Daily Archives: August 9, 2017

Kavya Kopparapu might be considered something of a whiz kid. After all, she had yet to enter her senior year of high school when she started Eyeagnosis, a smartphone app and 3D-printed lens that allows patients to be screened for diabetic retinopathy with a quick photo, avoiding the time and expense of a typical diagnostic procedure. In June 2016, Kopparapus grandfather had recently been diagnosed with diabetic retinopathy, a complication of diabetes that damages retinal blood vessels and can eventually cause blindness. He caught the symptoms in time to receive treatment, but it was close. A little too close for Kopparapus comfort. According to the IEEE Spectrum, Kopparapu, her 15-year-old brother Neeyanth and her classmate Justin Zhang trained an artificial intelligence system to scan photos of eyes and detect, and diagnose, signs of diabetic retinopathy. She unveiled the technology at the OReilly Artificial Intelligence conference in New York City in July. After diving into internet-based research and emailing opthamologists, biochemists, epidemiologists, neuroscientists and the like, she and her team worked on the diagnostic AI using a machine-learning architecture called a convolutional neural network. CNNs, as theyre called, parse through vast data sets -- like photos -- to look for patterns of similarity, and to date have shown an aptitude for classifying images. The network itself was the ResNet-50, developed by Microsoft. But to train it to make retinal diagnoses, Kopparapu had to feed it images from the National Institute of Healths EyeGene database, which essentially taught the architecture how to spot signs of retinal degeneration. One hospital has already tested the technology, fitting a 3D-printed lens onto a smartphone and training the phones flash to illuminate the retinas of five different patients. Tested against opthalmologists, the system went five for five on diagnoses. Kopparapus invention still needs lots of tests and additional data to prove its efficacy before it sees widespread clinical adoption, but so far, its off to a pretty good start. Eyeagnosis is operating in a space that's recently become interesting to some very large companies. Last fall, a team of Google researchers published a paper in the Journal of the American Medical Association showing that Google's deep learning algorithm, trained on a large data set of fundus images, can detect diabetic retinopathy with better than 90 percent accuracy. That algorithm was then tested on 9,963 deidentified images retrospectively obtained from EyePACS in the United States, as well as three eye hospitals in India. A second, publicly available research data set of 1,748 was also used. The accuracy was determined by comparing its diagnoses to those done by a panel of at least seven U.S. board-certified ophthalmologists. The two data sets had 97.5 percent and 96.1 percent sensitivity, and 93.4 percent and 93.9 percent specificity respectively.

And Google isnt the only player in that space. IBM has a technology utilizing a mix of deep learning, convolutional neural networks and visual analytics technology based on 35,000 images accessed via EyePACs; in research conducted earlier this year, the technology learned to identify lesions and other markers of damage to the retinas blood vessels, collectively assessing the presence and severity of disease. In just 20 seconds, the method was successful in classifying diabetic retinopathy severity with 86 percent accuracy, suggesting doctors and clinicians could use the technology to have a better idea of how the disease progresses as well as identify effective treatment methods.

Lower-tech options are also taking a stab at improving access to screenings. Using a mix of in-office visits, telemedicine and web-based screening software, the Los Angeles Department of Health Services has been able to greatly expand the number of patients in its safety net hospital who got screenings and referrals. In an article published in the journal JAMA Internal Medicine, researchers describe how the two-year collaboration using Safety Net Connects eConsult platform resulted in more screenings, shorter wait times and fewer in-person specialty care visits. By deploying Safety Net Connects eConsult system to a group of 21,222 patients, the wait times for screens decreased by almost 90 percent, and overall screening rates for diabetic retinopathy increased 16 percent. The digital program also eliminated the need for 14,000 visits to specialty care professionals

More:

Teenage team develops AI system to screen for diabetic retinopathy - MobiHealthNews

Artificial intelligence as a technology is rapidly growing, but much is still being learned about how AI and autonomous systems make decisions based on the information they collect and process.

To better explain those relationships so humans and autonomous systems can better understand each other and collaborate more deeply, researchers at PARC, the Palo Alto Research Center, have been awarded a multi-million dollar federal government contract to create an "interactive sense-making system" that could answer many related questions.

The research for the proposed system, called COGLE (CommonGroundLearning andExplanation), is being funded by the Defense Advanced Research Projects Agency (DARPA),using an autonomous Unmanned Aircraft System (UAS) test bed but would later be applicable to a variety of autonomous systems.

The idea is that since autonomous systems are becoming more widely used, it would behoove humans who are using them to understand how the systems behave based on the information they are provided, Mark Stefik, a PARC research fellow who runs the lab's human machine collaboration research group, told ITPro.

