The Human Effort Behind AI Successes

March 14, 2017

An article at Recode, “Watson Claims to Predict Cancer, but Who Trained It To Think,” reminds us that even the most successful AI software was trained by humans, using data collected and input by humans. We have developed high hopes for AI, expecting it to help us cure disease, make our roads safer, and put criminals behind bars, among other worthy endeavors. However, we must not overlook the datasets upon which these systems are built, and the human labor used to create them. Writer (and CEO of DaaS firm Captricity) Kuang Chen points out:

The emergence of large and highly accurate datasets have allowed deep learning to ‘train’ algorithms to recognize patterns in digital representations of sounds, images and other data that have led to remarkable breakthroughs, ones that outperform previous approaches in almost every application area. For example, self-driving cars rely on massive amounts of data collected over several years from efforts like Google’s people-powered street canvassing, which provides the ability to ‘see’ roads (and was started to power services like Google Maps). The photos we upload and collectively tag as Facebook users have led to algorithms that can ‘see’ faces. And even Google’s 411 audio directory service from a decade ago was suspected to be an effort to crowdsource data to train a computer to ‘hear’ about businesses and their locations.

Watson’s promise to help detect cancer also depends on data: decades of doctor notes containing cancer patient outcomes. However, Watson cannot read handwriting. In order to access the data trapped in the historical doctor reports, researchers must have had to employ an army of people to painstakingly type and re-type (for accuracy) the data into computers in order to train Watson.

Chen notes that more and more workers in regulated industries, like healthcare, are mining for gold in their paper archives—manually inputting the valuable data hidden among the dusty pages. That is a lot of data entry. The article closes with a call for us all to remember this caveat: when considering each new and exciting potential application of AI, ask where the training data is coming from.

Cynthia Murrell, March 14, 2017

Chipping Away at Social Content with Pictures

February 27, 2017

Analytics are catching up to content. In a recent ZDNet article, Digimind Partners with Ditto to Add Image Recognition to Social Media Monitoring, we are reminded images reign supreme on social media. Between Pinterest, Snapchat and Instagram, messages are often conveyed through images as opposed to text. Capitalizing on this, an intelligence software company Digimind has announced a partnership with Ditto Labs to introduce image-recognition technology into their social media monitoring software called Digimind Social. We learned,

 “The Ditto integration lets brands identify the use of their logos across Twitter no matter the item or context. The detected images are then collected and processed on Digimind Social in the same way textual references, articles, or social media postings are analysed. Logos that are small, obscured, upside down, or in cluttered image montages are recognised. Object and scene recognition means that brands can position their products exactly where there customers are using them. Sentiment is measured by the amount of people in the image and counts how many of them are smiling. It even identifies objects such as bags, cars, car logos, or shoes.”

 It was only a matter of time before these types of features emerged in social media monitoring. For years now, images have been shown to increase engagement even on platforms that began focused more on text. Will we see more watermarked logos on images? More creative ways to visually identify brands? Both are likely and we will be watching to see what transpires.

 Megan Feil, February 27, 2017

 

Finding Meaning in Snapchat Images, One Billion at a Time

February 27, 2017

The article on InfoQ titled Amazon Introduces Rekognition for Image Analysis explores the managed service aimed at the explosive image market. According to research cited in the article, over 1 billion photos are taken every single day on Snapchat alone, compared to the 80 billion total taken in the year 2000. Rekognition’s deep learning power is focused on identifying meaning in visual content. The article states,

The capabilities that Rekognition provides include Object and Scene detection, Facial Analysis, Face Comparison and Facial Recognition. While Amazon Rekognition is a new public service, it has a proven track record. Jeff Barr, chief evangelist at AWS, explains: Powered by deep learning and built by our Computer Vision team over the course of many years, this fully-managed service already analyzes billions of images daily. It has been trained on thousands of objects and scenes. Rekognition was designed from the get-go to run at scale.

The facial analysis features include markers for image quality, facial landmarks like facial hair and open eyes, and sentiment expressed (smiling = happy.) The face comparison feature includes a similarity score that estimates the likelihood of two pictures being of the same person. Perhaps the most useful feature is object and scene detection, which Amazon believes will help users find specific moments by searching for certain objects. The use cases also span vacation rental markets and travel sites, which can now tag images with key terms for improved classifications.

