Pages Navigation Menu

The blog of DataDiggers

Categories Navigation Menu

Google’s Translatotron converts one spoken language to another, no text involved

Posted by on May 15, 2019 in Artificial Intelligence, Google, machine learning, machine translation, Science, Translation | 0 comments

Every day we creep a little closer to Douglas Adams’ famous and prescient Babel fish. A new research project from Google takes spoken sentences in one language and outputs spoken words in another — but unlike most translation techniques, it uses no intermediate text, working solely with the audio. This makes it quick, but more importantly lets it more easily reflect the cadence and tone of the speaker’s voice.

Translatotron, as the project is called, is the culmination of several years of related work, though it’s still very much an experiment. Google’s researchers, and others, have been looking into the possibility of direct speech-to-speech translation for years, but only recently have those efforts borne fruit worth harvesting.

Translating speech is usually done by breaking down the problem into smaller sequential ones: turning the source speech into text (speech-to-text, or STT), turning text in one language into text in another (machine translation), and then turning the resulting text back into speech (text-to-speech, or TTS). This works quite well, really, but it isn’t perfect; each step has types of errors it is prone to, and these can compound one another.

Furthermore, it’s not really how multilingual people translate in their own heads, as testimony about their own thought processes suggests. How exactly it works is impossible to say with certainty, but few would say that they break down the text and visualize it changing to a new language, then read the new text. Human cognition is frequently a guide for how to advance machine learning algorithms.

Spectrograms of source and translated speech. The translation, let us admit, is not the best. But it sounds better!

To that end, researchers began looking into converting spectrograms, detailed frequency breakdowns of audio, of speech in one language directly to spectrograms in another. This is a very different process from the three-step one, and has its own weaknesses, but it also has advantages.

One is that, while complex, it is essentially a single-step process rather than multi-step, which means, assuming you have enough processing power, Translatotron could work quicker. But more importantly for many, the process makes it easy to retain the character of the source voice, so the translation doesn’t come out robotically, but with the tone and cadence of the original sentence.

Naturally this has a huge impact on expression, and someone who relies on translation or voice synthesis regularly will appreciate that not only what they say comes through, but how they say it. It’s hard to overstate how important this is for regular users of synthetic speech.

The accuracy of the translation, the researchers admit, is not as good as the traditional systems, which have had more time to hone their accuracy. But many of the resulting translations are (at least partially) quite good, and being able to include expression is too great an advantage to pass up. In the end, the team modestly describes their work as a starting point demonstrating the feasibility of the approach, though it’s easy to see that it is also a major step forward in an important domain.

The paper describing the new technique was published on Arxiv, and you can browse samples of speech, from source to traditional translation to Translatotron, at this page. Just be aware that these are not all selected for the quality of their translation, but serve more as examples of how the system retains expression while getting the gist of the meaning.


Source: The Tech Crunch

Read More

Get ready for a new era of personalized entertainment

Posted by on Apr 13, 2019 in Amazon, Artificial Intelligence, Column, computing, Content, Facebook, machine learning, Marketing, Multimedia, personalization, smart devices, Spotify, Streaming Media, streaming services, Twitter, virtual reality, world wide web | 0 comments

New machine learning technologies, user interfaces and automated content creation techniques are going to expand the personalization of storytelling beyond algorithmically generated news feeds and content recommendation.

The next wave will be software-generated narratives that are tailored to the tastes and sentiments of a consumer.

Concretely, it means that your digital footprint, personal preferences and context unlock alternative features in the content itself, be it a news article, live video or a hit series on your streaming service.

The title contains different experiences for different people.

From smart recommendations to smarter content

When you use Youtube, Facebook, Google, Amazon, Twitter, Netflix or Spotify, algorithms select what gets recommended to you. The current mainstream services and their user interfaces and recommendation engines have been optimized to serve you content you might be interested in.

Your data, other people’s data, content-related data and machine learning methods are used to match people and content, thus improving the relevance of content recommendations and efficiency of content distribution.

However, so far the content experience itself has mostly been similar to everyone. If the same news article, live video or TV series episode gets recommended to you and me, we both read and watch the same thing, experiencing the same content.

That’s about to change. Soon we’ll be seeing new forms of smart content, in which user interface, machine learning technologies and content itself are combined in a seamless manner to create a personalized content experience.

What is smart content?

Smart content means that content experience itself is affected by who is seeing, watching, reading or listening to content. The content itself changes based on who you are.

We are already seeing the first forerunners in this space. TikTok’s whole content experience is driven by very short videos, audiovisual content sequences if you will, ordered and woven together by algorithms. Every user sees a different, personalized, “whole” based on her viewing history and user profile.

At the same time, Netflix has recently started testing new forms of interactive content (TV series episodes, e.g. Black Mirror: Bandersnatch) in which user’s own choices affect directly the content experience, including dialogue and storyline. And more is on its way. With Love, Death & Robots series, Netflix is experimenting with episode order within a series, serving the episodes in different order for different users.

Some earlier predecessors of interactive audio-visual content include sports event streaming, in which the user can decide which particular stream she follows and how she interacts with the live content, for example rewinding the stream and spotting the key moments based on her own interest.

Simultaneously, we’re seeing how machine learning technologies can be used to create photo-like images of imaginary people, creatures and places. Current systems can recreate and alter entire videos, for example by changing the style, scenery, lighting, environment or central character’s face. Additionally, AI solutions are able to generate music in different genres.

Now, imagine, that TikTok’s individual short videos would be automatically personalized by the effects chosen by an AI system, and thus the whole video would be customized for you. Or that the choices in the Netflix’s interactive content affecting the plot twists, dialogue and even soundtrack, were made automatically by algorithms based on your profile.

Personalized smart content is coming to news as well. Automated systems, using today’s state-of-the-art NLP technologies, can generate long pieces of concise, comprehensible and even inventive textual content at scale. At present, media houses use automated content creation systems, or “robot journalists”, to create news material varying from complete articles to audio-visual clips and visualizations. Through content atomization (breaking content into small modular chunks of information) and machine learning, content production can be increased massively to support smart content creation.