"Machine learning is becoming increasing important," said Stefik. "As a consequence, if we are building systems that are autonomous, we'd like to know what decisions they will make. There is no established technique to do that today with systems that learn for themselves."

In the field of human psychology, there is an established history about how people form assumptions about things based on their experiences, but since machines aren't human, their behaviors can vary, sometimes with results that can be harmful to humans, said Stefik.

In one moment, an autonomous machine can do something smart or helpful, but then the next moment it can do something that is "completely wonky, which makes things unpredictable," he said. For example, a GPS system seeking the shortest distance between two points could erroneously and catastrophically send a user driving over a cliff or the wrong way onto a one-way street. Being able to delve into those autonomous "thinking" processes to understand them is the key to this research, said Stefik.

The COGLE research will help researchers pursue answers to these issues, he said. "We're insisting that the program be explainable," for the autonomous systems to say why they are doing what they are doing. "Machine learning so far has not really been designed to explain what it is doing."

The researchers involved with the project will essentially have roles educators and teachers for the machine learning processes to improve their operations and make it more useable and even more human like, said Stefik. "It's a sort of partnership where humans and machines can learn from each other."

That can be accomplished in three ways, he added, including reinforcement at the bottom level, using reasoning patterns like the ones humans use at the cognitive or middle level, and through explanation at the top sense-making level. The research aims to enable people to test, understand, and gain trust in AI systems as they continue to be integrated into our lives in more ways.

The research project is being conducted under DARPA's Explainable Artificial Intelligence (XAI) program, which seeks to create a suite of machine learning techniques that produce explainable models and enable human users to understand, appropriately trust, and effectively manage the emerging generation of artificially intelligent partners.

PARC, which is a Xerox company, is conducting the COGLE work with researchers at Carnegie Mellon University, West Point, the University of Michigan, the University of Edinburgh and the Florida Institute for Human & Machine Cognition. The key idea behind COGLE is to establish common ground between concepts and abstractions used by humans and the capabilities learned by a machine. These learned representations would then be exposed to humans using COGLE's rich sense-making interface, enabling people to understand and predict the behavior of an autonomous system.

Go here to read the rest:

Advancing AI by Understanding How AI Systems and Humans Interact - Windows IT Pro

Brands have long been able to search for company mentions on social media, but theyve lacked the ability to search for pictures of their logos or products in an easy way. Thats where Salesforces latest Einstein artificial intelligence feature comes into play.

Today the company introduced Einstein Vision for Social Studio, which provides a way for marketers to search for pictures related to their brands on social media in the same way they search for other mentions. The product takes advantage of a couple of Einstein artificial intelligence algorithms including Einstein Image Classification for image recognition. It uses visual search, brand detection and product identification. It also makes use of Einstein Object Detection to recognize objects within images including the type and quantity of object.

AI has gotten quite good at perception and cognition tasks in recent years. One result of this has been the ability to train an algorithm to recognize a picture. With cheap compute power widely available and loads of pictures being uploaded online, it provides a perfect technology combination for better image recognition.

Rob Begg, VP of product marketing for social and advertising products at Salesforce, says its about letting the machine loose on tasks for which its better suited. If you think of it from a company point of view, there is a huge volume of tweets and [social] posts. What AI does best is help surface and source the ones that are relevant, he says.

As an example, he says there could be thousands of posts about cars, but only a handful of those would be relevant to your campaign. AI can help find those much more easily.

Begg sees three possible use cases for this tool. First of all, it could provide better insight into how people are using your products. Secondly it could provide a way to track brand displays online hidden within pictures, and finally it could let you find out when influencers such as actors or athletes are using your products.

The product comes trained to recognize two million logos, 60 scenes (such as an airport), 200 foods and 1000 objects. That should be enough to get many companies started. Customizing isnt available in the first release, so if you have a logo or object not included out of the box, you will need to wait for a later version to be able to customize the content.

Begg says it should be fairly easy for marketers used to using Social Studio to figure out how to incorporate the visual recognition tools into their repertoire. The new functionality should be available immediately to Salesforce Social Studio users.

Follow this link:

Salesforce AI helps brands track images on social media | TechCrunch - TechCrunch

Marc Vontobel, CTO & Pascal Kaufmann, CEO, Starmind

Artificial intelligence is stuck today because companies are likening the human brain to a computer, according to Swiss neuroscientist and co-founder of Starmind Pascal Kaufmann. However, the brain does not process information, retrieve knowledge, or store memories like a computer does.