Chelsea Kerwin, February 27, 2017

Mobile App Usage on the Rise from 34% of Consumer Time in 2013 to 50% in 2016

February 24, 2017

Bad news, Google. The article titled Smartphone Apps Now Account for Half the Time Americans Spend Online on TechCrunch reveals that mobile applications are still on the rise. Throw in tablet apps and the total almost hits 60%. Google is already working to maintain relevancy with its In Apps feature for Androids, which searches inside apps themselves. The article explains,

This shift towards apps is exactly why Google has been working to integrate the “web of apps” into its search engine, and to make surfacing the information hidden in apps something its Google Search app is capable of handling.  Our app usage has grown not only because of the ubiquity of smartphones, but also other factors – like faster speeds provided by 4G LTE networks, and smartphones with larger screens that make sitting at a desktop less of a necessity.

What apps are taking up the most of our time? Just the ones you would expect, such as Facebook, Messenger, YouTube, and Google Maps. But Pokemon Go is the little app that could, edging out Snapchat and Pinterest in the ranking of the top 15 mobile apps. According to a report from Senor Tower, Pokemon Go has gone beyond 180 million daily downloads. The growth of consumer time spent on apps is expected to keep growing, but comScore reassuringly states that desktops will also remain a key part of consumer’s lives for many years to come.

Chelsea Kerwin, February 24, 2017

 

Upgraded Social Media Monitoring

February 20, 2017

Analytics are catching up to content. In a recent ZDNet article, Digimind partners with Ditto to add image recognition to social media monitoring, we are reminded images reign supreme on social media. Between Pinterest, Snapchat and Instagram, messages are often conveyed through images as opposed to text. Capitalizing on this, and intelligence software company Digimind has announced a partnership with Ditto Labs to introduce image-recognition technology into their social media monitoring software called Digimind Social. We learned,

The Ditto integration lets brands identify the use of their logos across Twitter no matter the item or context. The detected images are then collected and processed on Digimind Social in the same way textual references, articles, or social media postings are analysed. Logos that are small, obscured, upside down, or in cluttered image montages are recognised. Object and scene recognition means that brands can position their products exactly where there customers are using them. Sentiment is measured by the amount of people in the image and counts how many of them are smiling. It even identifies objects such as bags, cars, car logos, or shoes.

It was only a matter of time before these types of features emerged in social media monitoring. For years now, images have been shown to increase engagement even on platforms that began focused more on text. Will we see more watermarked logos on images? More creative ways to visually identify brands? Both are likely and we will be watching to see what transpires.

Megan Feil, February 20, 2017

 

Penn State Research Team Uses Big Data to Explore Crime Rates

February 2, 2017

The article on E&T titled Social Media and Taxi Data Improve Crime Pattern Picture delves into a fascinating study that uses big data involving taxi routes and social media location labels from sites like Foursquare to discover a correlation between taxis, locations of interest, and crime. The study was executed by Penn State researchers who are looking for a more useful way to estimate crime rates rather than the traditional approach targeting demographics and geographic data only. The article explains,

The researchers say that the analysis of crime statistics that encompass population, poverty, disadvantage index and ethnic diversity can provide more accurate estimates of crime rates … the team’s approach likens taxi routes to internet hyperlinks, connecting different communities with each other… One surprising discovery is that the data suggests areas with nightclubs tend to experience lower crime rates – at least in Chicago.  The explanation may be that it reflects people’s choices to be there.

This research will be especially useful to city planners interested in how certain spaces are being used, and whether people want to go to those spaces. But the researcher Jessie Li, an assistant professor of information sciences, explained that while the correlation is clear, the underlying cause is not yet known.

Chelsea Kerwin, February 2, 2017

 

Twitter: Selling and Banning Its Way to Its Future

January 23, 2017

Twitter is making news again. The company sold some tools to the Google. Google, wisely Beyond Search thinks, has not yet built up the gumption to buy the whole Twitter enchilada. And Twitter continues to annoy some professionals who use Twitter data to figure out the who, what, and why of certain illegal activities.

Twitter Bans Award-Winning London, Ont., Company for Helping Police Track Protesters” explains:

A London, Ont., data mining company has been banned from Twitter and is being reviewed by Facebook for selling surveillance software to North American police services to monitor people at Black Lives Matter events and other public protests.

The company in question is Media Sonar, one of a number of firms which developed tools to make sense of messages and metadata generated by the folks who send information via Twitter “tweets”. (You can watch a video explaining some of the firm’s methods at this link.) Another example of a social media analysis outfit is Geofeedia which has been given a bloody nose by spasmodic Silicon Valley wizards.

The write up reports:

Media Sonar did not return calls to CBC News but its website states that it works to help clients analyze the sentiment of social media posts and can use location-based data to monitor threats.

