Smarter bots are coming to Facebook, Google and Amazon assistants

We keep hearing that robots are going to take our jobs, but a company called MindMeld is giving us an idea how with its “Deep-Domain Conversational AI Platform.” It’ll allow bots that can essentially replace customer service agents and even baristas by answering complicated voice or text queries over Google Assistant, Amazon Echo, Facebook Messenger and other popular platforms. Uniqlo, for one, will offer a Facebook bot that can answer questions about its products, services and retail locations with more detail than ever.

The “MindMeld Language Parser” helps it understand and answer questions with “human-like accuracy,” whether by talk or text, the company says. It adds that while current AI toolkits from Google, Facebook and others are great, they provide “little data suitable for production-quality applications” — in other words, they can do deep learning, but don’t give companies anything to teach them. It reckons that customers have been “disappointed and underwhelmed” with first-gen bots — something Facebook pretty much said not long ago.

In a video (below) the company showed some (simulated) scenarios of the bot in action. It understands relatively complicated searches, like a coffee order for a “venti, hot Pike Place Roast with room,” (I have no idea myself what that actually is). The bot can then process the order and give you a pickup time and price, or else auto-suggest more options via buttons at the bottom of the screen.

Other features include question responses across “any dynamic knowledge graph,” dialogue management and state tracking, large-scale training data generation and cloud-based or on-premise deployment. Basically, all of those buzzwords mean that companies can put together a bot without needing to do a lot of extra work (for a price, of course).

The company thinks bots will be able to do tasks like “placing a take-out order at a local restaurant, booking a flight or hotel reservation, creating a service appointment at an auto repair shop or doctor’s office, or finding retail store and production information.” It’ll eventually work across “Amazon Alexa, Facebook Messenger, Google Assistant, Skype, Slack, Google Home and more,” says MindMeld CEO Tim Tuttle.

The platform is available starting today, and MindMeld customers and investors already include Google, Samsung, Intel, Uniqlo, Spotify, Telefonica and others. Suffice to say, if you work in customer service or other areas, you may want to start buffing up your resume.

Facebook’s fake news activity spiked near election day

Mark Zuckerberg said it’s unlikely that the small amount of fake stories circulating on Facebook affected the outcome of the presidential elections. According to Buzzfeed News’ investigation, though, that “small amount” of hoaxes outperformed (Google Doc) stories from legit sources on the social network during the final three months of the campaign period. The top 20 fake stories from August to election day apparently generated 8,711,000 shares, reactions, and comments on Facebook, whereas the top 20 from 19 mainstream news websites only generated 7,367,000.

Before that period, mainstream sources generated up to four times the engagement fake sources did. It’s not exactly clear what caused the shift, but it’s worth noting that in the end of June this year, Facebook updated its News Feed algorithm to prioritize the updates your friends share over posts from other sources.

Buzzfeed says 17 of the 20 top-performing hoaxes were overtly pro-Trump or anti-Clinton. That supports earlier reports that Facebook didn’t address its fake news problem, because it would’ve removed a lot of conservative content from people’s Feeds. Three of the top five hoaxes, for instance, claim that the Pope endorsed the President-Elect, that Hillary Clinton sold weapons to ISIS and that the FBI agent suspected of leaking Clinton’s emails was found dead.

There were a handful of fake anti-Trump stories that made the rounds, including one that claims Mike Pence called Michelle Obama the “most vulgar first lady we’ve ever had.” But Buzzfeed found that most of the fake content in the list ultimately came from hoax websites or from hyperpartisan right-wing publishers. Some of those publishers even registered their domains merely a few months before the elections.

[Image credit: Buzzfeed]

As Buzzfeed noted, though, Facebook likes, shares and comments don’t necessarily translate into traffic. A spokesperson for the social network also told the publication that these 20 fake stories are but a small fraction of everything that’s shared on the website. It doesn’t necessarily represent the Facebook activity of all the election-related posts, whether fake or real, shared on the social network. “There is a long tail of stories on Facebook,” he said. “It may seem like the top stories get a lot of traction, but they represent a tiny fraction of the total.” Still, there are dozens of employees within the company who feel like Facebook didn’t do enough to prevent fake news from spreading. They’re now reportedly investigating the situation from within.

Source: Buzzfeed News

Facebook Halts Collection of WhatsApp User Data Across Europe

Facebook has temporarily halted the collection of WhatsApp user data across Europe after pressure from privacy watchdogs, according to the Financial Times. The move follows news last week that the social media company, which owns the popular chat platform, agreed to pause data collection in the U.K. while a government probe there inquired into WhatsApp’s privacy policy.

The Irish Data Protection Commissioner’s office, which acts as Facebook’s European regulatory body, confirmed to FT that the company had suspended European data collection since last week.