Say that a news article you read or listen to is about a specific political topic that is unfamiliar to you. When comparing the same article with your friend, your version of the story might use different concepts and offer a different angle than your friend’s who’s really deep into politics. A beginner’s smart content news experience would differ from the experience of a topic enthusiast.

Content itself will become a software-like fluid and personalized experience, where your digital footprint and preferences affect not just how the content is recommended and served to you, but what the content actually contains.

Automated storytelling?

How is it possible to create smart content that contains different experiences for different people?

Content needs to be thought and treated as an iterative and configurable process rather than a ready-made static whole that is finished when it has been published in the distribution pipeline.

Importantly, the core building blocks of the content experience change: smart content consists of atomized modular elements that can be modified, updated, remixed, replaced, omitted and activated based on varying rules. In addition, content modules that have been made in the past, can be reused if applicable. Content is designed and developed more like a software.

Currently a significant amount of human effort and computing resources are used to prepare content for machine-powered content distribution and recommendation systems, varying from smart news apps to on-demand streaming services. With smart content, the content creation and its preparation for publication and distribution channels wouldn’t be separate processes. Instead, metadata and other invisible features that describe and define the content are an integral part of the content creation process from the very beginning.

Turning Donald Glover into Jay Gatsby

With smart content, the narrative or image itself becomes an integral part of an iterative feedback loop, in which the user’s actions, emotions and other signals as well as the visible and invisible features of the content itself affect the whole content consumption cycle from the content creation and recommendation to the content experience. With smart content features, a news article or a movie activates different elements of the content for different people.

It’s very likely that smart content for entertainment purposes will have different features and functions than news media content. Moreover, people expect frictionless and effortless content experience and thus smart content experience differs from games. Smart content doesn’t necessarily require direct actions from the user. If the person wants, the content personalization happens proactively and automatically, without explicit user interaction.

Creating smart content requires both human curation and machine intelligence. Humans focus on things that require creativity and deep analysis while AI systems generate, assemble and iterate the content that becomes dynamic and adaptive just like software.

Sustainable smart content

Smart content has different configurations and representations for different users, user interfaces, devices, languages and environments. The same piece of content contains elements that can be accessed through voice user interface or presented in augmented reality applications. Or the whole content expands into a fully immersive virtual reality experience.

In the same way as with the personalized user interfaces and smart devices, smart content can be used for good and bad. It can be used to enlighten and empower, as well as to trick and mislead. Thus it’s critical, that human-centered approach and sustainable values are built in the very core of smart content creation. Personalization needs to be transparent and the user needs to be able to choose if she wants the content to be personalized or not. And of course, not all content will be smart in the same way, if at all.

If used in a sustainable manner, smart content can break filter bubbles and echo chambers as it can be used to make a wide variety of information more accessible for diverse audiences. Through personalization, challenging topics can be presented to people according to their abilities and preferences, regardless of their background or level of education. For example a beginner’s version of vaccination content or digital media literacy article uses gamification elements, and the more experienced user gets directly a thorough fact-packed account of the recent developments and research results.

Smart content is also aligned with the efforts against today’s information operations such as fake news and its different forms such as “deep fakes” (http://www.niemanlab.org/2018/11/how-the-wall-street-journal-is-preparing-its-journalists-to-detect-deepfakes). If the content is like software, a legit software runs on your devices and interfaces without a problem. On the other hand, even the machine-generated realistic-looking but suspicious content, like deep fake, can be detected and filtered out based on its signature and other machine readable qualities.


Smart content is the ultimate combination of user experience design, AI technologies and storytelling.

News media should be among the first to start experimenting with smart content. When the intelligent content starts eating the world, one should be creating ones own intelligent content.

The first players that master the smart content, will be among tomorrow’s reigning digital giants. And that’s one of the main reasons why today’s tech titans are going seriously into the content game. Smart content is coming.


Source: The Tech Crunch

Read More

Blind users can now explore photos by touch with Microsoft’s Seeing AI

Posted by on Mar 12, 2019 in accessibility, Apps, Artificial Intelligence, Augmented Reality, Blindness, Computer Vision, Disabilities, machine learning, Microsoft, Mobile | 0 comments

Microsoft’s Seeing AI is an app that lets blind and limited-vision folks convert visual data into audio feedback, and it just got a useful new feature. Users can now use touch to explore the objects and people in photos.

It’s powered by machine learning, of course, specifically object and scene recognition. All you need to do is take a photo or open one up in the viewer and tap anywhere on it.

“This new feature enables users to tap their finger to an image on a touch-screen to hear a description of objects within an image and the spatial relationship between them,” wrote Seeing AI lead Saqib Shaikh in a blog post. “The app can even describe the physical appearance of people and predict their mood.”

Because there’s facial recognition built in as well, you could very well take a picture of your friends and hear who’s doing what and where, and whether there’s a dog in the picture (important) and so on. This was possible on an image-wide scale already, as you can see in this image:

But the app now lets users tap around to find where objects are — obviously important to understanding the picture or recognizing it from before. Other details that may not have made it into the overall description may also appear on closer inspection, such as flowers in the foreground or a movie poster in the background.

In addition to this, the app now natively supports the iPad, which is certainly going to be nice for the many people who use Apple’s tablets as their primary interface for media and interactions. Lastly, there are a few improvements to the interface so users can order things in the app to their preference.

Seeing AI is free — you can download it for iOS devices here.


Source: The Tech Crunch

Read More

Superb AI generates customized training data for machine learning projects

Posted by on Feb 25, 2019 in Artificial Intelligence, machine learning, Startups, Superb AI, Y Combinator | 0 comments

One of the big challenges of developing a machine learning project can be simply getting enough relevant data to train the algorithms. That’s where Superb AI, a member of the Y Combinator Winter 2019 class, can help. The startup helps companies create customized data sets to meet the requirements of any project, using AI to speed up the tagging process.

Hyun Kim, who is CEO and co-founder at the startup, says one of the big stumbling blocks for companies trying to incorporate AI and machine learning into their applications is coming up with a set of suitable data to train the models. “Superb AI uses AI to make customized AI training data for large tech companies. Clients work with us to develop machine learning-based features in their products multiple times faster than they could themselves,” Kim told TechCrunch.

