Linguists, update your resumes because Baidu thinks it has cracked fast AI translation

AI can translate between languages in real time as people speak, according to fresh research from Chinese search giant Baidu and Oregon State University in the US.

Human interpreters need superhuman concentration to listen to speech and translate at the same time. There are, apparently, only a few thousand qualified simultaneous interpreters and the job is so taxing that they often work in pairs, swapping places after 20 to 30 minute stints. And as conversations progress, the chance for error increases exponentially.

Machines have the potential to trump humans at this task, considering they have superior memory and don’t suffer from fatigue. But it’s not so easy for them either, as researchers from Baidu and Oregon State University found.

They built a neural network that can translate between Mandarin Chinese to English in almost real time, where the English translation lags behind by up to at least five words. The results have been published in a paper on arXiv.

The babble post-Babel

Languages have different grammatical structures, where the word order of sentences often don’t match up, making it difficult to translate quickly. The key to a fast translation is predicting what the speaker will say next as he or she talks.

With the AI engine an encoder converts the words in a target language into a vector representation. A decoder predicts the probability of the next word given the words in the previous sentences. The decoder is always behind the encoder and generates the translated words until it processes the whole speech or text.

“In one of the examples, the Chinese sentence ‘Bush President in Moscow…’ would suggest the next English word after ‘President Bush’ is likely ‘meets’”, Liang Huang, principal scientist at Baidu Research, explained to The Register.

“This is possible because in the training data, we have a lot of “Bush meeting someone, like Putin in Moscow” so the system learned that if “Bush in Moscow”, he is likely “meeting” someone.

You can also listen to other examples here.

The problem with languages

The difficulty depends on the languages being translated, Huang added. Languages that are closely related, such as French and Spanish for example, have similar structures where the order of words are aligned more.

Japanese and German sentences are constructed with the subject at the front, the object in the middle, and the verb at the end (SOV). English and Chinese also starts with the subject, but the verb is in the middle, followed by the object (SVO).

Translating between Japanese and German to English and Chinese, therefore, more difficult. “There is a well-known joke in the UN that a German-to-English interpreter often has to pause and “wait for the German verb”. Standard Arabic and Welsh are verb-subject-object , which is even more different from SVO,” he said.

The new algorithm can be applied to any neural machine translation models and only involves tweaking the code slightly. It has already been integrated to Baidu’s internal speech-to-text translation and will be showcased at the Baidu World Tech Conference next week on 1st November in Beijing.

“We don’t have an exact timeline for when this product will be available for the general public, this is certainly something Baidu is working on,” Liang said.

“We envision our technology making simultaneous translation much more accessible and affordable, as there is an increasing demand. We also envision the technology [will reduce] the burden on human translators.”

Source: Linguists, update your resumes because Baidu thinks it has cracked fast AI translation • The Register

Aptoide, alternative app store: EU National Court Rules Against Google in Anti-Trust Process, it has to be shown in the Play Store and can’t be removed by Google

The Portuguese Courts issued today a decision against Google in relation to the injunction filed by Aptoide. It is applicable on 82 countries including UK, Germany, USA, India, among others. Google will have to stop Google Play Protect from removing the competitor Aptoides app store from users phone without users knowledge which has caused losses of over 2.2 million users in the last 60 days.

The acceptance of the injunction is totally aligned with Aptoide’s claim for Google to stop hiding the app store in the Android devices and showing warning messages to the users.

Aptoide is now working alongside its legal team to next week fill in courts the main action, demanding from Google indemnity for all the damages caused.

Aptoide, with over 250 million users, 6 billion downloads and one of the top stores globally, has presented this July, a formal complaint to the European Union’s anti-trust departments against Google.

Paulo Trezentos, Aptoide’s CEO, says that, “For us, this is a decisive victory. Google has been a fierce competitor, abusing his dominant position in Android to eliminate App Store competitors. Innovation is the reason for our 200 million users base. This court’s decision is a signal for startups worldwide: if you have the reason on your side don’t fear to challenge Google.”

