Facebook teaches machines to negotiate with humans

Though Facebook is rarely mentioned alongside Apple, Microsoft and Amazon in discussions about conversational AI, the company has published a hoard of papers that underscore a deep interest in dialog systems. As has become clear with Siri, Cortana and Alexa, dialog is hard it requires more than just good speech recognition to deliver a killer experience to users. From the sidelines Facebook has been tinkering with big challenges like natural language understanding and text generation. And today the Facebook AI Research team added to its portfolio with a paper bringingnegotiation into the conversation (all puns intended).

Facebooks team smashed game theory together with deep learning to equip machines to negotiate with humans. By applying rollout techniques more commonly used in game-playing AIs to a dialog scenario, Facebook was able to create machines capable of complex bargaining.

To start, Facebook dreamed up an imaginary negotiation scenario. Humans on Amazons Mechanical Turk were given an explicit value function and told to negotiate in natural language to maximize reward by splitting up a pot of random objects fivebooks, three hats and two balls. The game was capped at ten rounds of dialog, the rules stated that nobody would receive any reward if that limit was exceeded.

Because each agent had distinct hidden preferences, the two had to engage in dialog to sort out which objects should be given to which agent. Over the course of the interactions, machines naturally adopted many common negotiation tactics like placing false emphasis on a low-value item in an attempt to use it as a more valuable bargaining chip later.

Under the hood, Facebooks rollout technique takes the form of a decision tree. Decision trees are a critical component of many intelligent systems. They allow us to model future states from the present to make decisions. Imagine a game of tic-tac-toe, at any given point of the game, there is a finite option set (places you can place your X on the board.

In that scenario, each move has an expected value. Humans dont usually consider this value in an explicit way but if you decompose your decision process when playing the game, you are effectively short-handing this math in your head.

Games like Tic Tac Toe are simple enough that they can be completely solved in a decision tree. More complex games like Go and Chess require strategies and heuristics to reduce the total number of states (its an almost unimaginable number of possible states). But even Chess and Go are relatively simple compared to dialog.

Dialog doesnt draw from a finite set of outcomes. This means that for any question, there is an infinite number of possible human responses. To model a conversation, researchers have to take extra effort to bound the uncertainty problem into a reasonable size and scope. Opting to model a negotiation scheme makes this possible. The language itself can exist in an infinite number of states but its intent generally clusters around simple outcomes (Ill take the deal or reject it).

But even in a bounded world, its still difficult to get machines to interact with humans in a believable way. To this avail, Facebook trained its models on negotiations between pairs of people. Once this was done, the machines were set up to negotiate with each other using reinforcement learning. At the end of each round of conversation, agents received rewards to guide improvement.

FAIR researchersMichael Lewis andDhruv Batra explained to me that their algorithms were better at preventing individuals from making bad decisions than ensuring individuals made the best decisions. This is still important the team told me to imagine a calendar application that doesnt try to schedule meetings for the best time for everyone but instead tries to just ensure the meeting actually happens.

As with a lot of research, the application of this technology isnt necessarily as explicit as the scenario simulated for the paper. Engineers often employ adversarial relationships between machines to improve outcomes think using generative adversarial networks to generate training data by having a machine generate data looking to fool another gatekeeper machine.

Semi-cooperative, semi adversarial relationships, like the relationship between a coach and an athlete, could be an interesting next frontier further connecting game theory and machine learning.

Facebook has open sourced its code from this research project. If youre interested, you can read additional details about the work in the full paper here.

Read more: https://techcrunch.com/2017/06/14/facebook-teaches-machines-to-negotiate-with-humans/

Advertisements

Apples Watch can detect an abnormal heart rhythm with 97% accuracy, UCSF study says

According to a study conducted through heartbeat measurement app Cardiogram and the University of California, San Francisco, the Apple Watch is 97 percent accurate in detecting the most common abnormal heart rhythm when paired with an AI-based algorithm.

The study involved 6,158 participants recruited through the Cardiogram app on Apple Watch. Most of the participants in the UCSF Health eHeart study had normal EKG readings. However, 200 of them had been diagnosed with paroxysmal atrial fibrillation (an abnormal heartbeat). Engineers thentrained a deep neural network to identify these abnormal heart rhythms from Apple Watch heart rate data.