Kim and his co-founders (CTO Jung Kwon Lee, machine learning engineers Jonghyuk Lee and Moonsu Cha and Hyundong Lee, head of APAC sales and operations, who is based in Seoul, South Korea) all were working in the field when they identified the data problem and decided to launch a company to solve it.

Traditionally, companies working on a machine learning project will hire human workers to tag data, but this has been expensive and error prone, assuming you even had the data to work with. Kim and his co-founders, who worked on AI projects and studied the subject in college, came up with the idea of putting AI to work on the tagging part of the problem.

“Instead of relying on slow and error-prone manual labor, Superb AI uses proprietary deep learning AI that assists humans to achieve up to 10x faster labeling of images and videos,” Kim explained. The company will also help find data sources for companies that don’t have any data to begin with.

Kim says that they don’t take humans out of the process completely, but they do enhance tagging accuracy by combining human workers with artificial intelligence underpinnings. He says that this involves a couple of steps. First, it splits training data into as many components as possible in order to automate each piece one at a time. If the data is too complex, and the AI tools can’t automate the tagging, they use a second approach called “human in the loop.” As humans label data, the AI can learn over time and eventually take over more and more of the process.

The co-founders decided to apply to Y Combinator to gain a foothold in Silicon Valley, where they could expand their market beyond their native South Korea. “It’s definitely been a game changer. The amount of knowledge and experience we gained from the YC partners and fellow entrepreneurs is really unbelievable. And also the vast YC network helped us find our early customers in the Valley,” Kim said.

The company, which launched last October, is up to 13 employees, including the co-founders. It has raised $300,000 in seed investment and has already generated the same amount in revenue from the product, according to Kim.


Source: The Tech Crunch

Read More

Briq, the next building block in tech’s reconstruction of the construction business, raises $3 million

Posted by on Feb 22, 2019 in blockchains, Briq, California, computing, construction, construction software, eniac, eniac ventures, General Partner, Hyperledger, IBM, machine learning, procore, Real Estate, real-time, TC, Technology, Tim Young | 0 comments

Bassem Hamdy has been in the construction business for a long time.

He spent the last few years at the construction software business Procore, now a $3 billion company developing technology for the construction industry, and now Hamdy is ready to unveil his next act as chief executive and co-founder of Briq, a new software service for the industry.

Hamdy started Briq with his own cash, amassed through secondary sales as Procore climbed the ranks of startups to reach its status as a construction industry unicorn. And the company has just raised $3 million in financing to fund its expansion.

“With enough secondaries you can afford to make your own decisions,” Hamdy says. 

His experience in construction dates back to his earliest days. Hailing from a family of construction engineers, Hamdy describes himself as a black sheep who went into the financial services industry — but construction kept pulling him back.

Beginning in the late nineties with CMIC, which was construction enterprise resource planning, and continuing through to Procore, Hamdy has had success after success in the business, but Briq is the culmination of all of that experience, he says. 

“As much as data entry helps people it’s data intelligence software that changes things,” says Hamdy. 

Briq chief executive Bassem Hamdy

The Santa Barbara, Calif.-based company is part of a growing number of Southern California technology startups building businesses to service large swaths of specific industries — specifically real estate and construction.

Already, Procore is a $3 billion behemoth, and ServiceTitan has become a billion-dollar company as well, with its software and services for air conditioning and appliance repairmen.

Now Hamdy’s Briq, with backing from Eniac Ventures and MetaProp NYC, is hoping to join their ranks.

“Bassem built and helped run the most successful construction software businesses in the world. It is rare and humbling to have an opportunity to help build a company from the ground up with an industry legend,” says Tim Young, founding general partner at Eniac Ventures . “The technology Bassem and his team are building will do something the industry has never seen before: break down data silos to leverage information in real time. Bassem has built and run the most successful construction software businesses in the world, and his knowledge of the construction space and the data space is second to none.”

The company, formerly called Brickschain, uses a combination of a blockchain-based immutable ledger and machine learning tools to provide strategic insights into buildings and project developments.

Briq’s software can predict things like the success of individual projects, where demand for new projects is likely to occur and how to connect data around construction processes.

Briq has two main offerings, according to Hamdy. ProjectIQ, which monitors and manages individual projects and workflows — providing data around different vendors involved in a construction project; and MarketIQ, which provides market intelligence around where potential projects are likely to occur and which projects will be met with the most demand and success.

Joining Hamdy in the creation of Briq is Ron Goldschmidt, an experienced developer of quantitative-based trading strategies for several businesses. Hamdy, a former Wall Streeter himself, has long realized the power of data in the construction business. And with the new tools at his disposal — including the blockchain-based ledger system that forms the backbone of Briq’s project management software, Hamdy thinks he has developed the next big evolution in technology for the industry.

Briq already counts Webcore, a major contractor and developer, as one of its clients, along with Kobayashi, Probuild, Hunter Roberts OEG and Gartner Builders. In all, the company has contracts with nearly 12 developers and contractors.

All of the insights that Briq can provide through its immutable ledger can add up to big savings for developers. Hamdy estimates that there’s roughly $1 trillion in waste in the construction industry.

Briq relies on IBM’s Hyperledger for its blockchain backbone and through that, the company has a window into all of the decisions made on a project. That ledger forms the scaffolding on which Briq can build out its projections and models of how much a building will cost, and how could conceivably be made on a project.

“Construction and infrastructure are integral to society, but the decision-making process behind how, when, where, and why we build is no longer working,” said Hamdy, in a statement. “We aren’t just solving a construction problem, we are solving a societal problem. If we are to meet the infrastructure needs of both the developed and developing world, we must improve our decision-making and analysis around the data we have. We are thrilled to have the support of Eniac Ventures as we enter the next phase of our journey.”


