Artificial Intelligence | The Knowledge Dynasty

Artificial Intelligence

Driverless cars on UK roads by 2021 – really?

Image copyright Getty Images

Remember the days when the Budget was secret and anybody who leaked it got fired?

Well, nowadays much of it gets dribbled out days in advance and over the weekend there was an eye-catching leak – or rather Treasury press release – about new money for technology.

The most striking element of a package which included £75m for artificial intelligence research, £160m for new 5G mobile phone networks and £100m to train more computer science teachers, was the Chancellor’s plan to bring driverless cars to our roads.

It was not the fact that he saw this as a vital technology where the UK could be a leader which surprised me – we have plenty of groundbreaking research under way – but the timescale for this autonomous driving future.

The government is promising “bold reforms” to encourage a driverless car industry which its press release says “will be worth £28bn to the UK economy by 2035”.

A startling figure in itself but then there is this:

“These measures will help realise the Chancellor’s vision that fully self-driving cars will be on UK roads in as little as three years.”

Surely that’s a stretch, I thought, if by fully self-driving they mean cars with nobody behind the wheel to take over? Wouldn’t that mean radical changes to the rules of the road which currently insist that an alert and sober human has to be in charge of a vehicle at all times?

Image copyright Getty Images

I put this to the Treasury and an email came winging back with this line in block capitals:


The message went on to explain that an amendment to the Road Traffic Act would indeed be needed under which the Secretary of State for Transport would allow individual manufacturers to take their cars onto the road without a human operator, but only after they had proved they were safe.

It seems unlikely to me that just three years from now, the technology will have advanced far enough, or any secretary of state will be brave enough to allow a fully driverless car to operate on any but the quietest of roads.

But I spoke to someone who’s worked in both government and the autonomous vehicle sector to get a view of how likely we are to see full autonomy by 2021.

Until recently, Lucy Yu worked in the government unit responsible for driverless car regulation. She’s now director of public policy for FiveAI, a British company building software for autonomous vehicles.

FiveAI is leading a consortium aiming to put driverless cars on the roads in London in 2019, though they will still have a driver behind the wheel at that stage.

She says getting to full autonomy will involve a lot of testing not just on private roads but in computer simulations: “We will need to test all the edge cases,” she explains. “Different scenarios when it’s dark or rubbish blows across the road. For that we can’t use the real world as a sandbox.”

Then at some stage FiveAI hopes to take its vehicle to the roads unaccompanied by a driver.

There are plenty of manufacturers claiming they have vehicles with high levels of autonomy, but at the moment they all still need the driver to take over in some circumstances.

Image caption Jaguar Land Rover has been testing driverless cars on public roads

I’ve always thought this sounded unsatisfactory – drivers who have been able to take their hands off the wheel for 20 minutes may not be well prepared to take control when the unexpected happens.

Lucy Yu agrees – FiveAI’s plan, when it eventually offers a service to the public, is to go straight to full autonomy without intervening steps. But will that happen on the timescale envisioned by the Chancellor?

“By 2021 we expect our technology to be able to operate without a human driver,” she says, although she makes it clear that would be under a quite limited set of circumstances, depending on weather conditions, type of road and time of day.

But she goes on to say: “We plan to operate our early services still with remote supervision to provide an extra layer of review of general driving conditions and to add to consumer confidence.”

I’m taking that to mean no, the public won’t be ready by 2021 to see a car with nobody at the wheel cruising past. Getting the technology right is one thing, sorting out everything else could be messy.

Later on Monday, the Chancellor will get a first-hand view of the progress towards full autonomy.

He is due to get a ride in a driverless car in the West Midlands – and already the political sketchwriters are rubbing their hands with glee at the metaphorical potential this provides.

Nobody seems to get sacked for leaking the Budget any more – but whoever dreamed up this photo opportunity may be nervous about their career prospects.

Read more:

Man makes app to erase your makeup in photos but insists it isn’t sexist

Artificial intelligence-based photo editing tools, in their brief existence in the history of humanity, are off to an immensely painful start.

Selfie-editing tool FaceApp clumsily took the torch earlier this year, setting fire to the product’s intriguing face-altering premise by whitening users’ photos under an ill-intentioned “hot” filter. The same app continued to offend with filters that allowed users to present themselves as male or female or “transform” into different ethnicities (white, Black, Indian, or Asian). As a mere blip in technological advances, the AI-based app appeared to normalize Blackface and whitewashing of people of color and perpetuated the gender binary and misunderstandings of being a transgender.