Cardiogram began the study with UCSF in 2016 to discover whether the Apple Watch could detect an oncoming stroke.About a quarter of strokes are caused by an abnormal heart rhythm, according to Cardiogram co-founder and data scientist for UCSFs eHeart study Brandon Ballinger.

Cardiogram tested the deep neural network it had built against 51 in-hospital cardioversions (a procedure that restores the hearts normal rhythm) and says it achieved a 97 percent accuracy in the neural networks ability to find irregular heart activity.

So far this is just a study built on a preliminary algorithm but it holds promise in trying to identify and prevent stroke in the future. Atrial fibrillation, the most common abnormal heart rhythm, causes 1 in 4 strokes. Ballinger says two-thirds of those types of strokes are preventable with relatively inexpensive drugs.

And more people, including older populations most prone to stroke risk, are starting to use wearable technology such as Fitbit or the Apple Watch, which can double as heart monitors. Including algorithms trained to identify heart problems could help save lives in some of these more at-risk populations.

It should be noted mobile EKG readers have also made great strides in the past few years. The Mayo Clinic teamed up on a study involving AI and AliveCors version of an EKGreader, which sticks onto the back of a smartphone and uses the Kardia app to determine abnormal heart rhythm, and determined it was as good as other EKG devices used in the doctors office. The Mayo Clinic felt so strongly about this study it invested in AliveCors latest $30 million round.

In the meantime, Cardiogram and UCSF will continue its eHealth study and plan to further validate its deep neural network against multiple gold standards, incorporating the results into the Cardiogram app itself, and investigating the ability to detect health conditions beyond atrial fibrillation, according to Cardiogram.

Read more: https://techcrunch.com/2017/05/11/apples-watch-can-detect-an-abnormal-heart-rhythm-with-97-accuracy-ucsf-study-says/

Jack Dorsey just spent $9.5M buying more Twitter stock

Jack Dorsey is at it again after purchasing $7 million in Twitter stock back in February, the Twitter CEO bought another $9.5 million today.After making his purchase, Dorsey posted the news on Facebook Twitter, sparking a one percent gain in the companys share price in after-hours trading.

Sourced from Google Finance.

According to an SEC filing, Dorsey purchased574,002 shares of Twitter stock at roughly $16.62. That price is either a steal, if you compare it to Twitters peak share price of almost $70or a pretty bad deal if youre looking at last weeks share price that was $2 lower. Either way, the CEO is expressing confidence in his company that just posted its first decent earnings report in a while.

Earlier this week, Twitter reported to investors that it brought in $548 million in revenue in Q1 2017, beating investor expectations of $511.9 million. That sent Twitter stock on an upward trajectory, benefitting shareholders. But despite all the good news, Dorsey hasnt actually made much fromhis purchase of over 400,000 shares back in February.

With both purchases combined,Dorsey has now grabbed up over1 million additional shares of Twitter stock this year alone. In contrast, he has been dumping stock from his side-hustle Square. The share price for Square stock has been on the rise.The money from Dorseys sale of Square stock has previously gone to fund his ownStart Small Foundation.

Read more: https://techcrunch.com/2017/04/28/jack-dorsey-just-spent-9-5m-buying-more-twitter-stock/

Innovation in AI could see governments introduce human quotas, study says

Report predicts rise in robotics will usher in industrial revolution 4.0 altering working practices and legal frameworks

Innovation in artificial intelligence and robotics could force governments to legislate for quotas of human workers, upend traditional working practices and pose novel dilemmas for insuring driverless cars, according to a report by the International Bar Association.

The survey, which suggests that a third of graduate level jobs around the world may eventually be replaced by machines or software, warns that legal frameworks regulating employment and safety are becoming rapidly outdated.

The competitive advantage of poorer, emerging economies based on cheaper workforces will soon be eroded as robot production lines and intelligent computer systems undercut the cost of human endeavour, the study suggests.

While a German car worker costs more than 40 (34) an hour, a robot costs between only 5 and 8 per hour. A production robot is thus cheaper than a worker in China, the report notes. Nor does a robot become ill, have children or go on strike and [it] is not entitled to annual leave.

The 120-page report, which focuses on the legal implications of rapid technological change, has been produced by a specialist team of employment lawyers from the International Bar Association, which acts as a global forum for the legal profession.