Source: The Tech Crunch

Read More

Fabula AI is using social spread to spot ‘fake news’

Posted by on Feb 6, 2019 in Amazon, api, Artificial Intelligence, deep learning, Emerging-Technologies, Europe, European Research Council, Facebook, fake news, Imperial College London, London, machine learning, Mark Zuckerberg, Media, MIT, Myanmar, Social, Social Media, social media platforms, social media regulation, social network, social networks, Startups, TC, United Kingdom | 0 comments

UK startup Fabula AI reckons it’s devised a way for artificial intelligence to help user generated content platforms get on top of the disinformation crisis that keeps rocking the world of social media with antisocial scandals.

Even Facebook’s Mark Zuckerberg has sounded a cautious note about AI technology’s capability to meet the complex, contextual, messy and inherently human challenge of correctly understanding every missive a social media user might send, well-intentioned or its nasty flip-side.

“It will take many years to fully develop these systems,” the Facebook founder wrote two years ago, in an open letter discussing the scale of the challenge of moderating content on platforms thick with billions of users. “This is technically difficult as it requires building AI that can read and understand news.”

But what if AI doesn’t need to read and understand news in order to detect whether it’s true or false?

Step forward Fabula, which has patented what it dubs a “new class” of machine learning algorithms to detect “fake news” — in the emergent field of “Geometric Deep Learning”; where the datasets to be studied are so large and complex that traditional machine learning techniques struggle to find purchase on this ‘non-Euclidean’ space.

The startup says its deep learning algorithms are, by contrast, capable of learning patterns on complex, distributed data sets like social networks. So it’s billing its technology as a breakthrough. (Its written a paper on the approach which can be downloaded here.)

It is, rather unfortunately, using the populist and now frowned upon badge “fake news” in its PR. But it says it’s intending this fuzzy umbrella to refer to both disinformation and misinformation. Which means maliciously minded and unintentional fakes. Or, to put it another way, a photoshopped fake photo or a genuine image spread in the wrong context.

The approach it’s taking to detecting disinformation relies not on algorithms parsing news content to try to identify malicious nonsense but instead looks at how such stuff spreads on social networks — and also therefore who is spreading it.

There are characteristic patterns to how ‘fake news’ spreads vs the genuine article, says Fabula co-founder and chief scientist, Michael Bronstein.

“We look at the way that the news spreads on the social network. And there is — I would say — a mounting amount of evidence that shows that fake news and real news spread differently,” he tells TechCrunch, pointing to a recent major study by MIT academics which found ‘fake news’ spreads differently vs bona fide content on Twitter.

“The essence of geometric deep learning is it can work with network-structured data. So here we can incorporate heterogenous data such as user characteristics; the social network interactions between users; the spread of the news itself; so many features that otherwise would be impossible to deal with under machine learning techniques,” he continues.

Bronstein, who is also a professor at Imperial College London, with a chair in machine learning and pattern recognition, likens the phenomenon Fabula’s machine learning classifier has learnt to spot to the way infectious disease spreads through a population.

“This is of course a very simplified model of how a disease spreads on the network. In this case network models relations or interactions between people. So in a sense you can think of news in this way,” he suggests. “There is evidence of polarization, there is evidence of confirmation bias. So, basically, there are what is called echo chambers that are formed in a social network that favor these behaviours.”

“We didn’t really go into — let’s say — the sociological or the psychological factors that probably explain why this happens. But there is some research that shows that fake news is akin to epidemics.”

The tl;dr of the MIT study, which examined a decade’s worth of tweets, was that not only does the truth spread slower but also that human beings themselves are implicated in accelerating disinformation. (So, yes, actual human beings are the problem.) Ergo, it’s not all bots doing all the heavy lifting of amplifying junk online.

The silver lining of what appears to be an unfortunate quirk of human nature is that a penchant for spreading nonsense may ultimately help give the stuff away — making a scalable AI-based tool for detecting ‘BS’ potentially not such a crazy pipe-dream.

Although, to be clear, Fabula’s AI remains in development at this stage, having been tested internally on Twitter data sub-sets at this stage. And the claims it’s making for its prototype model remain to be commercially tested with customers in the wild using the tech across different social platforms.

It’s hoping to get there this year, though, and intends to offer an API for platforms and publishers towards the end of this year. The AI classifier is intended to run in near real-time on a social network or other content platform, identifying BS.

Fabula envisages its own role, as the company behind the tech, as that of an open, decentralised “truth-risk scoring platform” — akin to a credit referencing agency just related to content, not cash.

Scoring comes into it because the AI generates a score for classifying content based on how confident it is it’s looking at a piece of fake vs true news.

A visualisation of a fake vs real news distribution pattern; users who predominantly share fake news are coloured red and users who don’t share fake news at all are coloured blue — which Fabula says shows the clear separation into distinct groups, and “the immediately recognisable difference in spread pattern of dissemination”.

In its own tests Fabula says its algorithms were able to identify 93 percent of “fake news” within hours of dissemination — which Bronstein claims is “significantly higher” than any other published method for detecting ‘fake news’. (Their accuracy figure uses a standard aggregate measurement of machine learning classification model performance, called ROC AUC.)

The dataset the team used to train their model is a subset of Twitter’s network — comprised of around 250,000 users and containing around 2.5 million “edges” (aka social connections).

For their training dataset Fabula relied on true/fake labels attached to news stories by third party fact checking NGOs, including Snopes and PolitiFact. And, overall, pulling together the dataset was a process of “many months”, according to Bronstein, He also says that around a thousand different stories were used to train the model, adding that the team is confident the approach works on small social networks, as well as Facebook-sized mega-nets.

Asked whether he’s sure the model hasn’t been trained to identified patterns caused by bot-based junk news spreaders, he says the training dataset included some registered (and thus verified ‘true’) users.

“There is multiple research that shows that bots didn’t play a significant amount [of a role in spreading fake news] because the amount of it was just a few percent. And bots can be quite easily detected,” he also suggests, adding: “Usually it’s based on some connectivity analysis or content analysis. With our methods we can also detect bots easily.”

To further check the model, the team tested its performance over time by training it on historical data and then using a different split of test data.

“While we see some drop in performance it is not dramatic. So the model ages well, basically. Up to something like a year the model can still be applied without any re-training,” he notes, while also saying that, when applied in practice, the model would be continually updated as it keeps digesting (ingesting?) new stories and social media content.

