What’s happened in AI: March 16th-23rd

Arguably the biggest week in AI this year in terms of new developments. Not all developments are positive and rosy though. This week we witnessed the tragic and unfortunate death of Elaine Herzberg, who was struck by one of Uber’s autonomous vehicles in Arizona.

While I’m a strong proponent of autonomous vehicles, especially given the potential positive long-term benefits they offer, I’m deeply saddened to learn about Elaine’s death. Most AI experts have said this crash was avoidable with today’s LIDAR and autonomous vehicle software, which only makes this fatal accident even more disturbing. This week’s post is dedicated to Elaine.

Company developments:

Facebook escalates AI talent wars, makes key hire from Paul Allen’s AI2 in push to grow Seattle team – Mar. 23, 2018 (GeekWire)

• The social network has signed up Luke Zettlemoyer, a computer science professor at the University of Washington who was most recently a senior research manager at Seattle’s Allen Institute for Artificial Intelligence, or AI2
• Zettlemoyer is considered a rising star in the AI field, and was among the researchers featured last November in a New York Times article about the bidding war for artificial intelligence talent. He left AI2 last week and began at Facebook this week. His move comes as the Allen Institute, created by Paul Allen, looks to ramp up its own hiring in a tight market for AI talent with help from an additional $125 million in funding from the Microsoft co-founder
• Facebook spokesman Ari Entin said Zettlemoyer will be working with the company’s Seattle-based computational photography team, which was formed in 2015 and is led by Microsoft veteran Rick Szeliski. But Entin added that Zettlemoyer will report to Menlo Park, Calif., under the broader aegis of Facebook AI Research, also known as FAIR

Instagram heeds user feedback, tweaks machine learning algorithm – Mar. 23, 2018 (CIO Dive)

• On Thursday, Instagram announced changes to its feed to make newer posts appear higher up for users. The change was prompted by user feedback criticizing a move made a couple years ago from a chronological feed to one curated by machine learning algorithms, which prioritized posts users were more likely to like and interact with
• Machine learning has also helped the company reduce cyberbullying and trolling by pinpointing inappropriate words and phrases in comments, said Systrom in an interview with WIRED. The DeepText system, which uses word embeddings to determine context, was brought on from Facebook in 2016 and has been taught to pinpoint obvious cases of spam and mean comments (such as an explicit word) as well as more subtle slurs and harassment

A 21-year-old Swedish AI prodigy wants to revolutionize the $6 trillion education industry – and Tim Cook is impressed – Mar. 23, 2018 (Business Insider Nordic)

• Founded by Joel Hellermark, a 21-year-old Swedish expert on artificial intelligence, Sana Labs wants to insert the latest advances in AI into education. “We want any education company in the world to be able to implement adaptivity in just a few days,” said Hellermark, also Sana Labs’ CEO
• Sana Labs is an education tech startup founded by Joel Hellermark, 21. It provides an artificial-intelligence platform designed to individualize a student’s learning in subjects like language and math. Applying AI to education has so far proved difficult, and Sana Labs hopes its scalable platform will change that

IDEMIA’s Facial Recognition Get’s Top NIST Ranking in Webcam and Selfie Categories – Mar. 22, 2018 (Mobile ID World)

• IDEMIA‘s facial recognition technology has attained the top ranking in the Webcam and Selfie categories of the National Institute of Standards and Technology’s latest Face Recognition Vendor Test. IDEMIA’s algorithm proved more accurate than those of 32 competitors
• The FRVT was launched at the start of last year, and tests biometric matching systems against images collected in civil and criminal databases, assessing them based on not only accuracy but also speed, memory, and resilience. Unlike previous NIST evaluation programs, FRVT performs its assessments on an ongoing basis, with organizations invited to submit their technologies at any time, so long as a given submission is not made within 120 days of a previous one
• In a statement announcing its latest FRVT wins, IDEMIA noted that this technology is being used in its MorphoFACE and ONELook solutions, the latter of which is a competitor in the 2018 Biometric Technology Rally, a Department of Homeland Security-organized event revolving around live trials of biometric technologies. Commenting on the FRCT victories, IDEMIA Identity & Security Business CEO Ed Casey said they “further confirm IDEMIA’s technology leadership in the highly competitive facial recognition marketplace.”

