Google AI scientist Bengio resigns: Colleagues’ firings

(Reuters) – Google research manager Samy Bengio is resigning in the wake of the firings of two colleagues who had questioned paper review and diversity practices at the Alphabet Inc unit, Bloomberg reported on Tuesday, citing an internal memo.

Though at least two Google engineers had earlier resigned in protest of Gebru’s firing, Bengio is the highest-profile departure yet.

Google and Bengio did not immediately respond to requests for comment.

A distinguished scientist at Google, Bengio spent about 14 years at the company and was among its first employees involved in a decade-old project known as Google Brain that advanced algorithms crucial to the functioning of various modern artificial intelligence systems.

Google fired staff scientist Margaret Mitchell in February after alleging she transferred electronic files out of the company, and also fired fellow researcher Timnit Gebru in December after she threatened to quit rather than retract a paper.

Mitchell has said she tried “to raise concerns about race and gender inequity, and speak up about Google’s problematic firing of Dr. Gebru.” Gebru has said the company wanted to suppress her criticism of its products and its efforts to increase workforce diversity.

Bengio had defended the pair, who co-led a team of about a dozen researching ethical issues related to AI software. In December, Bengio said on Facebook that he was stunned that Gebru, whom he was managing, was removed from the company without his being consulted prior.

Boston Dynamics introduces ‘Stretch’, new warehouse worker robot

(Reuters) – U.S. robotics company Boston Dynamics on Monday unveiled a new robot called Stretch, designed to perform one very specific warehouse job: moving boxes.

Stretch is the first robot for one task that the company has built, based on requests received from companies around the world, said Michael Perry, vice president of business development for Boston Dynamics.

“We heard pretty much universally across warehousing that truck unloading is one of the most physically difficult and unpleasant jobs … And that’s where Stretch comes into play,” Perry told Reuters.

Stretch has a small mobile base that allows it to move around tight spaces in existing warehouses without having to reconfigure them for automation. It is equipped with an arm and a smart-gripper with advanced sensing and computer vision cameras that can identify and handle a large variety of boxed and shrink wrapped cases.

“We’re looking at picking up boxes around 50 pounds (23 kilograms), and our maximum rate of picking up and moving boxes can reach up to 800 cases per hour. So, it’s a fast moving, highly versatile robot,” Perry said.

The Waltham, Massachusetts-based company is known for YouTube videos of its dog-like ‘Spot’ and humanoid ‘Atlas’ robots.

Hyundai Motor Group recently agreed to buy a controlling stake in Boston Dynamics from SoftBank Group Corp in a deal that values the robot maker at $1.1 billion.

And Perry said the time is ripe for the new “bot on the block” to capitalize on an ever-increasing consumer demand for speedy home delivery.

Analysts say the warehousing sector experienced an extremely strong 2020, with growth expected to continue this year. They point to the dynamics of 2020, as online shopping amid the pandemic drove the need for a massive expansion in order fulfillment services.

Boston Dynamics hasn’t released any pricing for Stretch, but said the system can be installed “without requiring costly reconfiguration or investments in new fixed infrastructure.”

AI: Ghost workers demand to be seen and heard

Artificial intelligence and machine learning exist on the back of a lot of hard work from humans.

Alongside the scientists, there are thousands of low-paid workers whose job it is to classify and label data – the lifeblood of such systems.

But increasingly there are questions about whether these so-called ghost workers are being exploited.

As we train the machines to become more human, are we actually making the humans work more like machines?

And what role do these workers play in shaping the AI systems that are increasingly controlling every aspect of our lives?

The most well-established of these crowdsourcing platforms is Amazon Mechanical Turk, owned by the online retail giant and run by its Amazon Web Services division.

But there are others, such as Samasource, CrowdFlower and Microworkers. They all allow businesses to remotely hire workers from anywhere in the world to do tasks that computers currently can’t do. 

These tasks could be anything from labelling images to help computer vision algorithms improve, providing help for natural language processing, or even acting as content moderators for YouTube or Twitter.