Somewhat terrifyingly, the model could also be used to predict virality, according to Bronstein — raising the dystopian prospect of the API being used for the opposite purpose to that which it’s intended: i.e. maliciously, by fake news purveyors, to further amp up their (anti)social spread.

“Potentially putting it into evil hands it might do harm,” Bronstein concedes. Though he takes a philosophical view on the hyper-powerful double-edged sword of AI technology, arguing such technologies will create an imperative for a rethinking of the news ecosystem by all stakeholders, as well as encouraging emphasis on user education and teaching critical thinking.

Let’s certainly hope so. And, on the educational front, Fabula is hoping its technology can play an important role — by spotlighting network-based cause and effect.

“People now like or retweet or basically spread information without thinking too much or the potential harm or damage they’re doing to everyone,” says Bronstein, pointing again to the infectious diseases analogy. “It’s like not vaccinating yourself or your children. If you think a little bit about what you’re spreading on a social network you might prevent an epidemic.”

So, tl;dr, think before you RT.

Returning to the accuracy rate of Fabula’s model, while ~93 per cent might sound pretty impressive, if it were applied to content on a massive social network like Facebook — which has some 2.3BN+ users, uploading what could be trillions of pieces of content daily — even a seven percent failure rate would still make for an awful lot of fakes slipping undetected through the AI’s net.

But Bronstein says the technology does not have to be used as a standalone moderation system. Rather he suggests it could be used in conjunction with other approaches such as content analysis, and thus function as another string on a wider ‘BS detector’s bow.

It could also, he suggests, further aid human content reviewers — to point them to potentially problematic content more quickly.

Depending on how the technology gets used he says it could do away with the need for independent third party fact-checking organizations altogether because the deep learning system can be adapted to different use cases.

Example use-cases he mentions include an entirely automated filter (i.e. with no human reviewer in the loop); or to power a content credibility ranking system that can down-weight dubious stories or even block them entirely; or for intermediate content screening to flag potential fake news for human attention.

Each of those scenarios would likely entail a different truth-risk confidence score. Though most — if not all — would still require some human back-up. If only to manage overarching ethical and legal considerations related to largely automated decisions. (Europe’s GDPR framework has some requirements on that front, for example.)

Facebook’s grave failures around moderating hate speech in Myanmar — which led to its own platform becoming a megaphone for terrible ethnical violence — were very clearly exacerbated by the fact it did not have enough reviewers who were able to understand (the many) local languages and dialects spoken in the country.

So if Fabula’s language-agnostic propagation and user focused approach proves to be as culturally universal as its makers hope, it might be able to raise flags faster than human brains which lack the necessary language skills and local knowledge to intelligently parse context.

“Of course we can incorporate content features but we don’t have to — we don’t want to,” says Bronstein. “The method can be made language independent. So it doesn’t matter whether the news are written in French, in English, in Italian. It is based on the way the news propagates on the network.”

Although he also concedes: “We have not done any geographic, localized studies.”

“Most of the news that we take are from PolitiFact so they somehow regard mainly the American political life but the Twitter users are global. So not all of them, for example, tweet in English. So we don’t yet take into account tweet content itself or their comments in the tweet — we are looking at the propagation features and the user features,” he continues.

“These will be obviously next steps but we hypothesis that it’s less language dependent. It might be somehow geographically varied. But these will be already second order details that might make the model more accurate. But, overall, currently we are not using any location-specific or geographic targeting for the model.

“But it will be an interesting thing to explore. So this is one of the things we’ll be looking into in the future.”

Fabula’s approach being tied to the spread (and the spreaders) of fake news certainly means there’s a raft of associated ethical considerations that any platform making use of its technology would need to be hyper sensitive to.

For instance, if platforms could suddenly identify and label a sub-set of users as ‘junk spreaders’ the next obvious question is how will they treat such people?

Would they penalize them with limits — or even a total block — on their power to socially share on the platform? And would that be ethical or fair given that not every sharer of fake news is maliciously intending to spread lies?

What if it turns out there’s a link between — let’s say — a lack of education and propensity to spread disinformation? As there can be a link between poverty and education… What then? Aren’t your savvy algorithmic content downweights risking exacerbating existing unfair societal divisions?

Bronstein agrees there are major ethical questions ahead when it comes to how a ‘fake news’ classifier gets used.

“Imagine that we find a strong correlation between the political affiliation of a user and this ‘credibility’ score. So for example we can tell with hyper-ability that if someone is a Trump supporter then he or she will be mainly spreading fake news. Of course such an algorithm would provide great accuracy but at least ethically it might be wrong,” he says when we ask about ethics.

He confirms Fabula is not using any kind of political affiliation information in its model at this point — but it’s all too easy to imagine this sort of classifier being used to surface (and even exploit) such links.

“What is very important in these problems is not only to be right — so it’s great of course that we’re able to quantify fake news with this accuracy of ~90 percent — but it must also be for the right reasons,” he adds.

The London-based startup was founded in April last year, though the academic research underpinning the algorithms has been in train for the past four years, according to Bronstein.

The patent for their method was filed in early 2016 and granted last July.

They’ve been funded by $500,000 in angel funding and about another $500,000 in total of European Research Council grants plus academic grants from tech giants Amazon, Google and Facebook, awarded via open research competition awards.

(Bronstein confirms the three companies have no active involvement in the business. Though doubtless Fabula is hoping to turn them into customers for its API down the line. But he says he can’t discuss any potential discussions it might be having with the platforms about using its tech.)

Focusing on spotting patterns in how content spreads as a detection mechanism does have one major and obvious drawback — in that it only works after the fact of (some) fake content spread. So this approach could never entirely stop disinformation in its tracks.

Though Fabula claims detection is possible within a relatively short time frame — of between two and 20 hours after content has been seeded onto a network.

“What we show is that this spread can be very short,” he says. “We looked at up to 24 hours and we’ve seen that just in a few hours… we can already make an accurate prediction. Basically it increases and slowly saturates. Let’s say after four or five hours we’re already about 90 per cent.”