Google’s François Chollet: AI researchers with a conscience shouldn’t work at Facebook – Mar. 22, 2018 (VentureBeat)

• Facebook is under fire from a lot of critics this week as fallout continues from the Cambridge Analytica scandal, including from an unexpected source: Google. In a series of tweets published Thursday, Google researcher François Chollet warns that the problem with Facebook isn’t just the privacy breach or recent lack of trust, it’s the fact that Facebook, powered by AI, can soon become a “totalitarian panopticon.”
• We’re looking at a powerful entity that builds fine-grained psychological profiles of over two billion humans, that runs large-scale behavior manipulation experiments, and that aims at developing the best AI technology the world has ever seen. Personally, it really scares me,” Chollet said in a Twitter thread. “If you work in AI, please don’t help them. Don’t play their game. Don’t participate in their research ecosystem. Please show some conscience.”

EA has started training AI players in Battlefield 1 – Mar. 22, 2018 (The Verge)

• At GDC today, EA announced that it’s been training AI agents in 2016’s WWI shooter Battlefield 1. The company says it’s the first time this sort of work has been done in a high-budget AAA title (which is disputable), but more importantly, it says the methods it’s developing will help improve future games: providing tougher, more realistic enemies for human players and giving developers new ways to debug their software
• EA’s AI agents — which, unlike bots, are expected to learn how to play instead of merely following instructions — are being trained using a combination of two standard methods: imitation learning and reinforcement learning. Both work exactly as you’d expect. The first part involves the agent watching human players and then attempting to mimic them. That constitutes roughly 2 percent of their knowledge, EA tells us, and sets them up on the right path

Self-driving Uber car kills Arizona woman crossing street – Mar. 19, 2018 (Reuters)

• An Uber self-driving car hit and killed a woman crossing the street in Arizona, police said on Monday, marking the first fatality involving an autonomous vehicle and a potential blow to the technology expected to transform transportation
• The ride services company said it was suspending North American tests of its self-driving vehicles, which are currently going on in Arizona, Pittsburgh and Toronto
• “This tragic accident underscores why we need to be exceptionally cautious when testing and deploying autonomous vehicle technologies on public roads,” said Democratic Senator Edward Markey, a member of the transportation committee, in a statement

New Uber patent shows how self-driving cars might talk to pedestrians – Mar. 19, 2018 (TechRadar)

• Per a recently filed patent: flashing lights and signs on the outside of self-driving cars that tell pedestrians exactly what’s happening. One image shows a “please proceed to cross” sign on the front grill
• These “intention outputs” will show other road users when an autonomous vehicle is about to change lanes, set off, pull to a stop, and so on. It’ll be a bit like watching for the green crossing signal, only embedded on a car
• The system is designed to replace the nod of the head or the flash of the lights we might use today, Uber’s Sean Chin told The Verge: “While we don’t have final implementation, what we’re considering is what is a new language we can create to give people that information.”

General Motors: Autonomous Vehicle Production Begins Next Year – Mar. 18, 2018 (The Drive)

• Though no news about GM being greenlit for operation of autonomous vehicles has surfaced, GM has stated a belief that its vehicle will satisfy policymakers, and on Thursday, it declared that the Cruise AV would enter production in 2019 at the company’s Orion Township facility in Michigan, where the Bolt EV and Sonic are built
• “We will launch the Cruise AV in a ridesharing environment where the vehicles can be managed in a fleet,” stated Sullivan. Launch markets for the Cruise AV and associated mobility services are not yet known, but Waymo-friendly California and Arizona are considered likely candidates

Tencent is putting a robot research lab in China’s manufacturing heartland – Mar. 16, 2018 (MIT News)