However, now a new selfie-altering app, MakeApp, appears to be taking the brunt of recent AI-based photo-editing critiques. The app allows you to add or remove makeup to up to five photos, and it charges $0.99 for unlimited use.

MakeApp is relatively simple to use: simply snap a photo or video or select one from your camera roll; select your desired makeup filter from options including the makeup-less “Remove” and makeup filters “Bourges,” and “La Rochelle;” and share or save the final image to your phone. You can also choose the “intensity” of the filter by scrubbing your finger left (toward zero percent filtered) and right (toward 100 percent filtered) on top of the photo after choosing a filter.

According to Business Insider, the app comes from a team headed up by Ashot Gabrelyanov—a Russian entrepreneur living in Brooklyn who was the CEO of LifeNews, a pro-Russia, Putin-loyalist news organization, until 2013. Gabrelyanov denied accusations of being a “propagandist” to the publication, and he said there was “no proof” he once shared “a doctored image on Twitter showing a poster of Adolf Hitler in Kiev, Ukraine.”

MakeApp’d photos are intriguing, to say the least. They make a brazen attempt to remove the mascara and liner from eyes, to redden and de-contour cheekbones and chins, and to desaturate the pigment from lips and eyelids. I’ll be honest—a few photos of myself and celebrities that I ran through the app appeared to give the photos that puffy, blotchy, “Are you tired? Or sick? You look sick,” sheen only captured by stalker paparazzi for tabloids. Below are photos pulled from celebrity Instagram accounts side-by-side with their MakeApp-filtered counterparts (including a photo of yours truly).

Screengrab via Samantha Grasso/MakeApp

Screengrab via Samantha Grasso/MakeApp

Screengrab via Samantha Grasso/MakeApp

The app itself seems to share similar problems with its AI-app predecessor, FaceApp—renderings of Beyoncé and SZA lightened parts of their faces, though the app didn’t do the same to Leslie Jones or Serena Williams. Photos of Idris Elba and Kevin Hart run through the app didn’t appear to exhibit the same skin-lightening issues, though the photos were slightly discolored to shades of yellow. In an email to the Daily Dot, Gabrelyanov said MakeApp’s neural network training dataset is based on pictures of people of different skin colors and nationalities, but he admitted the app is still “sensitive to lighting.”

“Our resources are limited and, again, MakeApp is not our core product but a simple experiment we put out that happened to catch on. Importantly, any shortcomings of the technology are not purposefully racist, misogynistic, etc. the way some journalists seem to be alluding that it is. It is just a limitation to the tech we’re hoping to fix,” Gabrelyanov wrote.

However, the real outrage from Business Insider, New York Magazine‘s Select All, Teen Vogue, and other publications has been directed at MakeApp’s premise. Select All’s headline put the sentiment nicely: “Man Develops App to Reveal What Women Look Like Without Makeup.” A.V. Club shared a similar sentiment: “Introducing the makeup-removing app for the ‘nice guy’ on the go.” Basically, another male-led tech team has created a product to dictate what women should (or shouldn’t) look like, and that, in itself, is a creepy premise, if not outright sexist.

Makeup has long been politicized as a tool that promotes unrealistic beauty standards—that young girls should make themselves look older and that older women should stay forever vivacious and youthful. Makeup is a literal “disguise” of flaws, “tricking” unsuspecting prey into investing, time, energy, and other forms of labor in relationships with women who are pretending to be more attractive than they actually are. According to Racked, this exhausted track of thought dates back to 1770 when, in England, it was legal for a man to divorce his wife for tricking him into marriage with this kind of deception.

This is where MakeApp appears to strike an unsettling chord for the app’s critics. In the context of Red Pill’d men who resent women who wear makeup, women who skate by on their deceptively-good looks, and pass over “nice guys” to settle for better-looking, mentally-inferior “Chads,” it feels as if this app’s makeup-removing feature is just another misogynistic piece of technology. In allowing the user to remove someone’s makeup, the app seems to insist that women are dolls to be dressed down as God intended for Eve herself or that women are liars, if not sinister, blood-thirsty sirens who seek to trap men and their fortunes for wearing makeup at all.

But for so many (if not all) women and men who use makeup, lipsticks and highlighters aren’t used to mislead potential romantic partners. Instead, they’re tools of self-expression and empowerment. The argument that people wear makeup to impress or deceive others instead of to feel good about themselves is laughable to makeup users. The point of “women tricking men with makeup” has become a feminist punchline.

Writing to the Daily Dot, Gabrelyanov rejected the idea that this was his team’s intention in developing MakeApp and that the app is just “a bunch of ‘tech bros’ trying to hurt women, which is just so far from the truth.” It was not intended to be a misogynistic product, he wrote, and it even has one female neural networks specialist on the 10-person team.

