What’s happened in AI: October 22nd-28th

By | October 31, 2018

Big highlight this week is Baidu’s development of simultaneous language translation through one of their AI programs. It’s considered a major step forward in the natural language processing realm.

Elsewhere on the M&A side Lyft made an acquisition of Blue Vision to further enhance their autonomous vehicle program.

Company developments:

CrunchBot AI, Inc. receives Qlik’s TED accreditation for their AI based Natural Language Extension – Oct. 28, 2018 (PR Web)

  • CrunchBot AI, Inc., the world’s first AI based Analytics ChatBot, today announced achieving accreditation for its analytics extension through Qlik’s Trusted Extension Developer (TED) program. The accreditation highlights the solution’s enterprise readiness and proven integration of the offering which is available today only with the Qlik platform. The solution will help organizations looking to democratize data and raise business user data literacy by spreading analytics insights across the enterprise using natural language query capabilities. Using the CrunchBot extension, users will be able to have more relevant and deeper conversations with all their enterprise data assets within their current workflows

Facial recognition startup Kairos founder continues to fight attempted takeover – Oct. 27, 2018 (TechCrunch)

  • There’s some turmoil brewing over at Miami-based facial recognition startup Kairos. Late last month, New World Angels President and Kairos board chairperson Steve O’Hara sent a letter to Kairos founder Brian Brackeen notifying him of his termination from the role of chief executive officer. The termination letter cited willful misconduct as the cause for Brackeen’s termination. Specifically, O’Hara said Brackeen misled shareholders and potential investors, misappropriated corporate funds, did not report to the board of directors and created a divisive atmosphere

Tesla Pushing Out “Drive On Navigation” Autopilot Feature Tonight – Oct. 27, 2018 (Clean Technica)

  • The new Drive on Navigation feature will allow Tesla owners with Autopilot hardware 2.5 and Enhanced Autopilot software to use Autopilot to navigate many of the scenarios commonly encountered on highways, including transitioning from one highway to another, automatic lane changes after driver confirmation, and even taking the exit — as long as the vehicle is following a navigation route in the car’s nav system
  • The feature will also include the ability to automagically change lanes while on the freeway, though this feature will initially require a confirmation from the driver. Elon clarified the implementation in a response to Jaden Smith that the feature will require the driver to hit the turn signal for the first couple of million miles. Assuming the feature passes the test, the requirement to hit the turn indicator will be removed and the car will just fly along on the freeway towards its destination

Google releases AI based open source reinforcement learning framework – Oct. 24, 2018 (TechieExpert)

  • Google is now helping its researchers team to train the Artificial Intelligence models with the help of an open sourcing a reinforcement learning frameworks which are mainly used for its projects. This new release also consists of a set of collaboration which clarifies how to use the open source framework. Google framework has been designed with three new focuses” stability, flexibility, and productivity
  • The company is also providing the 15 code examples for the Arcade learning Environment, a platform which uses the video games that helps to evaluate the performance of the Artificial Intelligence technology along with the help of a four machine learning concept: Implicit Quantile Network, Rainbow agent, c51 and the aforementioned DQN

Facebook says it removed 8.7M child exploitation posts with new machine learning tech – Oct. 24, 2018 (TechCrunch)

  • Facebook announced today that it has removed 8.7 million pieces of content last quarter that violated its rules against child exploitation, thanks to new technology. The new AI and machine learning tech, which was developed and implemented over the past year by the company, removed 99 percent of those posts before anyone reported them, said Antigone Davis, Facebook’s global head of safety, in a blog post
  • The new technology examines posts for child nudity and other exploitative content when they are uploaded and, if necessary, photos and accounts are reported to the National Center for Missing and Exploited Children. Facebook had already been using photo-matching technology to compare newly uploaded photos with known images of child exploitation and revenge porn, but the new tools are meant to prevent previously unidentified content from being disseminated through its platform

Apple’s Tim Cook makes blistering attack on the ‘data industrial complex’ – Oct. 24, 2018 (TechCrunch)

  • Apple’s CEO Tim Cook has joined the chorus of voices warning that data itself is being weaponized against people and societies — arguing that the trade in digital data has exploded into a “data industrial complex”. “Our own information — from the everyday to the deeply personal — is being weaponized against us with military efficiency,” warned Cook. “These scraps of data, each one harmless enough on its own, are carefully assembled, synthesized, traded and sold
  • “Taken to the extreme this process creates an enduring digital profile and lets companies know you better than you may know yourself. Your profile is a bunch of algorithms that serve up increasingly extreme content, pounding our harmless preferences into harm.”