• Tencent, the tech titan behind China’s biggest social networking and chat platform, WeChat, is about to bring its AI research to life by opening a robotics lab in the country’s center of manufacturing, Shenzhen
• Tencent has already researched many kinds of AI algorithms, and even has a rival to DeepMind’s Go-playing program AlphaGo, called Fine Art. But it’s more difficult to have AI software control systems that operate in the messy real world. The challenge of interacting with real objects can also feed back into AI research on vision and language

Wave Computing close to unveiling its first AI system – Mar. 16, 2018 (ZDnet)

• The rapid evolution of deep learning has started an AI arms race. Last year, venture capitalists poured more than $1.5 billion into semiconductor start-ups and there are now some 45 companies designing chips purpose-built for artificial intelligence tasks including Google with its Tensor Processing Unit (TPU)
• After quietly testing its “early access” system for nearly a year, one of these startups, Wave Computing, is close to announcing its first commercial product. And it is promising that a novel approach will deliver some big gains in terms of both performance and ease of use for training neural networks
• Wave’s dataflow architecture is different. The Dataflow Processing Unit (DPU) does not need a host CPU and consists of thousands of tiny, self-timed processing elements designed for the 8-bit integer operations commonly used in neural networks

Partnerships:

deepsense.ai Becomes NVIDIA Deep Learning Partner – Mar. 22, 2018 (PR Newswire)

• The company joins the prestigious NVIDIA Service Delivery Partner program as one of a handful of preferred partners worldwide providing professional services in deep learning
• Tomasz Kułakowski, deepsense.ai CEO, emphasizes the importance of this partnership: “We are very proud to join the select group of companies recommended by NVIDIA as professional deep learning service providers. This ensures that we invest our efforts in the right way and build our competencies around NVIDIA’s cutting-edge AI technology.”
• deepsense.ai has experience in delivering both research and commercial data science projects based on machine learning, deep learning and reinforcement learning. The solutions designed by the company include various prediction models for customer analytics, recommendation systems, predictive maintenance, image processing models such as defect detection, product recognition, brand visibility analysis, satellite and drone imagery analytics, as well as reinforcement learning-based vehicle automation and artificial limbs control

Apple, IBM add machine learning to partnership with Watson-Core ML coupling – Mar. 19, 2018 (TechCrunch)

• Today, they took that friendship a step further when they announced they were providing a way to combine IBM Watson machine learning with Apple Core ML to make the business apps running on Apple devices all the more intelligent
• Apple introduced Core ML at the Worldwide Developers Conference last June as a way to make it easy for developers to move machine learning models from popular model building tools like TensorFlow, Caffe or IBM Watson to apps running on iOS devices
• IBM also announced a cloud console to simplify the connection between the Watson model building process and inserting that model in the application running on the Apple device. Over time, the app can share data back with Watson and improve the machine learning algorithm running on the edge device in a classic device-cloud partnership

China Develops Artificial Intelligence for “Human-Like” Autonomous Driving, in Partnership with Canadian University – Mar. 16, 2018 (Global Research)

• The partnership between Waterloo and the Qingdao Academy of Intelligent Industries (QAII) and the State Key Laboratory for Management and Control of Complex Systems (SKL-MCCS) was solidified in an agreement recently signed by all parties
• The agreement outlines a number of initiatives, including the establishment of a shared research centre for automated driving, faculty and graduate student exchanges, a Waterloo PhD program focused on autonomous vehicles, and the potential for Chinese startup companies to establish research and development facilities in the Waterloo Region
• The Chinese partners will collectively provide up to $1M CDN per year for five years to initiate collaborative activities. Waterloo has committed to providing $4M CDN to build a new autonomous lab facility in 2018 and is seeking further government matching funds to support this initiative

Government / policy:

Niti Aayog to come out with national policy on artificial intelligence soon – Mar. 21, 2018 (The Economic Times)