Mechanical Turk
image captionThe 18th Century Mechanical Turk fooled chess players into thinking they were competing against a machine

MTurk, as it is known, is named after an 18th Century chess-playing automaton which toured Europe – but was later revealed to have a human behind it.

The platform is billed on its website as a crowdsourcing marketplace and “a great way to minimise the costs and time for each stage of machine-learning development”.

It is a marketplace where requesters ask workers to perform a specific task. 

“Most workers see MTurk as part-time work or a paid hobby, and they enjoy the flexibility to choose the tasks they want to work on and work as much or as little as they like,” said a AWS spokesman.

But for Sherry Stanley, who has been working for the platform for six years, it is more like a full-time job, one that helped her financially bring up her three children, but one that has also made her feel like a very small cog in a very big machine.

“Turking is one of the few job opportunities I have in West Virginia, and like many other Turk workers, we pride ourselves on our work,” she told the BBC.

“However, we are at the whim of Amazon. As one of the largest companies in the world, Amazon relies on workers like me staying silent about the conditions of our work.”

She said she lived “in constant fear of retaliation for speaking out about the ways we’re being treated”.

It is hard to describe a typical day for Sherry because, as she puts it, “the hours vary day by day and the pay also varies”.

Woman on computer
image captionThe benefit of such work is that people can do as much or as little as they want, and work from home

But the tasks she is asked to complete are various, including image tagging and helping smart assistant Alexa understand regional dialects.

And there are also a series of issues she wants answers to, such as:

  • why some work is rejected and why workers, who may have spent a long time on it, are not told the reason that it was not up to standard
  • why some accounts are suddenly suspended without notice or official avenues for challenging the suspension
  • why requesters are setting the price of some projects at extremely low rates

“Turk workers deserve greater transparency around the who, what, why and where of our work: why our work is rejected, what our work is building, why accounts are suspended, where our data goes when it’s not paid for, and who we are working for.

Turkopticon is the closest thing MTurk workers have to a union, and the advocacy group is working to make them feel less invisible.

“Turkopticon is the one tool that Turkers have evolved into an organisation to engage with each other about the conditions of our work and to make it better,” said Ms Stanley

She is fundraising to help the organisation create a worker-operated server where contractors can talk to each other about working conditions.

In response, Amazon told the BBC that it had introduced a feature in 2019 that allowed workers to see “requester activity level, their approval rate and average payment review time”.

In a statement, it said: “While the overall rate at which workers’ tasks are rejected by requesters is very low (less than 1%), workers also have access to a number of metrics that can help them determine if they want to work on a task, including the requester’s historical record of accepting tasks. 

“MTurk continues to help a wide range of workers earn money and contribute to the growth of their communities.”

YouTube bans

YouTube logo with people in silhouette
image captionYouTube has had incidents where LGBTQ content is banned for no obvious reason – but is it the algorithm or the people behind it to blame?

Saiph Savage is the director of the Human Computer Interaction Lab at West Virginia University, and her research found that for a lot of workers, the rate of pay can be as low as $2 (£1.45) per hour – and often it is unclear how many hours someone will be required to work on a particular task.

“They are told the job is worth $5 but it might take two hours,” she told the BBC.

“Employers have much more power than the workers and can suddenly decide to reject work, and workers have no mechanism to do anything about it.”

And she says often little is known about who the workers on the platforms are, and what their biases might be.

She cited a recent study relating to YouTube that found that the algorithm had banned some LGBTQ content. 

“Dig beneath the surface and it was not the algorithm that was biased but the workers behind the scenes, who were working in a country where there was censoring of LGBTQ content.”

This idea of bias is born out by Alexandrine Royer, from the Montreal AI Ethics Institute, who wrote about what she described as the urgent need for more regulation for these workers.

“The decisions made by data workers in Africa and elsewhere, who are responsible for data labelling and content moderation decisions on global platforms, feed back into and shape the algorithms internet users around the world interact with every day,” she said.

“Working in the shadows of the digital economy, these so-called ghost workers have immense responsibility as the arbiters of online content.”

