would be underselling WWDC 2019. Core ML has been no exception. So far with Core ML 2 we saw some amazing updates and that made on device training amazingly simple. However, there was still a lot to be desired and if you wanted to implement newer models like YOLO you needed to drop down to Metal and do a lot of leg work to get a model up and running. Now we have Core ML 3 and honestly outside of optimization alone I'm not too sure why you would need to drop down to metal after this new update.


5 myths about industrial AI

#artificialintelligence

Enterprise adoption of artificial intelligence (AI) has grown by more than 270% over the past four years, according to a Gartner report. Since AI has the capability to speed up business process and generate greater returns on investment (ROI), some 37% of organizations have now fully embraced the technology, the report found. As AI use cases have increased, many misconceptions have surfaced around the new technology. To help shed light on the conflicting viewpoints of AI, Teradata practice director Cheryl Wiebe shared the five most popular myths about industrial AI, and the truth behind each. Many organizations assume that tools like data analytics are able to immediately give organizations the exact answers they are looking for, according to Wiebe.