• The policy is expected to lay out short-, medium- and long-term goals to be achieved by 2022, 2026 and 2030, as India gears up to meet its commitment towards sustainable development goals, starting 2018
• Under the policy, deadlines for commercial rollout of AI may also be proposed in areas like agriculture, health, education, banking, retail and transportation. It may also suggest incentives for startups and venture capital funds that undertake research and adoption of AI. These guidelines would be based on half a dozen pilots the Aayog has undertaken on AI in areas of agriculture, health, education and creation of other social infrastructure
• A senior government official told ET that the government was considering to set up another high-level committee which would be responsible for the implementation and monitoring of AI

U.S. spending plan include $100 mln for autonomous cars research, testing – Mar. 21, 2018 (Yahoo Finance)

• A spending bill unveiled on Wednesday night includes $100 million for a highly automated “vehicle research and development” program and will include funding for assessing the job impact of self-driving cars
• The new funding includes $60 million for grants “to fund demonstration projects that test the feasibility and safety” of self-driving vehicles, the spending agreement said
• The funding includes $38 million for U.S. agencies to conduct research into self-driving cars, including cyber-security issues

Coalition presses Transportation Dept. for stricter oversight of driverless cars – Mar. 19, 2018 (The Hill)

• In a letter signed by more than 25 organizations, the group’s leaders call the Transportation Department (DOT) and the National Highway Traffic Safety Administration (NHTSA) “detached spectators instead of engaged safety regulators” on autonomous vehicles
• The letter comes as legislation that would speed up the development and testing of autonomous vehicles remains stuck in the Senate after unanimously passing through the Senate Commerce, Science and Transportation Committee last year

New bill would prepare US for artificial intelligence threat – Mar. 19, 2018 (Defense News)

• Rep. Elise Stefanik, chair of the House Armed Services Subcommittee on Emerging Threats and Capabilities, introduced legislation on AI Wednesday, which she hopes to add into this year’s defense policy bill. Her bill would develop a commission to review advances in AI, identify the nation’s AI needs and make recommendations to organize the federal government for the threat
• Stefanik’s legislation comes as U.S. officials are increasingly worried about China making a major government-led push on AI, and it lines up with calls from the military AI space for America to launch an all-of-government approach to the problem. Defense Secretary Jim Mattis told lawmakers last month that AI is making him question what impact AI will have on the nature of war, and the Pentagon plans to increase AI-related investments over the next two years
• Stefanik’s proposed national security commission on artificial intelligence would address and identify America’s national security needs with respect to AI. It would also look at ways to preserve America’s technological edge, foster AI investments and research, and establish data standards and open incentives to share data

Studies / breakthroughs:

China will win the A.I. race, according to Credit Suisse – Mar. 22, 2018 (CNBC)

• China will be number one in artificial intelligence due to the country’s lack of “serious law” about data protection, said Dong Tao, vice chairman for Greater China at Credit Suisse Private Banking Asia Pacific
• But, at some point, tighter data privacy laws will be introduced in China too, Tao said
• It would be a challenge for authorities to balance protecting the privacy of tech users and not hurting the sector’s growth, he added

China trails US in every area of AI development except big data, Oxford University report finds – Mar. 19, 2018 (South China Morning Post)

• China is given a score of 17 for its overall capacity for developing technologies in the field, compared with 33 for the US, its main competitor in the global AI race, in the report published by Oxford’s Future of Humanity Institute this month
• It found that except for “access to data”, China trails the US in every driver of AI development, namely hardware, research and algorithm and the commercialisation of the industry. The biggest obstacle is likely to be production of hardware such as microprocessors and chips, the study found
• Access to large quantities of data, aided by relatively lax privacy protection laws, has been cited as one of China’s biggest advantages in AI development. The country’s technology giants collect vast amount of data, and sharing among government agencies and companies is common

Scientists use Artificial Intelligence to map more than 6,000 new craters on the Moon – Mar. 17, 2018 (First Post)