When companies claim to be using AI to power "the next generation" of their products, what they are unknowingly referring to is the intersection of big data, analytics, and automation, Kaufmann told ZDNet.

"Today, so called AI is often just the human intelligence of programmers condensed into source code," said Kaufmann, who worked on cyborgs previously at DARPA.

"We shouldn't need 300 million pictures of cats to be able to say whether something is a cat, cow, or dog. Intelligence is not related to big data; it's related to small data. If you can look at a cat, extract the principles of a cat like children do, then forever understand what a cat is, that's intelligence."

He even said that it's not "true AI" that led to AlphaGo -- a creation of Google subsidiary DeepMind -- mastering what is revered as the world's most demanding strategy game, Go.

The technology behind AlphaGo was able to look at 10 to 20 potential future moves and lay out the highest statistics for success, Kaufmann said, and so the test was one of rule-based strategy rather than artificial intelligence.

The ability for a machine to strategise outside the context of a rule-based game would reflect true AI, according to Kaufmann, who believes that AI will cheat without being programmed not to do so.

Additionally, the ability to automate human behaviour or labour is not necessarily a reflection of machines getting smarter, Kaufmann insisted.

"Take a pump, for example. Instead of collecting water from the river, you can just use a pump. But that is not artificial intelligence; it is the automation of manual work ... Human-level AI would be able to apply insights to new situations," Kaufmann added.

While Facebook's plans to build a brain-computer interface and Elon Musk's plans to merge the human brain with AI have left people wondering how close we are to developing true AI, Kaufmann believes the "brain code" needs to be cracked before we can really advance the field. He said this can only be achieved through neuroscientific research.

Earlier this year, founder of DeepMind Demis Hassabis communicated a similar sentiment in a paper, saying the fields of AI and neuroscience need to be reconnected, and that it's only by understanding natural intelligence that we can develop the artificial kind.

"Many companies are investing their resources in building faster computers ... we need to focus more on [figuring out] the principles of the brain, understand how it works ... rather than just copy/paste information," Kaufmann said.

Kaufmann admitted he doesn't have all the answers, but finds it "interesting" that high-profile entrepreneurs such as Musk and Mark Zuckerberg, none of whom have AI or neuroscience backgrounds, have such strong and opposing views on AI.

Musk and Zuckerberg slung mud at each other in July, with the former warning of "evil AI" destroying humankind if not properly monitored and regulated, while the latter spoke optimistically about AI contributing to the greater good, such as diagnosing diseases before they become fatal.

"One is an AI alarmist and the other makes AI look charming ... AI, like any other technology, can be used for good or used for bad," said Kaufmann, who believes AI needs to be assessed objectively.

In the interim, Kaufmann believes systems need to be designed so that humans and machines can work together, not against each other. For example, Kaufmann envisions a future where humans wear smart lenses -- comparable to the Google Glass -- that act as "the third half of the brain" and pull up relevant information based on conversations they are having.

"Humans don't need to learn stuff like which Roman killed the other Roman ... humans just need to be able to ask the right questions," he said.

"The key difference between human and machine is the ability to ask questions. Machines are more for solutions."

Kaufmann admitted, however, that humans don't know how to ask the right questions a lot of the time, because we are taught to remember facts in school, and those who remember the most facts are the ones who receive the best grades.

He believes humans need to be educated to ask the right questions, adding that the question is 50 percent of the solution. The right questions will not only allow humans to understand the principles of the brain and develop true AI, but will also keep us relevant even when AI systems proliferate, according to Kaufmann.

If we want to slow down job loss, AI systems need to be designed so that humans are at the centre of it, Kaufmann said.

"While many companies want to fully automate human work, we at Starmind want to build a symbiosis between humans and machines. We want to enhance human intelligence. If humans don't embrace the latest technology, they will become irrelevant," he added.

The company claims its self-learning system autonomously connects and maps the internal know-how of large groups of people, allowing employees to tap into their organisation's knowledge base or "corporate brain" when they have queries.

Starmind platform

Starmind is integrated into existing communication channels -- such as Skype for Business or a corporate browser -- eliminating the need to change employee behaviour, Kaufmann said.

Questions typed in the question window are answered instantly if an expert's answer is already stored in Starmind, and new questions are automatically routed to the right expert within the organisation, based on skills, availability patterns, and willingness to share know-how. All answers enhance the corporate knowledge base.

"Our vision is if you connect thousands of human brains in a smart way, you can outsmart any machine," Kaufmann said.

