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

Banking with your personal assistant will be the future, whether you like it or not

Image: ambar del moral/mashable

It appears that asking your phone assistant for your bank account will soon be a reality.

Samsung recently partnered with three of South Korea’s largest banks to make voice command banking a reality for South Korean users.

Bixby which is shipped with the Samsung Galaxy S8 can make financial transactions by voice command via Samsung Pay and Samsung Pass, Samsung’s answer to Touch ID.

It’s a concept that’s similar to the partnership Amazon made with Virginia-based Capital One where Capital One customers could be able to check their account balances, recent transactions, and even pay bills just by speaking to Alexa.

The prevalence of personal and home assistants mean that banks can find a cheaper and better option by piggy-backing on Alexa, Siri or Bixby, since the technology to support voice-activated banking is already built into smartphones.

“It is in the bank’s interests to push the service as it helps cut back on manpower and costs,” Chia Tek Yew, head of Financial Services Advisory at KPMG Singapore, told Mashable.

“And at home, it will be a convenient channel of choice given the prevalence of home assistants Amazon Echo, Google Home and Apple HomePod,” Chia added. “[People] will be drawn to shop and transact via voice.”

A customer sets the Touch ID of his new iPhone 7 at an Apple Store in Madrid, Spain.

Image: Getty Images

Regulators are also unlikely to push back regarding voice-activated banking, said Michael Yeo, a Singapore-based research manager at marketing research firm IDC.

“We will continue to witness developments which should push the edge on what can be done with voice,” Yeo told Mashable. “We have already seen examples in China and Singapore where voice commands can issue payments.”

Voice-activated banking hasn’t been explicitly barred, said Chia, adding that in some countries, banks have already used voice assistants to perform banking transactions. Spain-based Santander, for one, is offering voice command banking in their mobile app.

Your face or your thumbprint will be part of this

If recent developments are of any clue, biometric scanning could be a major part of voice-activated banking.

Since Apple’s Touch ID made fingerprint scanning a must-have with smartphones, biometric authentication has been a mainstay. It’s likely that iris scanning will take off, with the iPhone 8 rumoured to have a front-facing 3D camera system that could unlock the phone based on detailed iris or facial recognition, similar to the S8’s iris scanner.

Image: lili sams/mashable

Bixby’s integration with South Korean banks could mean that South Korean users could say good-bye to their clunky banking dongles, for one.

That’s definitely more secure than Amazon’s integration with Capital One, where users are only asked to add a PIN number.

“Iris scanners and fingerprint recognition technology are generally beyond many existing methods, such as passwords, which are used by the majority of banks out there,” Yeo said.

Antony Eldridge, a financial services and fintech leader at PwC Singapore, said that iris and fingerprint scanners are more secure as they are “much more difficult to fake.”

“These appear more acceptable to both [the] public and authorities,” Eldridge said.

Still, most banking security standards written by regulatory bodies like the European Banking Authority and the FFIEC have required multi-factor authentication for bank transactions.

Multi-factor authentication depends on more than one of three things what you are, like your fingerprint or iris pattern; what you have, like a physical token, an ATM card or your phone; and what you know, like a password.

So while banks could integrate biometrics into a two-factor authentication framework, the whole aspect of having another method like your banking dongle or password isn’t likely to change, said Yeo.

“[The] whole essence of two-factor authentication [is that it] requires confirmation by other means,” he added.

But if your identity is tied to your phone like in Samsung Pass, where each device can only be programmed to recognise one set of irises it could still be, in the eyes of regulators, considered two-factor identification, said Chia.

While the industry has made moves to integrate banking with personal and home assistants, updates are coming slowly. Amazon’s partnership with Capital One came in March 2016, and there has yet to be anything like it since.

Whether voice-activated banking will take off would depend on how ubiquitous personal assistants are. It’s likely to remain an option, but just that, as some will never use this feature at all, Yeo said.

“Many still find it awkward to use Siri or Google Assistant on their smartphones,” Yeo said.

Read more: http://mashable.com/2017/06/10/personal-assistant-banking-now/

Microsoft brings Cortana to the Android lock screen

Following a series of beta trials, Microsoft officially announced this morning that its bringing its virtual assistant Cortana to the lock screen of Android devices. Initially, Microsoft had tested putting an overview of your day on the lock screen offering information like the days weather, commute times, flight or meeting information, and more where it could be seen at a glance.

The feature was designed to compete with Googles own Assistant technology, which also offers similar information.

During the beta, users could head into the Cortanas Setting menu to switch on lock screen integration. This would place a floating Cortana circle logo with Swipe to Open on the screen. If you swiped it, youd be shown your personalized feed, schedule and the other information Cortana offered.

The Cortana lock screen integration was rolled out earlier this month, Microsoft confirms, but itsnow widely available.

Today, youre also able to interact with Cortana above the lock, the company says. That means you can ask questions to their virtual assistant, as you could with Siri or Google Assistant, for example, as well as perform various tasks like setting reminders.

This functionality couldfurther challenge Google on its own platform. As Googles only entry in the PC space is Chrome OS, it may be more useful for Windows users to add Cortana to their phones, given its interoperability across desktop, laptop, and mobile including iOS and Android, not just Windows 10.

Microsoft says it has also improved the Cortana home on Android, making theinformation easier to read,along with other improvements like the ability to quickly add reminders and new calendar items, plus easier access to viewing, editing and adding items to lists.

Cortana in Decemberrolled outan updated design on iOS and Android, Microsoft reminded users in todays announcement. This upgrade had been focused on making it easier to get things done by putting the most important tasks more towards the center of the experience. The default experience is very purple, instead of the blue-and-black of the earlier Cortana look and feel. However, Microsoft added more color options in the weeks after the upgrade, including blue, green, and black.

In addition to its U.S. and U.K. availability, Cortana for iOS and Android is also launching in Australia today.

Read more: https://techcrunch.com/2017/03/27/microsoft-brings-cortana-to-the-android-lock-screen/