• A new artificial intelligence (AI)-based lunar mapping technology has accurately mapped over 6,000 new craters on Earth’s moon in just hours, the media reported
• For the study, published in the journal Icarus, the team first trained the convolutional neural network on a dataset covering two-thirds of the moon. They then tested the neural network on the remaining third of the moon
• The results yield 92 percent of craters from human-generated test sets and almost twice of the total number of crater detections

Hope for millions after life-saving AI is revealed that can diagnose prostate cancer as accurately as a doctor – Mar. 16, 2018 (Daily Mail)

• Prostate cancer can be diagnosed by new AI just as accurately as any doctor potentially saving millions of lives, new research suggests. Chinese scientists and doctors have developed an artificial intelligence system which they say can accurately diagnose and identify cancerous samples
• This artificial intelligence learning system, presented at the European Association of Urology congress in Copenhagen, Denmark, has shown similar levels of accuracy to a human pathologist.

Too Much Sun Could Wreak Havoc on Driverless Cars – Mar. 16, 2018 (Bloomberg)

• The threat comes from solar storms, those occasional eruptions of vast amounts of energy that can cause a massive spike in geomagnetic activity and radiation
• While these storms aren’t immediately evident to human drivers, they can sever the data connection between a vehicle’s global-position system and the satellites that supply location information. That’s what could spell trouble for driverless cars now under development, at least if engineers aren’t careful
• The engineers behind automated vehicles, meanwhile, are already taking steps to outsmart the sun. Self-driving systems mostly base navigation on a field of sensors, including laser pulses known as lidar, that read the immediate surroundings and speak directly to the vehicle’s computer nerve system

Hospital leans on machine learning to reduce sepsis-related mortality rate – Mar. 16, 2018 (Healthcare IT News)

• Implementing machine learning technology specifically designed to fight sepsis in part through clinician alerts, the organization saw the sepsis-related in-hospital mortality rate was 33.5 percent lower during the post-implementation period and the average sepsis-related hospital length of stay was 17.1 percent lower during the same period
• Analyses included 2,298 adult patients in the emergency department and intensive care unit
• InSight collects basic data from the EHR and computes sepsis risk by comparing the patient data profile and trends against statistical patterns from its reference library of millions of prior cases of patients with sepsis. InSight runs in the background, so there is no additional action needed from clinicians

Fundraising:

Skyline AI raises $3M from Sequoia Capital to help real estate investors make better decisions – Mar. 22, 2018 (TechCrunch)

• Skyline AI, an Israeli startup that uses machine learning to help real estate investors identify promising properties, announced today that it has raised $3 million in seed funding from Sequoia Capital. The round will be used to build its tech platform and hire experts in data science and machine learning
• Founded in 2017 and headquartered in Tel Aviv, Skyline AI predicts future property values and also analyzes the real estate market to help investors make important decisions such as when to raise rents, renovate or sell
• Co-founder and chief executive officer Guy Zipori told TechCrunch that Skyline AI’s founding team (who also includes chief technology officer Or Hiltch, chief revenue officer Iri Amirav and executive chairman Amir Leitersdorf) worked together for years at various artificial intelligence-based startups in sectors including security, healthcare and online video. After several of their companies exited, the four were in a position to find investment opportunities

AI game trainer Gosu.ai raises $1.9M to give gamers a virtual assistant – Mar. 21, 2018 (TechCrunch)

• It’s now raised a $1.9M funding round led by Runa Capital, with participation from Ventech and existing investor, Sistema_VC. Previously, the startup was backed by Gagarin Capital, a new Silicon Valley-based early-stage VC firm focusing on AI investments, which invested in Prisma and MSQRD, which exited to Facebook and Google, respectively
• Gosu.ai provides tools and guidance for users to improve their skills in competitive games. It analyzes their matches and makes personal recommendations. It also helps players prep, suggesting gear sets, starting items and offering ideas on how to take on a particular opponent. The platform currently works with Dota 2, with plans to support CS:GO and PUBG in the near future
• The company was founded by Alisa Chumachenko (pictured), who was the creator and former CEO of Game Insight, a big gaming world player. She says: “There are 2 billion gamers in the world now and 600 million of them play hardcore games, such as MOBAs, Shooters and MMOs. We can help those players reach their full potential with our AI assistants.”