Baidu launches simultaneous language translation AI – Oct. 23, 2018 (VentureBeat)

  • Baidu has developed an AI system capable of simultaneously translating two languages at once. Aptly dubbed Simultaneous Translation and Anticipation and Controllable Latency (STACL), the Beijing company claims it represents a “major breakthrough” in natural language processing
  • STACL, unlike most AI translation systems, is capable of beginning a translation just a few seconds into a speaker’s speech and finishing seconds after the end of a sentence. It’s the opposite of consecutive interpretation, where a translator waits until the speaker pauses to start translating

Ping An Good Doctor Launches China’s First Online AI TCM Smart Listening Diagnosis – Oct. 23, 2018 (Business Insider)

  • China’s leading one-stop healthcare ecosystem platform, Ping An Good Doctor (the “Company”, stock code: 1833.HK), today announced the online launch of its traditional Chinese medicine (“TCM”) Smart Listening Diagnosis, the first in China and developed independently by the Company. As an important part of the Modern Physician Hua Tuo Project, the Smart Listening Diagnosis combines artificial intelligence (AI) medical technology and TCM theory. By quickly collecting users’ voices and performing AI analysis, it can accurately identify the TCM constitution type of the users
  • Thanks to the big data collection of its 228 million registered users, Ping An Good Doctor has been able to compile a large number of audio samples. These samples were manually identified and labeled by TCM doctors across China before they were sent to the AI learning system. After continuous research and development, the Smart Listening Diagnosis has achieved a highly accurate recognition rate

Ford expands self-driving vehicle program to Washington, D.C. – Oct. 22, 2018 (TechCrunch)

  • Ford is bringing its autonomous vehicles to Washington, D.C., the fourth city to join the automaker’s testing program as it prepares to launch a self-driving taxi and delivery service in 2021. Ford will begin testing its self-driving vehicles in the district in the first quarter of 2019. The company is already is testing in Detroit, Pittsburgh and Miami

Google’s Fluid Annotation uses AI to annotate image datasets quickly – Oct. 22, 2018 (VentureBeat)

  • Annotation is often the most arduous part of the artificial intelligence (AI) model training process. That’s particularly true in computer vision — traditional labeling tools require human annotators to outline each object in a given image. Labeling a single pic in the popular Coco+Stuff dataset, for example, takes 19 minutes; tagging the whole dataset of 164,000 images would take over 53,000 hours
  • Fortunately, Google’s developed a solution that promises to cut down on labeling time dramatically. It’s called Fluid Annotation, and it employs machine learning to annotate class labels and outline every object and background region in a picture. Google claims it can accelerate the creation of labeled datasets by a factor of three

Google-incubated AdLingo client Valassis Digital sees 10 times more efficiency versus Facebook News Feed ads – Oct. 22, 2018 (VentureBeat)

  • “Discoverability — we know that’s a problem,” Google’s Vic Fatnani said onstage at VB Summit 2018. “Imagine if you could bring your conversational assistant to a display ad.” That’s the idea behind AdLingo, a startup Fatnani cofounded within Google’s Area 120 incubator. It’s the product of two years of development, and it essentially builds a Facebook Messenger-like experience within traditional web and mobile ad units
  • AdLingo works with a variety of chatbot tools, including Microsoft Bot Framework, LiveEngage, Blip, and Google’s Dialogflow. And to kick off a few early campaigns, the startup partnered with chatbot provider Take; LivePerson, a conversational commerce provider; and Valassis Digital, a digital marketing agency that works with more than 1,500 advertisers in the U.S., including Hyundai, Kia, Fiat Chrysler, and Ford