“We never worked with anything that was lower than hours but we could look,” he continues. “It really depends on the news. Some news does not spread that fast. Even the most groundbreaking news do not spread extremely fast. If you look at the percentage of the spread of the news in the first hours you get maybe just a small fraction. The spreading is usually triggered by some important nodes in the social network. Users with many followers, tweeting or retweeting. So there are some key bottlenecks in the network that make something viral or not.”

A network-based approach to content moderation could also serve to further enhance the power and dominance of already hugely powerful content platforms — by making the networks themselves core to social media regulation, i.e. if pattern-spotting algorithms rely on key network components (such as graph structure) to function.

So you can certainly see why — even above a pressing business need — tech giants are at least interested in backing the academic research. Especially with politicians increasingly calling for online content platforms to be regulated like publishers.

At the same time, there are — what look like — some big potential positives to analyzing spread, rather than content, for content moderation purposes.

As noted above, the approach doesn’t require training the algorithms on different languages and (seemingly) cultural contexts — setting it apart from content-based disinformation detection systems. So if it proves as robust as claimed it should be more scalable.

Though, as Bronstein notes, the team have mostly used U.S. political news for training their initial classifier. So some cultural variations in how people spread and react to nonsense online at least remains a possibility.

A more certain challenge is “interpretability” — aka explaining what underlies the patterns the deep learning technology has identified via the spread of fake news.

While algorithmic accountability is very often a challenge for AI technologies, Bronstein admits it’s “more complicated” for geometric deep learning.

“We can potentially identify some features that are the most characteristic of fake vs true news,” he suggests when asked whether some sort of ‘formula’ of fake news can be traced via the data, noting that while they haven’t yet tried to do this they did observe “some polarization”.

“There are basically two communities in the social network that communicate mainly within the community and rarely across the communities,” he says. “Basically it is less likely that somebody who tweets a fake story will be retweeted by somebody who mostly tweets real stories. There is a manifestation of this polarization. It might be related to these theories of echo chambers and various biases that exist. Again we didn’t dive into trying to explain it from a sociological point of view — but we observed it.”

So while, in recent years, there have been some academic efforts to debunk the notion that social media users are stuck inside filter bubble bouncing their own opinions back at them, Fabula’s analysis of the landscape of social media opinions suggests they do exist — albeit, just not encasing every Internet user.

Bronstein says the next steps for the startup is to scale its prototype to be able to deal with multiple requests so it can get the API to market in 2019 — and start charging publishers for a truth-risk/reliability score for each piece of content they host.

“We’ll probably be providing some restricted access maybe with some commercial partners to test the API but eventually we would like to make it useable by multiple people from different businesses,” says requests. “Potentially also private users — journalists or social media platforms or advertisers. Basically we want to be… a clearing house for news.”


Source: The Tech Crunch

Read More

Play Iconary, a simple drawing game that hides a deceptively deep AI

Posted by on Feb 5, 2019 in AI2, Artificial Intelligence, Computer Vision, machine learning, paul allen, Science, TC | 0 comments

It may not seem like it takes a lot of smarts to play a game like Pictionary, but in fact it involves a lot of subtle and abstract visual and linguistic skills. This AI built to play a game like it is similarly complex, and its interpretations and creations when you play it (as you can now) may seem eerily human — but it’s also refreshing to have such an agent working collaboratively with you rather than beating you with superhuman skills.

Iconary, as the game’s creators at the Allen Institute for AI decided to call it to avoid lawsuits from Mattel, has you drawing and arranging icons to form phrases, or guessing at the arrangements of the computer player.

For instance, if you were to get the phrase “woman drinking milk from a glass,” you’d probably draw a woman — a stick figure, probably, and then select the “woman” icon from the computer’s interpretations of your sketch. Then you’d draw a glass, and place that near the woman. Then… milk? How do you draw milk? There is actually a milk bottle icon if you look for it, but you could also draw a cow and put that in or next to the glass.

The computer then guesses at what you’ve put together, and after a few tries it would probably get it. You can also play it the other way, where the computer arranges icons and you have to guess.

Now, let’s get this right out of the way: This is very different from Google’s superficially similar “Quick, Draw” game. In that one the system can only guess whether your drawing is one of a few hundred pre-selected objects it’s been specifically trained to recognize.

Not only are there some 75,000 phrases supported in Iconary, with more being added regularly, but there’s no way to train the AI on them — the way that any one of them can be represented is uncountable.

“When you start bringing in phrases, the problem space explodes,” explained Ali Farhadi, one of the creators of the project; I talked with him and researcher Aniruddha Kembhavi about Iconary ahead of its release. “Sure, you can easily recognize a cat or a dog. But can you recognize a cat purring, or a dog scratching its back? There’s a huge diversity in elements people choose and how they position them.”

Although Pictionary may seem at first like a game that depends on your drawing skill, it’s really much more about arranging ideas and understanding the relationship with them — seeing the intent behind the drawing. How else can some people manage to recognize a word or phrase from a handful of crude shapes and some arrows?

The AI behind Iconary, then, isn’t a drawing recognition engine at all but one that has been trained to recognize relationships between objects, informed by their type, position, number and everything else. This is, the researchers say, the most significant example of AI collaborating meaningfully with humans yet created.

And this logic is kept fuzzy enough that several “person” icons gathered together could mean women, men, people, group, crowd, team or anything else. How would you know if it was a “team?” Well, if you put a soccer ball near it or put them on a play field, it becomes obvious. If there’s a blackboard there, it’s probably a class. And so on.

Of course, I say “and so on,” but that small phrase in a way encompasses the entirety of human intuition and years of training on how to view and interpret the visual world. Naturally Iconary isn’t nearly as good at it as we are, but its logic is frequently surprisingly human.

If you can only get part of the answer, you can ask the AI to draw again, and just like we do in Pictionary it will adapt its representation to address your needs.