The report covers both changes already transforming work and the future consequences of what it terms industrial revolution 4.0. The three preceding revolutions are listed as: industrialisation, electrification and digitalisation. Industry 4.0 involves the integration of the physical and software in production and the service sector. Amazon, Uber, Facebook, smart factories and 3D printing, its says, are among current pioneers.

The reports lead author, Gerlind Wisskirchen an employment lawyer in Cologne who is vice-chair of the IBAs global employment institute, said: What is new about the present revolution is the alacrity with which change is occurring, and the broadness of impact being brought about by AI and robotics.

Jobs at all levels in society presently undertaken by humans are at risk of being reassigned to robots or AI, and the legislation once in place to protect the rights of human workers may be no longer fit for purpose, in some cases … New labour and employment legislation is urgently needed to keep pace with increased automation.

Peering into the future, the authors suggest that governments will have to decide what jobs should be performed exclusively by humans for example, caring for babies. The state could introduce a kind of human quota in any sector, and decide whether it intends to introduce a made by humans label or tax the use of machines, the report says.

Increased mechanical autonomy will cause problems of how to define legal responsibility for accidents involving new technology such as driverless cars. Will it be the owner, the passengers, or manufacturers who pay the insurance?

The liability issues may become an insurmountable obstacle to the introduction of fully automated driving, the study warns. Driverless forklifts are already being used in factories. Over the past 30 years there have been 33 employee deaths caused by robots in the US, it notes.

Limits, it says, will have to be imposed on some aspects of machine autonomy. The study adopts the military principle, endorsed by the Ministry of Defence, that there must always be a human in the loop to prevent the development and deployment of entirely autonomous drones that could be programmed to select their own targets.

A no-go area in the science of AI is research into intelligent weapon systems that open fire without a human decision having been made, the report states. The consequences of malfunctions of such machines are immense, so it is all the more desirable that not only the US, but also the United Nations discusses a ban on autonomous weapon systems.

The term artificial intelligence (AI) was first coined by the American computer scientist John McCarthy in 1955. He believed that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. Software developers are still attempting to achieve his goal.

The gap between economic reality in the self-employed gig economy and existing legal frameworks is already growing, the lawyers note. The new information economy is likely to result in more monopolies and a greater income gap between rich and poor because many people will end up unemployed, whereas highly qualified, creative and ambitious professionals will increase their wealth.

Among the professions deemed most likely to disappear are accountants, court clerks and desk officers at fiscal authorities.

Even some lawyers risk becoming unemployed. An intelligent algorithm went through the European Court of Human Rights decisions and found patterns in the text, the report records. Having learned from these cases, the algorithm was able to predict the outcome of other cases with 79% accuracy … According to a study conducted by [the auditing firm] Deloitte, 100,000 jobs in the English legal sector will be automated in the next 20 years.

The pioneering nation in respect of robot density in the industrial sector is South Korea, which has 437 robots for every 10,000 employees in the processing industry, while Japan has 323 and Germany 282.

Robots may soon invade our home and leisure environments. In the Henn-na Hotel in Sasebo, Japan, actroids robots with a human likeness are deployed, the report says. In addition to receiving and serving the guests, they are responsible for cleaning the rooms, carrying the luggage and, since 2016, preparing the food.

The robots are able to respond to the needs of the guests in three languages. The hotels plan is to replace up to 90% of the employees by using robots in hotel operations with a few human employees monitoring CCTV cameras to see whether they need to intervene if problems arise.

The traditional workplace is disintegrating, with more part time employees, distance working, and the blurring of professional and private time, the report observes. It is being replaced by the latte macchiato workplace where employees or freelance workers in the cafe around the corner, working from their laptops.

The workplace may eventually only serve the purpose of maintaining social network between colleagues.

Read more: https://www.theguardian.com/technology/2017/apr/04/innovation-in-ai-could-see-governments-introduce-human-quotas-study-says

Facebook looks inward for new AI technical talent

The race is on to attract as much expertise in artificial intelligence as possible at tech companies large and small, and more than a few Silicon Valley giants are looking inward to convert tech talent they already possess into the AI resources they increasingly need. Facebook has its own AI course, which is oversubscribed, according to a new report by Wired, and which is led by one of the leading AI researchers in the world.