Lyft speeds ahead with its autonomous initiatives – Oct 23, 2018 (TechCrunch)

  • TechCrunch has learned that it is acquiring the London-based augmented reality startup Blue Vision Labs and unveiling its first test vehicle to advance its vision for self-driving cars Blue Vision will sit within Lyft’s Level 5 autonomous car division headed up by Luc Vincent (who joined the company last year as VP of engineering after creating and running Google Street View)
  • Blue Vision has developed technology that provides both street level mapping and interactive augmented reality that lets two people see the same virtual objects. The company has already built highly detailed maps that developers can now use to develop collaborative AR experiences — it’s like the maps of these spaces become canvasses for virtual objects to be painted on. Over time, we may see various uses of it throughout the Lyft platform, but for now the main focus is Level 5

Oracle acquires DataFox, a developer of ‘predictive intelligence as a service’ across millions of company records – Oct. 22, 2018 (TechCrunch)

  • Oracle announced that it has made another acquisition, this time to enhance both the kind of data that it can provide to its business customers, and its artificial intelligence capabilities: it is buying DataFox, a startup that has amassed a huge company database — currently covering 2.8 million public and private businesses, adding 1.2 million each year — and uses AI to analyse that to make larger business predictions. The business intelligence resulting from that service can in turn be used for a range of CRM-related services: prioritising sales accounts, finding leads, and so on

Fundraising / investment:

China’s ByteDance leapfrogs Uber to becomes world’s most valuable startup – Oct. 26, 2018 (TechCrunch)

  • Valuation is according to reports from Forbes and Bloomberg, both of which claim that the company has completed a $3 billion investment that values the company at $75 billion. A source with knowledge of the deal confirmed the round to TechCrunch and suggested that the value is pre-money, which, adding the round, would put ByteDance’s valuation at $78 billion. That’s ahead of Uber’s most recent $72 billion valuation, although the ride-hailing giant is being tipped to go public next year at a valuation of up to $120 billion
  • ByteDance operates a range of digital media platforms, but it is best known for Toutiao, its AI-based news aggregator that has become one of China’s most-used apps with over 120 million users, and short video platform TikTok, which recently gobbled up Musical.ly which ByteDance acquired via a $1 billion acquisition last year

Placer.ai raises $4 million to use AI to track foot traffic – Oct. 24, 2018 (VentureBeat)

  • Placer.ai, which uses smartphone tracking data to help businesses make decisions, announced the closure of a $4 million funding round today. The funds will be used to expand its service for tracking and predicting foot traffic and people’s movements and to grow its sales and marketing teams.
  • The company said it obtains its data through the 100 popular smartphone apps that use its SDK and that it can anonymously but accurately track the movements of 60 percent of Android users and 40 percent of iOS smartphone users in the United States

Customer service ‘behavioral pairing’ startup Afiniti quietly raised $130M at a $1.6B valuation – Oct. 23, 2018 (TechCrunch)

  • Afiniti, which uses machine learning and behavioral science to better match customers with customer service agents — “behavioral pairing” is how it describes the process — has closed a $130 million round of funding ($75 million cash, $60 million debt) — a Series D that Afiniti CEO Zia Chishti says values his company at $1.6 billion
  • If you are not familiar with the name Afiniti, you might not be alone. The company has been relatively under the radar, in part because it has never made much of an effort to publicise itself, and in part because the funding that it has raised up to now has largely been from outside the hive of VCs that swarm around many other startup deals that push those startups into the limelight. At the same time, its backers make for a pretty illustrious list. This latest round includes former Verizon CEO Ivan Seidenberg; Fred Ryan, the CEO and publisher of the Washington Post; and investors Global Asset Management, The Resource Group (which Chishti helped found), Zeke Capital, as well as unnamed Australian investors

Bright Machines lands $179M to bring smarter robotics to manufacturing – Oct. 23, 2018 (TechCrunch)