It was of course trained on human drawings collected via Mechanical Turk, but it isn’t just replicating what people drew. If the only thing it ever saw to represent a scientist was a man next to a microscope, how would it know to recognize the same idea in a woman, or standing next to an atom or rocket? In fact, the model has never been exposed to the phrases you can play with now. As the researchers write:

AllenAI has never before encountered the unique phrases in Iconary, yet our preliminary games have shown that our AI system is able to both successfully depict and understand phrases with a human partner with an often surprising deftness and nuance. This feat requires combining natural language understanding, computer vision, and the use of common sense to reason about phrases and their depictions within the constraints of a small vocabulary of possible icons. Being successful at Iconary requires skills beyond basic pattern recognition, including multi-hop reasoning, abstraction, collaboration, and adaptation

Instead of simply pairing “ball” with “sport,” it learned about why those objects are related, and how to exert some basic common sense — a sort of Holy Grail in AI, though this is only a small step in that direction. If one person draws “father” as a man bigger than a smaller person, it isn’t obvious to the computer that the father is the big one, not the small. And it’s another logical jump that a “mother” would be a similarly sized woman, or that the small one is a child.

But by observing how people used the objects and how they relate to one another, the AI built up a network of ideas about how different things are represented or related. “Child” is closer to “student” than “horse,” for instance. And “student” is close to “desk” and “laptop.” So if you draw a child by a desk, maybe it’s a student? This kind of robust logic is so simple to us that we don’t even recognize we’re doing it, but incredibly hard to build into a machine learning agent.

This type of AI is deceptively broad and intelligent, but it isn’t flashy the way that the human-destroying AlphaStar or AlphaGo are. It isn’t superhuman — in fact, it’s not even close to human. But board and PC games are tightly bounded problem spaces with set rules and limits. Visual expression of a complex phrase like “crowd celebrating a victory on a street” isn’t a question of how fast the computer can process, but the depth of its understanding of the concepts involved, and how others think about them.

This kind of learning is also more broadly applicable in the real world. Robots and self-driving cars do need to know how to exceed human capacity in some cases, but it’s also massively important to be able to understand the world around them in the same way people do. When it sees a person by a hospital bed holding a book, what does that mean? When a person leaves a knife out next to a whole tomato? And so on.

“Real life problems involve semantics, abstraction and collaboration,” said Farhadi. “They involve theory of mind.”

Interestingly, the agent is biased a bit (as these things tend to be) owing to the natural bias of our language. Images “read” from left to right, as people tend to draw them, since we also read in that direction, so keep that in mind.

Try playing a couple of games both drawing and guessing, and you may be surprised at the cleverness and weirdness of the AI’s suggestions. Don’t feel bad about skipping one — the agent is still learning, and sometimes its attempts to represent ideas are a bit too abstract. But I certainly found myself impressed more than baffled.

If you’d like to learn more, stay tuned: The team behind the system will be publishing a paper on it later this year. I’ll update this post when that happens.


Source: The Tech Crunch

Read More

Databricks raises $250M at a $2.75B valuation for its analytics platform

Posted by on Feb 5, 2019 in AI, Andreessen Horowitz, Apache Spark, Artificial Intelligence, big data, Cloud, Databricks, Enterprise, funding, machine learning, Recent Funding, Series E, Startups, TC | 0 comments

Databricks, the company behind the Apache Spark big data analytics engine, today announced that it has raised a $250 million Series E round led by Andreessen Horowitz. Coatue Management, Microsoft and NEA, also participated in this round, which brings the company’s total funding to $498.5 million. Microsoft’s involvement here is probably a bit of a surprise, but it’s worth noting that it also worked with Databricks on the launch of Azure Databricks as a first-party service on the platform, something that’s still a rarity in the Azure cloud.

As Databricks also today announced, its annual recurring revenue now exceeds $100 million. The company didn’t share whether it’s cash flow-positive at this point, but Databricks CEO and co-founder Ali Ghodsi shared that the company’s valuation is now $2.75 billion.

Current customers, which the company says number around 2,000, include the likes of Nielsen, Hotels.com, Overstock, Bechtel, Shell and HP.

While Databricks is obviously known for its contributions to Apache Spark, the company itself monetizes that work by offering its Unified Analytics platform on top of it. This platform allows enterprises to build their data pipelines across data storage systems and prepare data sets for data scientists and engineers. To do this, Databricks offers shared notebooks and tools for building, managing and monitoring data pipelines, and then uses that data to build machine learning models, for example. Indeed, training and deploying these models is one of the company’s focus areas these days, which makes sense, given that this is one of the main use cases for big data, after all.

On top of that, Databricks also offers a fully managed service for hosting all of these tools.

“Databricks is the clear winner in the big data platform race,” said Ben Horowitz, co-founder and general partner at Andreessen Horowitz, in today’s announcement. “In addition, they have created a new category atop their world-beating Apache Spark platform called Unified Analytics that is growing even faster. As a result, we are thrilled to invest in this round.”

Ghodsi told me that Horowitz was also instrumental in getting the company to re-focus on growth. The company was already growing fast, of course, but Horowitz asked him why Databricks wasn’t growing faster. Unsurprisingly, given that it’s an enterprise company, that means aggressively hiring a larger sales force — and that’s costly. Hence the company’s need to raise at this point.

As Ghodsi told me, one of the areas the company wants to focus on is the Asia Pacific region, where overall cloud usage is growing fast. The other area the company is focusing on is support for more verticals like mass media and entertainment, federal agencies and fintech firms, which also comes with its own cost, given that the experts there don’t come cheap.

Ghodsi likes to call this “boring AI,” since it’s not as exciting as self-driving cars. In his view, though, the enterprise companies that don’t start using machine learning now will inevitably be left behind in the long run. “If you don’t get there, there’ll be no place for you in the next 20 years,” he said.

Engineering, of course, will also get a chunk of this new funding, with an emphasis on relatively new products like MLFlow and Delta, two tools Databricks recently developed and that make it easier to manage the life cycle of machine learning models and build the necessary data pipelines to feed them.


Source: The Tech Crunch

Read More

DARPA wants to build an AI to find the patterns hidden in global chaos

Posted by on Jan 7, 2019 in Artificial Intelligence, DARPA, Government, machine learning | 0 comments

That most famous characterization of the complexity causality, a butterfly beating its wings and causing a hurricane on the other side of the world, is thought-provoking but ultimately not helpful. What we really need is to look at a hurricane and figure out which butterfly caused it — or perhaps stop it before it takes flight in the first place. DARPA thinks AI should be able to do just that.