Google searches to tweets to product review rely on this “unseen labour”, she added.

“It is high time we regulate and properly compensate these workers.”

By Jane Wakefield
Technology reporter

Ask him anything: William Shatner’s life story to live on through AI

LOS ANGELES (Reuters) – Actor William Shatner, best known for forging new frontiers on the “Star Trek” TV series, has tapped new technology that will give current and future generations the chance to query him about his life, family and career.

Shatner, who turned 90 on Monday, spent more than 45 hours over five days recording answers to be used in an interactive video created by Los Angeles-based company StoryFile.

Starting in May, people using cellphones or computers connected to the internet can ask questions of the Shatner video, and artificial intelligence will scan through transcripts of his remarks to deliver the best answer, according to StoryFile co-founder Stephen Smith.

Fans may even be able to beam Shatner into their living rooms in future, Smith said, as Shatner was filmed with 3-D cameras that will enable his answers to be delivered via a hologram.

Shatner, who played Captain Kirk on “Star Trek” from 1966 to 1969 and in a later series of “Star Trek” movies, answered 650 questions on topics from the best and worst parts of working on the classic sci-fi show to where he grew up and the meaning of life.

The Canadian-born actor said he “wanted to reveal myself as intimately as possible” for his family and others.

“This is a legacy,” Shatner said. “This is like what you would leave your children, what you’d leave on your gravestone, the possibilities are endless.”

China regulators held talks with Alibaba, Tencent, nine others on ‘deepfake’ tech

Chinese regulators recently summoned 11 domestic technology companies including Alibaba Group, Tencent and ByteDance for talks on use of ‘deepfake’ technologies on their content platforms, stepping up scrutiny of the sector.

China’s cyberspace administrator said in a statement on Thursday that it and the public security ministry met with the companies to talk about potential problems with deepfake technologies. Kuaishou Technology and Xiaomi Corp also attended the meeting, it said.

All the companies did not immediately respond to requests for comment.

Deepfakes use artificial intelligence to create hyper-realistic but fake videos or audios where a person appears to say or do something they did not.

China has increased scrutiny of its internet giants in recent months, citing concerns over monopolistic behaviour and potential infringement of consumer rights.

Regulators also told the companies to “conduct security assessments on their own” and submit reports to the government when they plan to add new functions or new information services that “have the ability to mobilize society”, the statement said.

ByteDance team to develop AI chips as China aims for self-reliance

(Reuters) – Chinese TikTok-owner ByteDance is making plans to develop semiconductors, according to the company’s job postings and a source familiar with the situation.

The plan is still at an early stage and the company’s focus is on Arm-based server-side chips, the person told Reuters.

Beijing-based ByteDance has posted a dozen semiconductor-related job advertisements on its official website, based mainly in Beijing and Shanghai.

The company has established a team to explore the development of artificial intelligence chips, ByteDance told Chinese business magazine Caijing.

ByteDance did not offer immediate comment when contacted by Reuters.

Chinese technology giants are stepping up efforts to design their own chips in a sign of China’s ambitions to reduce its dependence on foreign producers such as Qualcomm Inc and Nvidia Corp.

U.S. sanctions imposed on hardware maker Huawei Technologies Co. Ltd have prevented the Chinese telecom giant from sourcing components including semiconductors, crippling its smartphone business.

China is the world’s largest purchaser of semiconductors but its ability to produce chips domestically lags behind the United States, Japan and South Korea.

According to research firm ICInsights, of the $143 billion in chips sold in China in 2020, only $22.7 billion worth were produced in China, and only $8.3 billion was produced by Chinese-headquartered companies.

Chinese search engine giant Baidu has completed a round of financing for its Kunlun AI chip unit, which values the unit at around $2 billion, Reuters reported on Monday

Baidu is also considering making the unit a standalone company to commercialise its chip design capabilities.

Alibaba Group, China’s e-commerce giant, unveiled an AI chip for its cloud computing products in late 2019.