The change in data gathering means that WhatsApp users across Europe will no longer have select information shared with Facebook, including the phone number a user verifies during the registration process and the last time a user accessed the service.

Facebook acquired WhatsApp in 2014, but only began collecting data from users of the chat service in August this year. In September, German authorities ordered a halt to the practice in the country after ruling that it was an infringement of national data protection law.

Last month, pan-European privacy watchdogs signed an open letter to WhatsApp saying they had serious concerns over the privacy policy change, and urged Facebook to pause data collection until appropriate legal protections could be assured.

Tags: Facebook, WhatsApp
Discuss this article in our forums

Facebook finds more exaggerated ad data

After Facebook realized that it had been overstating video views for years, it conducted an internal review to search for more flaws messing with its ad data… and it’s not happy with what it found. The social network reports that it found multiple problems with how it calculated or represented the info that marketers thrive on. It wasn’t always counting end-to-end video playback properly, for example, since clip lengths would occasionally change when you started streaming. Facebook also over-reported how long people spent reading Instant Articles, and included more clicks and views than it should in some dashboards.

The company is promising to not only fix those mistakes, but improve its transparency in the future. It’s boosting the number of third-party reviews of its data, creating more measurement tools and promising both clearer and more frequent discussions of its stats. In short: advertisers should spend less time scratching their heads over the accuracy of data and more time using that data.

The flaws are relatively few (just four of the 220 metrics Facebook tracks), and they weren’t there for as long as the video view glitch that started it all. In one case, a dashboard bug had only been around since May. However, it’s not hard to see why Facebook is scrambling to come clean and show that it’s interested in accuracy going forward. Facebook’s massive profits revolve around ads — if companies don’t feel they can trust its metrics, they may scale back their ad campaigns or even jump ship.

Via: Wall Street Journal

Source: Facebook Newsroom

This is how the world looks on Facebook’s population maps

Facebook’s Connectivity Lab today released its high-resolution population maps for Malawi, South Africa, Ghana, Haiti and Sri Lanka, with the promise to make more datasets available over the coming months. The population maps are a joint effort between the Facebook Connectivity Lab, Columbia University and the World Bank, though Facebook is interested in the project as part of its effort to launch wireless communication services in rural regions around the globe.

Facebook and friends used software to identify buildings in commercially available satellite images, and then estimated population using census data and a few other surveys and programs. Convolutional neural networks powered a model capable of identifying individual buildings in images from across the world.

“There has been a lot of work recently on neural networks that can recognize individual buildings with very high accuracy, but these models are finely tuned on the local characteristics of the region where they are trained,” the Connectivity Lab’s Tobias Tiecke writes. “We found that these models do not perform well at a global scale with realistic amounts of training data. Therefore, we developed a model that is trained only on binary labeled data — if an image does or does not contain a building — which allows for fast acquisition of the training data.”

Facebook’s preliminary findings suggest that of the 23 countries it studied, 99 percent of the population lives within 63 km of the nearest city, with 44 percent of people living in cities.

“Hence, if we are able to develop communication technologies that can bridge 63 km with sufficiently high data rates, we should be able to connect 99 percent of the population in these 23 countries,” Tiecke says.

Facebook has been toying with a handful of options for rural communication technologies, including the shoebox-size OpenCellular platform and a satellite initiative. The population maps will help Facebook streamline its designs.

“We would love to hear ideas (preferably also in an open manner) on how to bridge this distance in an economically viable fashion for line of sight and non-line of sight use cases, and we hope that by sharing these datasets we will stimulate further research,” Tiecke says.

Source: CIESIN, Facebook Connectivity Lab

Trump’s Facebook focus helped him win the election

Despite Facebook’s arguments to the contrary, its social network helped Donald Trump win the US election… if not for the reasons you’d suspect. Wired has learned that the Trump campaign focused very heavily on its Facebook strategy, eclipsing a Clinton campaign that (while definitely digital-savvy) still devoted a lot of its attention to TV. Most of Trump’s $250 million in fundraising came from supporters on Facebook, and it was extremely aggressive in running and refining ads. Trump’s digital team ran between 40,000 to 50,000 variants of its promos per day, endlessly figuring out which strategies worked best. They knew if subtitles would help, or if a video was more effective than a still image.

Combine this with Trump’s fondness for Twitter and it’s clear that the campaign was constantly reaching out to the internet, making sure that anyone receptive to its ideas could interact. Contrast that with the relatively insular Clinton campaign. While her team certainly made an effort to influence undecided voters, its unofficial online support included massive private Facebook groups that served as echo chambers. Simply speaking, undecided voters more likely to hear from Trump while skimming their social feeds. This suggests that future American political candidates may have to emphasize Facebook if they want to win — they can’t just treat it as one part of a larger plan.

Source: Wired