A new program at the research agency is aimed at creating a machine learning system that can sift through the innumerable events and pieces of media generated every day and identify any threads of connection or narrative in them. It’s called KAIROS: Knowledge-directed Artificial Intelligence Reasoning Over Schemas.

“Schema” in this case has a very specific meaning. It’s the idea of a basic process humans use to understand the world around them by creating little stories of interlinked events. For instance when you buy something at a store, you know that you generally walk into the store, select an item, bring it to the cashier, who scans it, then you pay in some way, and then leave the store. This “buying something” process is a schema we all recognize, and could of course have schemas within it (selecting a product; payment process) or be part of another schema (gift giving; home cooking).

Although these are easily imagined inside our heads, they’re surprisingly difficult to define formally in such a way that a computer system would be able to understand. They’re familiar to us from long use and understanding, but they’re not immediately obvious or rule-bound, like how an apple will fall downwards from a tree at a constant acceleration.

And the more data there are, the more difficult it is to define. Buying something is comparatively simple, but how do you create a schema for recognizing a cold war, or a bear market? That’s what DARPA wants to look into.

“The process of uncovering relevant connections across mountains of information and the static elements that they underlie requires temporal information and event patterns, which can be difficult to capture at scale with currently available tools and systems,” said DARPA program manager Boyan Onyshkevych in a news release.

KAIROS, the agency said, “aims to develop a semi-automated system capable of identifying and drawing correlations between seemingly unrelated events or data, helping to inform or create broad narratives about the world around us.”

How? Well, they have a general idea but they’re looking for expertise. The problem, they note, is that schemas currently have to be laboriously defined and checked by humans. At that point you might as well inspect the information yourself. So the KAIROS program aims to have the AI teach itself.

At first the system will be limited to ingesting data in massive quantities to build a library of basic schemas. By reading books, watching news reports, and so on it should be able to create a laundry list of suspected schemas, like those mentioned above. It might even get a hint of larger, more hazy schemas that it can’t quite put its virtual finger on — love, racism, income disparity, etc — and how others might fit into them and each other.

Next it will be allowed to look at complex real-world data and attempt to extract events and narratives based on the schemas it has created.

The military and defense applications are fairly obvious: imagine a system that took in all news and social media posts and informed its administrators that it seemed likely there would be a run on banks, or a coup, or a new faction emerging from a declining one. Intelligence officers do their best to perform this task now, and human involvement will almost certainly never cease, but they would likely appreciate a computer companion saying, “there are multiple reports of stockpiling, and these articles on chemical warfare are being shared widely, this could point to rumors of terrorist attack” or the like.

Of course at this point it is all purely theoretical, but that’s why DARPA is looking into it: the agency’s raison d’etre is to turn the theoretical into the practical, or failing that, at least find out why they can’t. Given the extreme simplicity of most AI systems these days it’s hard to imagine one as sophisticated as they clearly want to create. Clearly we have a long way to go.


Source: The Tech Crunch

Read More

Tencent AI Lab loses key executive

Posted by on Jan 3, 2019 in andrew ng, Artificial Intelligence, Asia, Baidu, bytedance, computing, game publisher, machine learning, natural language processing, online games, optical character recognition, Seattle, Speech Recognition, Tencent, Toutiao, WeChat, Y Combinator | 0 comments

Chinese internet giant Tencent just lost a leading artificial intelligence figure. Zhang Tong, who previously worked at Yahoo, IBM and Baidu, has stepped down after directing Tencent’s AI Lab for nearly two years.

The scientist will return to academia and continue research in the AI field, Tencent confirmed with TechCrunch on Thursday, adding that it hasn’t appointed a successor.

”We are grateful for [Zhang]’s contributions to Tencent AI Lab and continue to explore fundamental and applied research that can make the benefits of AI accessible to everyone, everywhere,” Tencent said in a statement.

Zhang’s departure is the latest in a handful of top AI scientists quitting large Chinese tech firms. In 2017, search giant Baidu lost its chief scientist Andrew Ng who started Google’s deep learning initiative. Last year, the firm suffered another blow as renown AI expert Lu Qi resigned as chief operating officer and moved onto spearheading Y Combinator’s newly minted China program.

Talent is key to a tech firm’s AI endeavor, for a revered leader not only inspires employees but also boosts investor confidence. Baidu stocks plunged following Lu’s exit as markets weighed on the talent gap inside the company, which had poured resources into autonomous driving, smart speakers among other AI efforts. Tencent itself had poached Zhang from Baidu’s Big Data Lab to ramp up its own AI division.

Tencent is best known for its billion-user WeChat messenger and being the world’s largest video game publisher, but it’s also been doubling down on machine learning R&D to serve users and enterprise clients. It launched the AI Lab in April 2016 and opened its first U.S. research center in Seattle a year later to work on speech recognition and natural language processing (NLP).

The AI Lab dives into machine learning, computer vision, speech recognition and NLP. Meanwhile, the social and entertainment giant also works to put fundamental research to practical use, applying AI to its key businesses — content, social, online games and cloud computing.

One beneficiary has been WeChat, which applies NLP to enable seamless dialogues between users speaking different languages. Another case in point is Tencent’s news aggregator Tiantian Kuaibao, which deploys deep learning to recommend content based on readers’ past preference. Kuaibao is a direct competitor to Jinri Toutiao, the popular AI-powered news app run by TikTok’s parent company ByteDance.

To date, Tencent’s AI Lab has a team of 70 research scientists and 300 engineers, according to information on its website. Tencent operates another AI initiative called the Youtu Lab, which focuses on image understanding, face recognition, audio recognition and optical character recognition. While its sister AI Lab falls under Tencent’s research-focus Technology Engineering Group, Youtu is the brainchild of the Cloud & Smart Industries Group, a new unit that Tencent set up during its major organizational reshuffle in October to place more emphasis on enterprise businesses.


Source: The Tech Crunch

Read More