Beyond Search believes that some high flying Silicon Valley companies develop systems and do not think about how these systems will be used. Then  when the high flying Silicon Valley executives realize that their whizzy new creation has some interesting applications, the Twitter-type outfits take action. The approach is fascinating to watch.

On one hand, Twitter is struggling to develop its user base and get some sizzle back. On the other hand, the company is selling off grandma’s furniture and turning off revenue from licensees of the Twitter content stream.

Interesting stuff. Chaos monkeys in real life? Seems like it.

Stephen E Arnold, January 23, 2017

The Internet Is Once Again Anonymous

January 19, 2017

Let us reminiscence for a moment (and if you like you can visit the Internet archive) about the Internet’s early days, circa late 1990s.  It was a magic time, because there were chatrooms, instant messaging, and forums.  The Internet has not changed these forms of communication much, although chatrooms are pretty dead, but one great thing about the early days is that the Internet was mostly anonymous.  With the increase in tracking software, IP awareness, and social media, Internet anonymity is reserved for the few who are vigilant and never post anything online.    Sometimes, however, you want to interact online without repercussions and TechCrunch shares that “Secret Founder Returns To Anonymous Publishing With Launch Of IO.”

David Byttow, Secret co-founder, started the anonymous publishing app IO that is similar to Postcard Confessions.  IO’s purpose is to:

IO is a pseudo-resurrection of Secret that Byttow told us in November came into being partly because “the downsides of current social media products MUST be addressed,” an imperative he felt was especially urgent following the results of the last U.S. election. IO’s stated mission is to achieve “authentic publishing,” by which Byttow means that he’s hoping users having an option to publishing either anonymously, using a pseudonym or as their actual selves will allow for easier sharing of true thoughts and feelings.

IO really does not do much.  You can type something up, hit publish, but it is only shared with other people if you attach social media links.  You can remain anonymous and IO does include writing assistance tools.  I really do not get why IO is useful, but it does allow a person to create a shareable link without joining a forum, owning a Web site, etc.  Reddit seems more practical, though.

Whitney Grace, January 19, 2016

 

The Government Has a Sock Puppet Theater

January 13, 2017

Law enforcement officials use fake social media accounts and online profiles to engage with criminals.  Their goal is to deter crime, possibly even catching criminals in the act for a rock solid case.  While this happened way back in 2011, the comments are still coming.  In light of the recent presidential election and the violent acts of the past year, it is no wonder the comments are still fresh.  Tech Dirt talked about how the, “US Military Kicks Off Plan To Fill Social Networks With Fake Sock Puppet Accounts.”

The goal was for a company to develop a software that would allow one person to create and manage various social media profiles (including more than one profile on the same platform).  These accounts will then, and we are speculating on this given how dummy accounts have been used in the past, to catch criminals.  The article highlights how the government would use the sock puppet accounts:

Apparently a company called Ntrepid has scored the contract and the US military is getting ready to roll out these “sock puppet” online personas. Of course, it insists that all of this is targeting foreign individuals, not anyone in the US. And they promise it’s not even going to be used on US-based social networks like Facebook or Twitter, but does anyone actually believe that’s true?

Then the comments roll in a conversation that a span of five years the commentators argue about what it means to be American, reaffirming that the US government spies on its citizens, and making fun of sock puppets.

Whitney Grace, January 13, 2017

Yahoo Takes on ISIS, in Its Way

January 9, 2017

The article on VentureBeat titled Yahoo Takes Steps to Remove Content Posted From ISIS and Other Terrorist Groups remarks on the recent changes Yahoo made to its community guidelines. The updated guidelines now specify that any content or accounts involved with terrorist organizations, even those that “celebrate” violence connected to terrorist activity are up for deletion or deactivation. The article speaks to the relevance of these new guidelines that follow hard upon the heels of Orlando and San Bernardino,

Twitter has responded as well, “suspending over 125,000 accounts” related to terrorism. Messaging app Telegram has also blocked 78 channels that engaged in ISIS-related activity. Kathleen Lefstad, Yahoo’s policy manager for trust and safety, wrote that this new category is in addition to other types of content that are flagged, including hate speech, bullying or harassment, and sharing adult or sexualized content of someone without their consent.

ISIS has grown infamous for its social media presence and ability to draw foreign supporters through social media platforms. Yahoo’s crackdown is a welcome sign of awareness that these platforms must take some responsibility for how their services are being abused. Priorities, folks. If Facebook’s machine learning content security can remove any sign of a woman’s nipple within 24 hours, shouldn’t content that endorses terrorism be deleted in half the time?

Chelsea Kerwin, January 9, 2017

Next Page »

  • Archives

  • Recent Posts

  • Meta