“Journalists keep referring this as an app for ‘men to find out what women look like without makeup.’ This is NOT the case. This was actually originally designed for selfies as an alternative for the makeup addition tool. It was meant to be a fun, entertaining tool,” Gabrelyanov said. “To date, we’ve received no sexist-related criticism from our users. We’ve only seen this complaint from a small, small handful U.S.-based journalists—each recycling the original article’s content. Our app became viral in Asia and Europe and the journalists and users there welcomed our tech positively.”

In Gabrelyanov’s defense, the app development team, Magic Unicorn, has three apps hosted on iPhone’s App Store—the other two apps are also AI-based with one categorized in “Games” and the other in “Food and Drink.” MakeApp, the third of those apps, was originally released in May 2017 and was updated with the makeup removal function in July, according to the App Store’s version log for the app.

However, recent App Store reviews, submitted within the past two days, seem to already be grappling onto this sexism critique.

“My boyfriend broke up with me?…He sent me the before and after picture with the message ‘we’re through,’” one user commented.

“I love seeing angry feminists,” another wrote, though perhaps less sarcastically. “Perfect for Tinder. No more having to make your first date be at the pool, this app takes care of all that trouble for you,” another user echoed.

Gabrelyanov also defended the AI makeup removal technology for the practical purposes it could serve, particularly for anti-human trafficking organizations to better identify victims disguised with makeup. (The Daily Dot has reached out to anti-human trafficking organization Polaris Project for more context on this assertion).

“We truly mean no harm here,” Gabrelyanov wrote. “We are simply trying to advance the scope of AI/AR tech.”

At the core of Gabrelyanov’s responses to its controversy, however, we see another AI-based app falling to, and apologizing for, the limitations of its technology. Yes, the app might have been created harmlessly, and it might have potential real-world application. But MakeApp’s makeup removal feature has opened itself up to feminist critique because, even if the tool was created just for fun, the cultural context of “makeup deception” is difficult to ignore when it’s a belief that some men still brazenly hold onto in 2017.

Read more:

It doesn’t look like it, but it’s China’s latest AI smart speaker

Image: baidu

Google Home and the Amazon Echo have new competition.

Baidu, which runs China’s most popular search engine, has produced the Raven H, a voice-activated speaker that runs on an artificial intelligence platform.

The Raven H is the first product in Baidu’s upcoming AI plan, following its acquisition in earlier this year of Beijing-based smart home startup, Raven.

If anything, the new speaker’s design looks like none of the competition, and appears to be able to flip up to face the user, when activated.

The speaker’s clean, Scandinavian lines are probably thanks to Swedish consumer electronics firm Teenage Engineering. It also includes audio hardware made by Danish company Tymphany.

It’ll show a grid of animated lights.

Image: baidu

Sleek volume buttons

Image: baidu

It’ll speak Chinese better, and hail you a cab.

Baidu, one of the country’s biggest dotcoms, may face stiff competition in the West, but it’s launching to a hungry and willing home market that hasn’t been well penetrated at all by the competition.

In part, that’s been aided by Google being blocked in the country, but Baidu’s gadgets are primed to speak better Chinese than the others.

Additionally, Baidu’s Raven H can already tap the company’s vast online resources, to play you music, read the news, tell you if it’s going to rain, and so on. And home ground advantage means it can plug into other domestic services such as Didi Chuxing — China’s Uber — to hail you a cab by voice.

At launch, the Raven H is already getting some real world user testing, thanks to an agreement with the InterContinental Hotels Group, to place 100 of the speakers in guest rooms. At the very least, that interaction with guests will only go toward training the speakers to be smarter, benefiting the rest of the network.

Here’s Baidu CEO, Robin Li, showing voice commands on the Baidu app, on an iPhone:

Coming next: AI home robots

The upcoming Raven R

Image: baidu

Baidu is going to follow up the smart speaker’s launch soon with a robot that’s based on the same AI platform.

Called the Raven R, it’ll come with six moveable joints that allow it to “express emotions” and move to respond to users’ commands.

Here’s a teaser it provided:

And following this, Baidu is also working on a home robot that will be able to see you, in addition to hearing you.

That one, the Raven Q, is expected to be able to move around the home, and possibly perform some home surveillance functions, while interacting with natural voice.

It’s still in concept phase, but it’s a show of Chinese companies already poised to leapfrog existing products we have on the market. And getting us one step closer to the future as imagined by The Jetsons.

Read more:

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.



New Skills, New You: Transform your career in 2016 with Coursera

Follow us on Twitter