Discover more from Last Week in AI

Weekly text and audio summaries of the most interesting AI news, as well as editorials commenting on recent events

Over 20,000 subscribers

Skynet Today Last Week in AI News #52

Last Week in AI

Feb 17, 2020

Last Week in AI News #52

Subscribe for future emails here!

Mini Briefs

Group Asks Federal Trade Commission To Regulate Use Of Artificial Intelligence In Pre-Employment Screenings

The Electronic Privacy Information Center (EPIC) has filed a petition asking the Federal Trade Commission (FTC) to establish fair trade practices regarding the commercial use of artificial intelligence (AI) and charged that “HireVue, a leading provider of AI=based pre-employment screenings, is flouting national and international standards of transparency, fairness and accountability.” In particular, EPIC argued that applicants who seek employment at a company using HireVue’s assessment services have to submit to screening, and HireVue itself refuses to disclose its algorithm’s selection criteria.

Use of AI in employment screening is only one facet of EPIC’s petition, “which also cites the commercial use of AI for ranking tennis players, evaluating potential Airbnb guests, use of facial recognition for criminal justice purposes, etc.” According to EPIC, the unregulated use of AI technologies has caused serious harm to consumers who are subject to opaque and un-provable decision-making in multiple areas that is now carried out by AI. Even when business disclose their use of AI decision-making tools, they frequently fail to justify those tools’ accuracy, reliability, and necessity.

Emotion AI researchers say overblown claims give their work a bad name

When emotion recognition research is used for lucrative commercial applications, a great deal of nuance is lost. Companies like HireVue have deployed AI for a number of commercial applications from AI-assisted hiring to school assessments, claiming that the algorithms’ analysis of micro-expressions and voice can pick out the best workers or students. This hype around emotion recognition has a shaky basis in science, and “has created a backlash from tech ethicists and activists who fear that the technology could raise the same kinds of discrimination problems as predictive sentencing or housing algorithms for landlords deciding whom to rent to.”

This hype worries the researchers too, who agree that their work is being co-opted for commercial applications and say that a lack of government regulation is bad for both consumers and researchers. Emotion recognition is still a nascent technology, and no researcher in the field would claim that it can actually assess an individual’s internal emotions and experience. It can estimate how an individual’s emotions might be perceived by others or suggest trends, but the technology requires observing characteristics like body posture and gait, as well as using biometric sensors and audio, in addition to looking at someone’s face.

Many researchers think that a ban on emotion recognition technology is too broad, but also believe that a regulatory vacuum is harmful. In such a regulatory vacuum, companies are able to make outrageous claims about the emotion recognition technology they are using, putting buyers who are not well educated at risk. Because of the resulting use of these technologies in things like job assessments, the technologies have the power to determine people’s access to resources. With more restrictions, or perhaps even a ban, on the use of emotion recognition in commercial applications, science will have time to catch up to commercialization and either validate or repudiate the claims made by companies like HireVue.