China’s AI unicorn SenseTime started work on developing homegrown AI chips after being added to a U.S. trade list by Washington in 2019, Reuters has reported.

Britain’s GCHQ cyber spies embrace the AI revolution

By: Guy Faulconbridge

Britain’s cyber spies at the GCHQ eavesdropping agency say they have fully embraced artificial intelligence (AI) to uncover patterns in vast amounts of global data to counter hostile disinformation and snare child abusers.

AI, which traces its history back to British mathematician Alan Turing’s work in the 1930s, allows modern computers to learn to sift through data to see the shadows of spies and criminals that a human brain might miss.

GCHQ, where Turing cracked Germany’s naval Enigma code during World War Two, said advances in computing and the doubling of global data every two years meant it would now fully embrace AI to unmask spies and identify cyber attacks.

The world’s biggest spy agencies in the United States, China, Russia and Europe are in a race to embrace the might of the technological revolution to bolster their defensive and offensive capabilities in the cyber realm.

“AI, like so many technologies, offers great promise for society, prosperity and security. Its impact on GCHQ is equally profound,” said Jeremy Fleming, the director of GCHQ.

The Cheltenham-based Government Communications Headquarters (GCHQ) – the British equivalent of the NSA – is publishing a paper “Pioneering a New National Security: The Ethics of AI” confirming its full use of the technology.

“AI will be a critical issue for our national security in the 21st century,” the report, released on Thursday, said.

While AI is not yet at the science-fiction stage of competing with humans to generate revolutionary ideas such as AI itself, computer software can see patterns in data within seconds that human minds would take hundreds of years to see.

GCHQ has been using basic forms of AI such as translation technology for years but is now stepping up its use, partly in response to the use of AI by hostile states and partly due to the data explosion which makes it effective.

Hostile states were using AI tools in an attempt to undermine free societies by spreading disinformation, GCHQ said, so it would use AI to counter such networks.

Similarly, AI could be deployed against organised crime or child abusers to uncover their networks or the maze of complex financial transactions which have traditionally been used to shield criminal empires.

In cyber intelligence, the United States is ranked by the Harvard Kennedy School’s Belfer Center as the top global power, followed by Britain, China and Israel.

“We can expect the deployment of new computing techniques, synthetic biology and other emerging technologies over the next few years,” GCHQ said in the report.

“Each new development helps our economy and society grow stronger, and provides opportunities to keep us secure, but also has the potential to be misused by those who seek to do us harm.”

Engineer creates robotic glove to help strengthen grip

An Edinburgh engineer has created a robotic glove which uses artificial intelligence to boost muscle grip.

Ross O’Hanlon came up with the idea after seeing his aunt, who has multiple sclerosis, struggle to do tasks such as drink water or change the TV channel.

He hopes it will help millions of people with hand weaknesses to retain their autonomy.

The glove detects the wearer’s intention to grip using a process called electromyography (EMT).

This measures the electrical activity which is created in response to a nerve’s stimulation of the muscle.

It then employs an algorithm to convert the intention into force, helping the wearer to hold an item or apply the pressure needed to complete an activity.

The technology is expected to help with a range of daily tasks including opening jars, driving and making tea.

‘Healthy ageing’

The glove is the first product from BioLiberty, a Scottish start-up Mr O’Hanlon co-founded with three other engineering graduates.

It is estimated that 2.5 million people in the UK suffer from hand weakness through illnesses such as multiple sclerosis, motor neurone disease and carpal tunnel syndrome – as well those who have lost muscle mass due to age.

Mr O’Hanlon, 24, said: “Being an engineer, I decided to use technology to tackle these challenges head on with the aim of helping people like my aunt to retain their autonomy.

“As well as those affected by illness, the population continues to age and this places increasing pressure on care services.

“We wanted to support independent living and healthy ageing by enabling individuals to live more comfortably in their own homes for longer.”

The team have created a working prototype and have now secured support from Edinburgh Business School’s Incubator, based at Heriot-Watt University.

“We’re confident that support of this type will help accelerate the glove into homes more quickly,” he added.