On how this is different to asking a search engine a question, Kaufmann said Google is basically "a big data machine" and mines answers to questions that have been already asked, but is not able to answer brand new questions.

"The future of Starmind is we actually anticipate questions before they're even asked because we know so much about the employee. For example, we can say if you are a new hire and you consume a certain piece of content, there will be a 90 percent probability that you will ask the following three questions within the next three minutes and so here are the solutions."

Starmind is being currently used across more than 40 countries by organisations such as Accenture, Bayer, Nestl, and Telefonica Deutschland.

While Kaufmann thinks it is important at this point in time to enhance human intelligence rather than replicate it artificially, he does believe AI will eventually substitute humans in the workplace. But unlike the grim picture painted by critics, he doesn't think it's a bad thing.

"Why do humans need to work at all? I look forward to all my leisure time. I do not need to work in order to feel like a human," Kaufmann said.

When asked how people would make money and sustain themselves, Kaufmann said society does not need to be ruled by money.

"In many science fiction scenarios, they do not have money. When you look at the ant colonies or other animals, they do not have cash," Kaufmann said.

Additionally, if humans had continuous access to intelligent machines, Kaufmann said "the acceleration of human development will pick up" and "it will give rise to new species".

"AI is the ultimate tool for human advancement," he firmly stated.

Link:

True AI cannot be developed until the 'brain code' has been cracked: Starmind - ZDNet

GETTY

In parts of the US, when a suspect is taken in for questioning they are given a computerised risk assessment which works out the likelihood of the person reoffending.

The judges can then use this data when giving his or her verdict.

However, an investigation has revealed that the artificial intelligence behind the software exhibits racist tendencies.

Reporters from ProPublica obtained more than 7,000 test results from Florida in 2013 and 2014 and analysed the reoffending rate among the individuals.

GETTY

The suspects are asked a total of 137 questions by the AI system Correctional Offender Management Profiling for Alternative Sanctions (Compas) including questions such as Was one of your parents ever sent to jail or prison? or How many of your friends/acquaintances are taking drugs illegally?, with the computer generating its results at the end.

Overall, the AI system claimed black people (45 per cent) were almost twice as likely as white people (24 per cent) to reoffend.

In one example outlined by ProPublica, risk scores were provided for a black and white suspect, both of which on drug possession charges.

GETTY

The white suspect had prior offences of attempted burglary and the black suspect had resisting arrest.

Seemingly giving no indication as to why, the black suspect was given a higher chance of reoffending and the white suspect was considered low risk.

But, over the next two years, the black suspect stayed clear of illegal activity and the white suspect was arrested three more times for drug possession.

However, researchers warn the problem does not lie with robots, but with the human race as AI uses machine learning algorithms to pick up on human traits.

Joanna Bryson, a researcher at the University of Bath, told the Guardian: People expected AI to be unbiased; thats just wrong. If the underlying data reflects stereotypes, or if you train AI from human culture, you will find these things.

Asus

1 of 9

Asus Zenbo: This adorable little bot can move around and assist you at home, express emotions, and learn and adapt to your preferences with proactive artificial intelligence.

This is not an isolated incident either.

Microsofts TayTweets (AI) chatbot was unleashed on Twitter last year which was designed to learn from users.

However, it almost instantly turned to anti-semitism and racism, tweeting: "Hitler did nothing wrong" and "Hitler was right I hate the Jews.

See the original post:

REVEALED: AI is turning RACIST as it learns from humans - Express.co.uk

The U.S. Department of Energy wants to make investing in energy technology easier, less risky, and less expensive (for the government, at least).

A new initiative by the DOEs office of Energy Efficiency & Renewable Energy (EERE) is looking for ideas on how to reduce barriers to private investment in energy technologies. Rho AI, one of 11 companies awarded a grant through the EEREs US $7.8-million programcalled Innovative Pathways, plans to use artificial intelligence and data science to efficiently connect investors to startups. By using natural language processing tools to sift through publicly available information, Rho AI will build an online network of potential investors and energy technology companies, sort of like a LinkedIn for the energy sector.The Rho AI team wants to develop a more extensive network than any individual is capable of having on their own, and theyre relying on artificial intelligence to make smarter connections faster than a human could.

Youre limited by the human networking capability when it comes to trying to connect technology and investment, says Josh Browne, co-Founder and vice president of operations at Rho AI. Theres only so many hours in a day and theres only so many people in your network.

Using theUS $750,000 it received from the DOE, Rho AI has just two years to build, test, and prove the efficacy of its system. The two-year timeline for demonstrating proof of concept is a stipulation of the grant. With this approach,the DOE hopes to streamline the underlying process for getting new energy technologies to the market, instead of investing inparticularcompanies.