Advances & Business

‘A good dog’: Robo-pup helps people with dementia - Jennie the robotic dog has sensors and voice control to interact with people and her behaviour is customisable via a phone app. She is used as part of live animal therapy for people living with dementia.
VA doctors are using artificial intelligence to diagnose cancer - A team of researchers at the James A. Haley Veterans’ Hospital in Tampa, Florida, is revolutionizing the way cancer is documented by enlisting the help of a computer to diagnose the disease in one of the largest patient populations in the nation: veterans.
Play this bingo game with your kids to teach them about AI - Artificial intelligence is all around us. But we often don’t notice how much it’s incorporated into the different aspects of our lives. This game challenges you and your kid(s) to notice.
7-Eleven is the next retailer to test cashierless stores - 7-Eleven is the latest retailer to test the “cashierless” store concept, following Amazon’s big push into the market with its Amazon Go convenience stores that use technology, instead of people, to monitor stock levels, track purchases, and process payments.
A new model and dataset for long-range memory - This blog introduces a new long-range memory model, the Compressive Transformer, alongside a new benchmark for book-level language modelling, PG19.
An ever-changing room of Ikea furniture could help AI navigate the world - The Allen Institute wants to crowdsource navigation algorithms by letting researchers turn their robots loose in its physical and virtual apartments.
Boston Dynamics’ dog robot Spot is going to patrol an oil rig in Norway - Spot, a four-legged robot made by Boston Dynamics, has got one of its first real-world assignments: it will be start being used by oil and gas firm Aker BP to patrol a rig in the Skarv field in the Norwegian Sea this year.
Viziblezone taps AI and smartphone data to alert drivers to pedestrians - The World Health Organization reports that 1.5 million people were killed in road accidents in 2018, of which more than half were pedestrians and cyclists. Using a combination of AI and machine learning models, smartphones, and commodity car hardware, Viziblezone is capable of alerting drivers about potential nearby hazards while helping them to avoid accidents.
MIT CSAIL’s AI corrects outdated Wikipedia articles - The English Wikipedia has over 6 million articles, and the combined Wikipedias for all other languages contain over 28 billion words in 52 million articles in 309 languages. In search of an autonomous solution to Wikipedia’s need for constant, large-scale pruning, researchers at MIT have developed an system that uses AI and machine learning to address inconsistencies in Wikipedia articles.
Sony Envisions an AI-Fueled World, From Kitchen Bots to Games - Hiroaki Kitano, a research scientist at Sony, is leading a new effort to infuse cutting-edge AI across the company. The Japanese giant believes AI will create smarter cameras, more cunning videogame characters, and even the first helpful kitchen robots.
Intuition Robotics raises $36 million to bring AI companions to everyone - Israeli robotics startup Intuition Robotics has raised $36 million in a series B round of funding for its mission to create “social companion” robots and related technologies. The company’s initial focus is on reducing loneliness and isolation in elderly people.
This Fake Travis Scott Song Created By Artificial Intelligence Sounds Almost Like The Real Thing - Recent advances in machine learning have allowed musicians like Holly Herndon to use artificial intelligence programs while creating their music. Recently, digital agency space150 pushed the technology to its limits by creating an entirely AI-generated song modeled after Travis Scott’s music. While the track features many idiosyncrasies of Scott’s style, its nonsensical rhymes indicate that machine learning hasn’t surpassed human ability quite yet.
Learning to See Transparent Objects - To enable machines to better sense transparent surfaces, Google AI teamed up with researchers from Synthesis AI and Columbia University to develop ClearGrasp, a machine learning algorithm that is capable of estimating accurate 3D data of transparent objects from RGB-D images.
Boston Dynamics’ Spot robot is surprisingly good at pulling a rickshaw - Boston Dynamics has always talked about Spot’s ability to go places too dangerous for humans, whether it’s checking for gas leaks or disposing of bombs and other hazardous materials. But now, former MythBusters host Adam Savage has developed a new use for the expensive robodog: pulling him around in a homemade rickshaw.
How AI is stopping the next great flu before it starts - Immune systems across the globe have been working overtime this winter as a devastating flu season has taken hold. With numerous hospitalizations and deaths, along with the alarming spread of the coronavirus, there may be some good news on the horizon. In the near future, AI-augmented drug development could help produce vaccines and treatments fast enough to halt the spread of deadly viruses before they mutate into global pandemics.

Concerns & Hype

The controversial facial recognition tech from Clearview AI is also being used to identify child victims of sexual abuse - ClearView AI, a software startup that scraped billions of images from major web services, is selling its facial identification tool to law enforcement agencies across the United States. The tool helps identify crime suspects by matching unknown faces with publicly available photos. While ClearView AI has faced major criticism for taking its images without permission from major services, its service is also being used to help identify child victims of abuse.
A popular self-driving car dataset is missing labels for hundreds of pedestrians - The advent of self-driving cars is a particularly hyped area where machine learning is poised to bring about great societal change. But a poorly trained self-driving car can lead to fatalities–that’s why Roboflow was concerned to discover that a popular Udacity self-driving car dataset contains critical errors and omissions.
It’s 2020. Where are our self-driving cars? - In the mid 2010s, new sources from the Guardian to Business Insider predicted that self-driving cars would be a major presence–possibly with millions on the road–by 2020. But the future has arrived, and self-driving cars haven’t. Despite extraordinary efforts from many leading names and tech, fully autonomous cars remain out of reach except in special trial programs.

Analysis & Policy

The White House wants to spend hundreds of millions more on AI research - The White House is pumping hundreds of millions more dollars into artificial intelligence research, bumping funding for AI research at the Defense Advanced Research Projects Agency and the National Science Foundation, among other departments. These boosts come in the wake of security experts’ pressure on the Trump administration due to concerns that the US may fall behind China in the race to build next-gen technologies.

Expert Opinions & Discussion within the field

The Future of Deep Learning Is Unsupervised, AI Pioneers Say - Machines can perform well at recognizing images and understanding language when humans spoon-feed them labeled data. But, according to recent Turing Award winners Yann Lecun, Geoffrey Hinton, and Yoshua Bengio, the next step in AI advancement will require the technology to figure out how to learn on its own.
AI on steroids: Much bigger neural nets to come with new hardware, say Bengio, Hinton, and LeCun - According to the three Turing Award winners and deep learning pioneers, new hardware designed for deep learning will allow for much larger neural networks, and in addition to allow researchers to tinker more with the transformation built into such networks.
Google Head of Ethical AI Research on Data Biases and Ethics - Margaret (Meg) Mitchell, Co-Head of Ethical Research Group at Google AI, addresses data biases, algorithms, regulation, and more.