  • The startup wants to bring a software-driven approach to robotics, one that would let you take dumb robotics and program it in a more automated fashion to perform a set of tasks, taking advantage of artificial intelligence and machine learning in ways that they say most manufacturing companies simply aren’t equipped to handle right now
  • This is clearly not your typical Series A and Bright Machines does not appear to be a typical Series A company, feeling its way trying to get a product to market. Perhaps that’s because the company began life as incubated project inside Flex, a customized manufacturing company. It was then spun out as a startup called AutoLab AI and changed the name to Bright Machines today for the big company unveiling. It already boast over 300 employees and brought in CEO, Armar Hanspal, who was most recently co-CEO at Autodesk to run the show. Former Autodesk CEO Carl Bass is a board member. Other board members include Mike McNamara, CEO of Flex and Steve Luszo, CEO of Seagate. Eclipse led the round

Predictive sales tool People.ai racks up $30M Series B led by Andreessen Horowitz – Oct. 23, 2018 (TechCrunch)

  • People.ai, a startup that tracks every communication touchpoint between sales teams and customers, wants to solve this problem. Now, the company (and the youngest Y Combinator graduate to make the accelerator’s list of its most successful startups) has attracted the attention of Andreessen Horowitz, scoring a fresh $30 million to move forward on this mission. Also participating in the round were Series A investors Lightspeed Venture Partners, GGV Capital and Y Combinator. In addition to the investment, Andreessen Horowitz general partner Peter Levine is joining People.ai’s board
  • The startup, founded by Oleg Rogynskyy, previously raised $7 million. It started as a software meant to give sales managers a predictive playbook for the best way to close a deal, but investors have a master plan for the long term. While this kind of live data mapping tech resembles an acquisition target for Microsoft or Salesforce, it’s no secret that Andreessen likes to build massive software franchises like Skype, Airbnb and GitHub. As we enter Q4 of 2018, early-stage SaaS investment is stabilizing and public cloud stocks are soaring. Salesforce continues to pump more money into the AI sales concept, paving the way for startups like People.ai to thrive. But when it comes to exit strategy, selling to a large enterprise player is not the goal.

EyeSight scores $15M to use computer vision to combat driver distraction – Oct. 23, 2018 (TechCrunch)

  • Tel Aviv-based EyeSight has developed an in-car “AI vision” system that claims to be able to detect when a driver loses concentration or gets dangerously distracted. Using advanced facial processing, it tracks a driver’s gaze direction, pupil dilation, eye openness and head position and uses proprietary algorithms to determine attentiveness. The resulting “smart car” can either do something to alert the driver (e.g. sounds and vibrations) or potentially temporarily activate self-driving mode

Ople raises $8 million for automated AI model design – Oct. 23, 2018 (VentureBeat)

  • Ople enables customers to quickly develop AI models by using datasets to detect patterns and relationships. In that way, it’s not unlike R2.ai, a startup with a suite of machine learning services that train high-quality models, and Feature Labs, which automates feature engineering by using techniques that automatically create algorithms from datasets
  • But Ople’s differentiated in key ways. Its platform is available on Amazon Web Services, and it continuously learns from every model built while simultaneously gaining speed and accuracy. Alves calls the approach behavioral assimilation, or BASS, and claims it can increase a data science team’s ability to create a model by a factor of 10 compared to competing solutions. “By using Ople, companies are leaping past their competition, making better decisions — faster and positioned to seize new market opportunities first,” Alves said


Honda partners with universities to investigate human-like AI – Oct. 25, 2018 (VentureBeat)

  • Artificial intelligence (AI) that reasons like a human remains elusive, but Honda hopes to make inroads. The Tokyo company’s U.S.-based Research Institute today announced a collaboration with three academic institutions — the Massachusetts Institute of Technology (MIT), the University of Pennsylvania (Penn), and the University of Washington — to advance the field of artificial cognition
  • MIT’s Computer Science and Artificial Intelligence (CSAIL) lab, in partnership with Penn’s School of Engineering and Applied Science and the University of Washington’s Paul G. Allen School of Computer Science & Engineering, will develop prototypes, working examples, and demonstrations of what Honda calls the “mechanisms of curiosity.” Specifically, MIT CSAIL will focus its efforts on systems capable of predicting future percepts — concepts developed as a consequence of perception — and the effect of future actions, while Penn’s engineering department and the Paul G. Allen School will develop perception models informed by biology and robots that can work safely in human environments