Its a fairly small grant, relative to some of the larger grants where they invest in the actual hard technology, Browne says. In this case, theyre investing in ways to unlock money to invest in hard technology.

Rho AIs database will not only contain information about energy technology companies and investor interests, it will also track where money is coming from and who its going to in the industry. Browne imagines the interface will look something like a Bloomberg terminal.

To build the database, Rho AI will use Google Tensor Flow and Natural Language Toolkittools that can read and analyze human languageto scan public documents such as Securities and Exchange Commission filings and news articles on energy companies. The system will then use software tools that help analyze and visualize patterns in data, such as MUXviz and NetMiner, to understand how people and companies are connected.

In order to measure how well its helping investors and emerging clean technology companies find better business partners faster, Rho AI will compare the machine-built network with the real professional network of Carmichael Roberts, a leading venture capitalist in clean technology.

This tool is intended to emulate and perhaps surpass the networking capability of a leading clean tech venture capitalist, Browne says. It should be able to match their network, and it should be able to very rapidly be ten times their network.

Rho AIs program should create a longer, more comprehensive list of possible investments than Roberts canwithin seconds. The intention is for the final product to be robust enough that members of the private sector could and would adopt it after one year.

If Rho AI is able to be successful in what theyre building, that will be in some sense self-scaling, says Johanna Wolfson, director of the tech-to-market program at the DOE.

In other words, Rho AI could grow on its own and the industry could start seeing the effects of these connections. Investors and clean energy technology companies could find each other directly, while reducing the burden on the government to invest so much in energy innovation.

Improving the underlying pathway for getting new energy technology to market actually can be done for relatively small dollar amounts, relative to what the government sometimes supports, in ways that can be catalytic, but sustained by the private sector, said Wolfson.

Editors note: This post was corrected on August 8to reflect thespecifications of the DOEs grant.

IEEE Spectrums energy, power, and green tech blog, featuring news and analysis about the future of energy, climate, and the smart grid.

Sign up for the EnergyWise newsletter and get biweekly news on the power & energy industry, green technology, and conservation delivered directly to your inbox.

CCS still not close to widespread deployment, in spite of continued government support16Aug2011

South Carolina utilities abandoned a pair of troubled reactors projected to cost more than twice their original price 2Aug

Why the carmakers audacious plan to electrify every Volvo from 2019 may stall out 21Jul

Clean coal technology suffered a setback when efforts to start up the gasification portion of an IGCC plant in Mississippi were halted 30Jun

The world running on wind, sunlight, and hydropowerchampioned by Stanford's Mark Jacobsonhas captured the public imagination, but it faces a fierce attack from 21 climate and power-grid experts 19Jun

Leaders of major tech companies say they have good reason to stay the courseeven if the federal government won't 2Jun

Methane rules and the Paris Accord expose friction within the GOP and the Trump administration over climate and energy policy 15May

Values are personal, not necessarily logical, and when applied to electricity choices they can impact the market in unpredictable ways 11May

The incubator agencys 2017 budget victory last week still says little about its fate in 2018 9May

A crisis that threatened Southern California's electric grid enabled energy storage to demonstrate its flexibility and rapid deployment 8May

When considering only the portion of fossil fuel support that relates to electric power, renewables receive far more federal help 4May

Unfavorable market conditions are forcing some nuclear power plants to close, removing a carbon-free source of power from the U.S. grid. Plant operators are fighting back 18Apr

A mathematical rethink suggests that carbon dioxide will warm Earth more in the future than it does today. But better satellitessuch as those Trump wants to scrapare needed to reduce climate uncertainty 17Apr

The sweeping attack on climate action that President Trump demanded in his executive order is likely to prove but short-lived relief for coal miners who cheered him at the EPA 29Mar

How changes in regulations allowed wind power to increase in Texas without increasing operational reserves 23Mar

The new president could use advice from a practical problem solver 22Mar

President Donald Trump outlined a fiscal 2018 budget request that asks Congress to stamp out federal climate science and slash investment in energy innovation 17Mar

NASAs new geosensing satellites may be on the chopping block. The timing could hardly be worse 9Mar

Market-calibrated forecasts for natural gas prices show historical trends, bode well for future updates with recent data sets 7Mar

The EPA approaches this task differently for greenhouse gases than for other pollutants 23Feb

Go here to read the rest:

DOE Backs AI for Clean Tech Investors - IEEE Spectrum