About Aptoide

Founded in 2011 and based in Lisbon with offices in Shenzhen and Singapore, Aptoide is one of the top three Android app stores in the world. With over 200 million users, 4 billion downloads and 1 million apps, Aptoide is an app store that reinvents the app discovery experience through an online community, tailored recommendations and the opportunityfor users to create and share their own personal app stores. The Aptoide App Store is available for mobile and TV android devices and is accessible in over 40 languages. With an ever-growing community of users and partners worldwide, Aptoide is now one of the leading players in the world of Apps.

Source: Aptoide: EU National Court Rules Against Google in Anti-Trust Process

Wolf Data, Government Spyware Vendor Left Customer, Victim Data Online for Everyone to See

A startup that claims to sell surveillance and hacking technologies to governments around the world left nearly all its data—including information taken from infected targets and victims—exposed online, according to a security firm who found the data.

Wolf Intelligence, a Germany-based spyware company that made headlines for sending a bodyguard to Mauritania and prompting an international incident after the local government detained the bodyguard as collateral for a deal went wrong, left a trove of its own data exposed online. The leak exposed 20 gigabytes of data, including recordings of meetings with customers, a scan of a passport belonging to the company’s founder, scans of the founder’s credit cards, and surveillance targets’ data, according to researchers.

Security researchers from CSIS Security discovered the data on an unprotected command and control server and a public Google Drive folder. The researchers showed screenshots of the leaked data during a talk at the Virus Bulletin conference in Montreal, which Motherboard attended.

“This is a very stupid story in the sense that you would think that a company actually selling surveillance tools like this would know more about operational security,” CSIS co-founder Peter Kruse told Motherboard in an interview. “They exposed themselves—literally everything was available publicly on the internet.”

Source: Government Spyware Vendor Left Customer, Victim Data Online for Everyone to See – Motherboard

Apple, Samsung fined in Italy for slowing people’s phones.

In a statement on Wednesday, the Italian competition authority, the Autorità Garante della Concorrenza e del Mercato (AGCM), said both companies had violated consumer protection laws by “inducing customers to install updates on devices that are not able to adequately support them.”

It fined Apple €10m ($11.4m): €5m for slowing down the iPhone 6 with its iOS 10 update, and a further €5m for not providing customers with sufficient information about their devices’ batteries, including how to maintain and replace them. Apple banks millions of dollars an hour in profit.

Samsung was fined €5m for its Android Marshmallow 6.0.1 update which was intended for the Galaxy Note 7 but which lead to the Note 4 malfunctioning due to the upgrade’s demands.

Both companies deny they deliberately set out to slow down older phones, but the Italian authorities were not persuaded and clearly felt it was a case of “built-in obsolescence” – where products are designed to fall apart before they need to in order to drive sales of newer models.

Source: Finally, someone takes a stand against Apple, Samsung for slowing people’s phones. Just a few million dollars, tho • The Register

Oxford study claims data harvesting among Android apps is “out of control”

It’s no secret that mobile apps harvest user data and share it with other companies, but the true extent of this practice may come as a surprise. In a new study carried out by researchers from Oxford University, it’s revealed that almost 90 percent of free apps on the Google Play store share data with Alphabet.

The researchers, who analyzed 959,000 apps from the US and UK Google Play stores, said data harvesting and sharing by mobile apps was now “out of control.”

“We find that most apps contain third party tracking, and the distribution of trackers is long-tailed with several highly dominant trackers accounting for a large portion of the coverage,” reads the report.

It’s revealed that most of the apps, 88.4 percent, could share data with companies owned by Google parent Alphabet. Next came a firm that’s no stranger to data sharing controversies, Facebook (42.5 percent), followed by Twitter (33.8 percent), Verizon (26.27 percent), Microsoft (22.75 percent), and Amazon (17.91 percent).

According to The Financial Times, which first reported the research, information shared by these third-party apps can include age, gender, location, and information about a user’s other installed apps. The data “enables construction of detailed profiles about individuals, which could include inferences about shopping habits, socio-economic class or likely political opinions.”

Big firms then use the data for a variety of purposes, such as credit scoring and for targeting political messages, but its main use is often ad targeting. Not surprising, given that revenue from online advertising is now over $59 billion per year.

According to the research, the average app transfers data to five tracker companies, which pass the data on to larger firms. The biggest culprits are news apps and those aimed at children, both of which tend to have the most third-party trackers associated with them.