Nvidia partners with Scripps to investigate the role of AI in genomics processing and analysis – Oct. 23, 2018 (VentureBeat)

  • The firms’ joint research will focus on whole genomic sequences (the complete DNA sequence of an organism’s genome), continuous physiological wearable devices and other sensors, and illness prevention — specifically, digital sensing prediction of atrial fibrillation, an irregular heartbeat that increases the risk of stroke. Future work will involve as-yet unselected diseases and datasets
  • Scripps will supply the bulk of the research datasets, one of which contains over 1,000 continuous heart rhythm recordings (another comprises the entire genomic sequences of 1,400 people ages 80 and older who’ve never been sick). And researchers from both companies will employ a combination of bespoke neural networks and pretrained models in research data experiments. Assuming all goes well, they’ll package up their work and tools and open-source them

Addison Lee aims to deploy self-driving cars in London by 2021 – Oct. 22, 2018 (The Guardian)

  • Self-driving car services could be on the streets of London within three years under a partnership between the private hire firm Addison Lee and the British driverless car pioneers Oxbotica. The companies have signed a deal to develop and deploy autonomous vehicles in the city by 2021
  • Oxbotica will start mapping more than 250,000 miles of public roads in and around London from next month, using its technology to create a comprehensive map of every traffic feature. While the link-up could eventually allow Addison Lee’s fleet of black Mercedes and Prius cabs to be driven autonomously, the 5,000 drivers in London will remain employed, the firm says. However, it could also offer a cheaper, autonomous ride-sharing version of its hire service. The first stage is likely to be in corporate shuttles, around airports or campuses

Research / studies:

AIs trained to help with sepsis treatment, fracture diagnosis – Oct. 27, 2018 (Ars Technica)

  • They used a reinforcement learning algorithm because those are considered effective when there are what they term “sparse reward signals.” In other words, a patient population this large is going to have a lot going on other than sepsis, and a lot of it will influence the results of any treatments, so the signal from effective treatments is going to be small and hard to spot. This approach was designed to increase the odds of spotting one
  • As was the large number of data points used to train the software: over 17,000 intensive care unit patients, and another 79,000 general hospital admissions from a total of over 125 hospitals. Data on the patients included 48 different bits of information, from vital signs and lab tests to demographic information. The algorithm used the data to identify the treatments that would maximize the 90-day survival of the patients. The researchers termed the resulting software AI Clinician

When it comes to driverless cars, old people and pets are out of luck, finds MIT – Oct. 26, 2018 (V3)

  • Millions of people have answered questions posed by MIT’s ‘Moral Machine’, about the ethical decisions that driverless cars will have to make when they reach our roads – and the results don’t look good for old folks. In this grisly referendum on life and death, the real winners were the young and the sociable. Given various scenarios, users around the globe consistently told the driverless car to squish animals and the elderly, and to spare the lives of people in groups and children
  • Although the results broadly followed the same trends around the world, there were some interesting regional differences. People in China and Japan, for example, were less inclined to spare the young over the old, and they also cared less about saving the lives of those with lots of money compared to their European and North American counterparts (for the record, your V3 team’s results showed that we put a high value on people with important office jobs, like journalists)

Government / policy:

Microsoft Says It Will Sell Pentagon Artificial Intelligence and Other Advanced Technology – Oct. 26, 2018 (New York Times)

  • Microsoft said on Friday that it would sell the military and intelligence agencies whatever advanced technologies they needed “to build a strong defense,” just months after Google told the Pentagon it would refuse to provide artificial intelligence products that could build more accurate drones or compete with China for next-generation weapons
  • The announcement — made quietly in a small, town-hall-style meeting with the software giant’s leadership and employees on Thursday, then published on a blog on Friday afternoon — starkly illustrated the radically different paths these leading American technology companies are taking as they struggle with their role in creating a new generation of cyberweapons to help, and perhaps someday replace, American warriors