Chinese AI Medical Imaging Start-up Infervision Raises $47M From Sequoia, Qiming – Mar. 19, 2018 (China Money Network)

• Infervision, a Chinese artificial intelligence start-up developing medical imaging technology, has announced to complete a RMB300 million (US$47 million) series B+ round from existing investors Sequoia Capital China, Qiming Venture Partners and Genesis Capital
• New investors in the round include Xianghe Capital, a venture capital firm funded by Hesong Tang, Baidu’s former head of corporate development, and Advantech Capital, a private equity fund focusing on tech and medical industries
• Founded in 2016 and headquartered in Beijing, Infervision applies deep learning technology to assist medical image diagnosis. Its product, AI-CT, works to identify the core characteristics of lung caner and detect suspicious lesions in computerized tomography scans. The company says the technology is especially sensitive to hard-to-detect nodules such as semi-solid and ground glass nodules. Another product, AI-DR, which targets cardiothoracic lesions, can reduce digital radiography analysis time from ten minutes to as little as five seconds, the company say

Universities:

Natural-language processing technology boosts BI at Penn Medicine – Mar. 20, 2018 (Tech Target)

• Penn Medicine implemented Linguamatics Health I2E, a natural-language processing platform designed for healthcare organizations, to build queries and automatically mine data from various sources containing unstructured information, such as doctor’s notes recorded in electronic health records and specialist reports
• Such documents contained different forms of unstructured data — notably free text and text containing specialized medical language such as pathology reports as well as documents that contain a combination of discreet data points and free text
• Birtwell said the organization selected technology from Linguamatics, based in Marlborough, Mass., after surveying the marketplace and speaking to peer institutions who had implemented natural-language processing technology. After running a few projects as proofs of concept with the platform, Penn Medicine implemented it more widely in 2015

Top schools for AI: New study ranks the leading U.S. artificial intelligence grad programs – Mar. 20, 2018 (GeekWire)

• Here are the top five graduate-level artificial intelligence programs, according to the report:
  1. Carnegie Mellon University
  2. Massachusetts Institute of Technology
  3. Stanford University
  4. University of California — Berkeley
  5. University of Washington

Events:

At SXSW, artificial intelligence talk focused on more pragmatic, near-term possibilities – Mar. 23, 2018 (512 Tech)

• Unfortunately, artificial intelligence can’t always tell the difference, and that missing context ranks among the key challenges developers face as they try to craft a smoother and more-useful interaction between humans and machines, the principal program manager at Microsoft Research AI said at South by Southwest earlier this month
• While several panels still contemplated our long-term future — one, for example, considered the possibility of consciousness in machines — most of the researchers, developers and pundits set their sights on the challenges AI faces in the here and now
• Adam Cheyer, the co-founder of Viv Labs and one of Siri’s creators, joined Watson on an “Exploring Innovations in AI” panel and shared her sense that significant breakthroughs were coming in the near future. Cheyer said most people will use an AI-powered assistant for almost everything they currently use the Web for, and it will happen within the next two to three years

AI will be CITE 2018’s major attraction – Mar. 19, 2018 (ECNS.CN)

• The artificial intelligence exhibition hall will be the major attraction in the 2018 China Information Technology Expo, set to be held in the Shenzhen special economic zone next month
• “Big-name Chinese AI and electronic companies, including Huawei, Lenovo, TCL, Hisense, Haier, Skyworth, Changhong, Oppo and Vivo will participate in CITE 2018, which has become a major event in the AI and electronics industries in the country,” Song said at a press conference in Guangdong’s provincial capital over the weekend
• In addition to the AI exhibition hall, CITE 2018 will also establish special exhibition halls for digital families, the internet of things, new energy cars, new types of displays, smart terminals, lithium batteries and high-end chips to display China’s great achievements and new developments in high-tech industries