Source: New study claims data harvesting among Android apps is “out of control” – TechSpot

How A Massive Ad Fraud Scheme Exploited Android Phones To Steal Millions Of Dollars

Last April, Steven Schoen received an email from someone named Natalie Andrea who said she worked for a company called We Purchase Apps. She wanted to buy his Android app, Emoji Switcher. But right away, something seemed off.

“I did a little bit of digging because I was a little sketched out because I couldn’t really find even that the company existed,” Schoen told BuzzFeed News.

The We Purchase Apps website listed a location in New York, but the address appeared to be a residence. “And their phone number was British. It was just all over the place,” Schoen said.

It was all a bit weird, but nothing indicated he was about to see his app end up in the hands of an organization responsible for potentially hundreds of millions of dollars in ad fraud, and which has funneled money to a cabal of shell companies and people scattered across Israel, Serbia, Germany, Bulgaria, Malta, and elsewhere.

Schoen had a Skype call with Andrea and her colleague, who said his name was Zac Ezra, but whose full name is Tzachi Ezrati. They agreed on a price and to pay Schoen up front in bitcoin.

“I would say it was more than I had expected,” Schoen said of the price. That helped convince him to sell.

A similar scenario played out for five other app developers who told BuzzFeed News they sold their apps to We Purchase Apps or directly to Ezrati. (Ezrati told BuzzFeed News he was only hired to buy apps and had no idea what happened to them after they were acquired.)

“A significant portion of the millions of Android phone owners who downloaded these apps were secretly tracked as they scrolled and clicked inside the application.”
The Google Play store pages for these apps were soon changed to list four different companies as their developers, with addresses in Bulgaria, Cyprus, and Russia, giving the appearance that the apps now had different owners.

But an investigation by BuzzFeed News reveals that these seemingly separate apps and companies are today part of a massive, sophisticated digital advertising fraud scheme involving more than 125 Android apps and websites connected to a network of front and shell companies in Cyprus, Malta, British Virgin Islands, Croatia, Bulgaria, and elsewhere. More than a dozen of the affected apps are targeted at kids or teens, and a person involved in the scheme estimates it has stolen hundreds of millions of dollars from brands whose ads were shown to bots instead of actual humans. (A full list of the apps, the websites, and their associated companies connected to the scheme can be found in this spreadsheet.)

One way the fraudsters find apps for their scheme is to acquire legitimate apps through We Purchase Apps and transfer them to shell companies. They then capture the behavior of the app’s human users and program a vast network of bots to mimic it, according to analysis from Protected Media, a cybersecurity and fraud detection firm that analyzed the apps and websites at BuzzFeed News’ request.

This means a significant portion of the millions of Android phone owners who downloaded these apps were secretly tracked as they scrolled and clicked inside the application. By copying actual user behavior in the apps, the fraudsters were able to generate fake traffic that bypassed major fraud detection systems.

“This is not your run-of-the-mill fraud scheme,” said Asaf Greiner, the CEO of Protected Media. “We are impressed with the complex methods that were used to build this fraud scheme and what’s equally as impressive is the ability of criminals to remain under the radar.”

Another fraud detection firm, Pixalate, first exposed one element of the scheme in June. At the time, it estimated that the fraud being committed by a single mobile app could generate $75 million a year in stolen ad revenue. After publishing its findings, Pixalate received an email from an anonymous person connected to the scheme who said the amount that’s been stolen was closer to 10 times that amount. The person also said the operation was so effective because it works “with the biggest partners [in digital advertising] to ensure the ongoing flow of advertisers and money.”

In total, the apps identified by BuzzFeed News have been installed on Android phones more than 115 million times, according to data from analytics service AppBrain. Most are games, but others include a flashlight app, a selfie app, and a healthy eating app. One app connected to the scheme, EverythingMe, has been installed more than 20 million times.

Once acquired, the apps continue to be maintained in order to keep real users happy and create the appearance of a thriving audience that serves as a cover for the cloned fake traffic. The apps are also spread among multiple shell companies to distribute earnings and conceal the size of the operation.

Source: How A Massive Ad Fraud Scheme Exploited Android Phones To Steal Millions Of Dollars