Facebooks Larry Zitnick, who is a key leader at the social networking companys Artificial Intelligence Research Lab, as well as a Microsoft Research and CMU Robotics alum, teaches a class on deep learning for Facebook employees that draws over-capacity crowds. Zitnicks course sparks strong competition among engineers who already rank among the best in the world, each vying to come to grips with and excel at a field outside of their original purview, but one that few fail to recognize is the hottest in tech.

On the other hand, AI and deep learning increasingly touch all aspects of the technology business, so experts with understanding of where the overlap might prove most useful in their own original discipline are also going to be very much in demand. There are external efforts underway to help create more of these polyglot deep learning pros, including at online educational firms like Udacity, but new talent isnt rolling in fast enough from outside sources, traditional and non-traditional alike.

Facebook also offers an AI immersion program, which embeds prospects within the work its doing in the field. The goal, again, is to spread expertise across the company, and thread deep learning know-how into the organizations very DNA. Expect this to be the rule for big tech company behavior for the foreseeable future.

Read more: https://techcrunch.com/2017/03/27/facebook-looks-inward-for-new-ai-technical-talent/

12 ways to hack-proof your smartphone

Protect your privacy, data and peace of mind with this guide to beating thieves, whether theyre online or on the street

As weve recently seen from leaked CIA documents, no one is immune to hacking attacks. Heres how to protect yourself against them, whether they come from opportunist thieves or state-sponsored spies.

1. Keep up to date and dont open up holes yourself

When it comes to protecting yourself against hackers, step one is always to install software updates as soon as they become available: thats as true on smartphones as it is on computers. Yes, updating can be a tiresome and intrusive process, and it sometimes brings annoying changes to the interface that youre used to. All the same, a huge proportion of successfulhacks exploit vulnerabilities that have already been patched; exposing yourself unnecessarily is justdaft.

Id also strongly advise against using unofficial tools to root your phone (known as jailbreaking on iOS), unless you know exactly what youre doing. On a rooted phone, technical safeguards can be defeated, allowing apps to perform all sorts of actions that are normally prohibited and that can include snooping on your personaldata.

2. Be careful of what you install

When you install a smartphone app, you may be asked to grant it various permissions, including the ability to read your files, access your camera or listen in to your microphone. There are legitimate uses for these capabilities, but theyre potentially open to abuse: think before you approve the request. That applies especially to Android users, as Googles app-vetting process isnt as strict as Apples, and there have been reports of malicious apps spending months on the Play Store before being spotted and taken down.

Android also lets you install apps from third-party sources: this allows services such as Amazons competing Appstore to operate, but it also provides an easy way for rogue apps to get onto your phone. Id strongly advise against installing anything from an unfamiliar website.

3. Review whats already on your phone

Even if the apps on your phone seemed simple and safe when you installed them, subsequent updates could have turned them into something more sinister. Take two minutes to review all the apps on your smartphone, and see which permissions theyre using: on iOS, youll find lots of relevant information under Settings > Privacy.

On Android, its harder to get an overview of which apps have which permissions, but there are plenty of security apps that help here, including free packages from Avast and McAfee. These tools can also jump in and alert you if youre trying to install an app thats known to be malicious, and warn you if a phishing attack is trying to trick you into entering a password into an untrusted app orwebpage.

4. Make it hard for intruders to get in

If a thief gets physical access to your phone, they can cause all sorts of trouble. For a start, your email app probably contains a trove of personal information. Make sure your phone is locked when not in use: both Android and iOS can be set to require a six-digit passcode. Your device may offer other options too, like fingerprints or facial recognition. Such methods arent perfect a really determined hacker could copy your fingerprints from a drinking glass, or trick a camera with a photograph of you but theyre a lot better than nothing.

And be wary of smart unlock features, which automatically unlock your phone when youre at home, or when your smartwatch is near; these could let a thief bypass your unlock code altogether.

5. Be prepared to track and lock your phone

Plan ahead, so even if your phone is stolen, you know your data is safe. One option is to set your phone to automatically erase itself after a certain number of incorrect attempts to enter the passcode.

If that seems a bit drastic, dont forget that both Apple and Google operate find my device services that can locate your phone on a map, and remotely lock or erase it. For Apple users, this is accessed through the iCloud website you can check its enabled on the phone in Settings > iCloud > Find My iPhone. Android users can access Googles service at google.co.uk/android/devicemanager. You can also make a missing phone ring helpful for drawing attention to the thief, or tracking down a handsetthatsbeen merely mislaid.