China tests facial recognition at border crossing of Hong Kong-Zhuhai-Macau Bridge – Oct. 24, 2018 (South China Morning Post)

  • Hong Kong drivers travelling to the mainland via the newly opened Hong Kong-Zhuhai-Macau Bridge may in future leave their home return permits at home as the Chinese authorities test a combination of facial recognition and fingerprint analysis to speed up border crossings
  • High-resolution cameras, fingerprint matching and thermal-scanning technology are now being deployed for one immigration lane at the border control in Zhuhai, according to Shenzhen-based artificial intelligence firm Intellifusion, which is supplying the technology. The driver will be cleared to proceed if the fingerprints, facial image and car license-plate images match pre-registered information in the immigration database, without the need to show the ID card issued by mainland authorities to Hong Kong permanent residents for cross-border travel

Nvidia delivers its self-driving car safety report to the feds – Oct. 23, 2018 (The Verge)

  • Nvidia, one of the world’s best known manufacturers of computer graphics cards, released its autonomous driving safety report on Tuesday. The Santa Clara-based company, which for several years has been engaged in a high-stakes venture to build the “brains” that power self-driving cars for major automakers like Volvo, Volkswagen, and Mercedes-Benz parent Daimler, is only the fifth company to delivery its voluntary safety report to the US National Highway Traffic Safety Administration
  • The report’s release was timed to coincide with Nvidia’s GPU Technology Conference in Washington, DC, this week. “We are dedicated to working with regulators to deploy self-driving technologies for safer, more efficient roads, and hope this collaboration will bring us closer to a new era of transportation,” a spokesperson said

China launches national AI public service platform – Oct. 22, 2018 (Global Times)

  • A national public service platform of artificial intelligence (AI) resources has been launched in southwest China’s Chongqing Municipality, the Science and Technology Daily reported Monday
  • The data capacity of the platform will be not lower than 5 petabytes (one petabyre equals to 1,024 terabytes), with an image recognition rate higher than 95 percent, character recognition rate higher than 93 percent and voice recognition rate higher than 97 percent, according to standards set by the National Development and Reform Commission (NDRC)

Events / other:

A 19-year-old developed the code for the AI portrait that sold for $432,000 at Christie’s – Oct. 26, 2018 (Washington Post)

  • Robbie Barrat was 17 and bored in West Virginia when he started experimenting with artificial intelligence and art. First, he trained a computer to write original rap songs by feeding it 6,000 Kanye West lyrics. Then, he taught it how to make landscape paintings and nude portraits by feeding it thousands of images scraped off the Internet. He uploaded the code to GitHub, the code-sharing platform, so that others could download it and learn from it. And many did — including the French art collective known as Obvious, whose AI portrait sold for $432,500 at Christie’s on Thursday in an internationally celebrated art auction. That’s nearly 45 times higher than Christie’s original estimated sale price of $10,000, the auction house said in an article on its website
  • Obvious could not immediately be reached for comment Thursday, but in an interview with the Verge, Obvious co-founder Hugo Caselles-Dupré confirmed elements were borrowed from Barrat, who said he began corresponding on GitHub with the group in October 2017. “If you’re just talking about the code, then there is not a big percentage that has been modified,” Caselles-Dupré said. “But if you talk about working on the computer, making it work, there is a lot of effort there.”

Tableau Hosts Global Partner Summit and Unveils New Product Integrations at Tableau Conference in New Orleans – Oct. 22, 2018 (PR Newswire)

  • Tableau Software (NYSE: DATA), the leading analytics platform, welcomed more than 750 system integrators, value added resellers, and technology partners to its annual Global Partner Summit at Tableau Conference in New Orleans. At the event, the company announced plans to expand the global partner program with new certifications and trainings as well as support for bundled offers to better serve joint customers. Additionally, the company announced new product integrations with industry leading technology providers, from mapping to data cataloging, that extend Tableau’s powerful analytics capabilities to better fit customers’ needs. Tableau’s partner community has generated a series of new and exciting extensions and integrations showcased in the Tableau Conference Data Village