6. Dont leave online services unlocked

Auto-login is a very convenient feature, especially since a virtual keyboard can make typing passwords a chore. Its also a huge liability: an intruder simply needs to open your browser to gain access to all your online accounts.

Ideally, therefore, you shouldnt use auto-login features at all. If you must, use a password manager app that requires you to regularly re-enter a master password. And dont use the same password for more than one app or service: if that one password gets found out, it can be used to access a whole range of private information. This applies even if youre perfectly scrupulous about keeping your smartphone secure: hackers regularly break into online services to steal user credentials, which they then try out on other sites.

7. Adopt an alter ego

If youve followed this advice so far, it should be very difficult for anyone to get into your phone. However, some major hacks have been pulled off without any access to the victim at all. If someone can find out (for example) your date of birth, home town and mothers maiden name all stuff that can be easily picked up from a site like Facebook thats often all they need to reset a password and start breaking into your accounts. You can see off such attacks by fictionalising your past with details that are unlikely to be guessed; perhaps, for the purposes of security, you were born in 1999 to MrsVictoriaBeckham, ne Adams.Just remember what you claimed, or you could end up locking yourself out.

Personal
Personal information can easily be gleaned from sites such as Facebook.

8. Beware open wifi

We all know theres a risk involved in using an open wireless network. But you may not realise how severe it is: anyone in the vicinity can snoop on what youre doing online. This sort of attack demands specialist software and skills, so its unlikely to be a hazard in your local cafe, but its not a danger that can be ignored.

If youre at all doubtful about a wireless network, dont connect stick with your phones mobile internet connection. Or use a VPN tool such as CyberGhost or TunnelBear (both available free for Android and iOS). These tools route your traffic through a private encrypted channel, so even if someone is monitoring your traffic they wont be able to see what youre up to.

9. Dont let lockscreen notifications give the game away

Lots of apps pop up messages and notifications on your phones lockscreen. Its worth thinking about what these notifications may reveal. If you work for a big banking company, for example, a visible email from a work colleague or a meeting remindertells a thief that this might be a particularly interesting phone tosteal.

On iOS, also consider disabling access to Siri from the lockscreen. Siri isnt supposed to give away personal information before you enter your passcode to unlock the iPhone, but past hacks have let intruders use Siri to unlock the device, access details of contacts and view photos. Its safest to shut the feature off entirely: youll find the option under Settings > Touch ID & Passcode > Disable Siri on theLockscreen.

10. Lock individual apps

A strong passcode helps keep thieves out of your phone, but what if a stranger snatches your phone while youre using it? Or asks to borrow it to check a website, then bolts off down the street?

On Android, as a second line of defence, you can lock individual apps, so even if someone can get past your lockscreen, they cant open your email or banking app without a second password. This capability isnt built into the OS, but there are plenty of free apps that provide it, such as AVG AntiVirus Free. iOS users cant directly lock individual apps, but check out Folder Lock free on the App Store which can password-protect your documents and folders, reducing the amount of information a thief canaccess.

11. Get a warning when your phone goes walkies

If youre on the fence about investing in a smartwatch, heres a little-known feature that could swing it: Apple Watch and Android Wear devices can warn you immediately if they lose Bluetooth contact with your phone. If you get this notification while youre in a public place, theres a good chance someones just picked your pocket, and is currently making off with yourphone.

The device will normally be less than 50 metres away when the connection drops, so the warning gives you a chance to ring the phone right away, hopefully drawing attention to the thief and prompting them to jettison it. Failing that, you can lock it before the culprit has a chance to starttryingtobreak in and steal yourdata.

12. Keep an eye on things behind the scenes

No matter how cautious you are, you cant completely eradicate the danger of your phone being hacked not unless you refuse to install any apps or visit any websites. What you can do is supplement your on-device security measures with an online service. LogDog available for both Android and iOS is an app that monitors your identity on sites such as Gmail, Dropbox and Facebook. It alerts you to suspicious activity, such as logins from unfamiliar places, giving you a chance to step in and change your credentials before serious harm can be done. As a bonus, LogDog will also scan your email and highlight messages containing sensitive data such as credit card details and passwords, which you can then purge to ensure they dont fall into the wrong hands.

Read more: https://www.theguardian.com/technology/2017/mar/26/12-ways-to-hack-proof-your-smartphone-privacy-data-thieves