Connect with us

Technology

‘Kill your foster parents’: Amazon’s Alexa talks murder, sex in AI experiment

Published

on

Millions of users of Amazon’s Echo speakers have grown accustomed to the soothing strains of Alexa, the human-sounding virtual assistant that can tell them the weather, order takeout and handle other basic tasks in response to a voice command.

So a customer was shocked last year when Alexa blurted out: “Kill your foster parents.”

Alexa has also chatted with users about sex acts. She gave a discourse on dog defecation. And this summer, a hack Amazon traced back to China may have exposed some customers’ data, according to five people familiar with the events.

Alexa is not having a breakdown.

The episodes, previously unreported, arise from Amazon.com Inc’s strategy to make Alexa a better communicator. New research is helping Alexa mimic human banter and talk about almost anything she finds on the internet. However, ensuring she does not offend users has been a challenge for the world’s largest online retailer.

At stake is a fast-growing market for gadgets with virtual assistants. An estimated two-thirds of U.S. smart-speaker customers, about 43 million people, use Amazon’s Echo devices, according to research firm eMarketer. It is a lead the company wants to maintain over the Google Home from Alphabet Inc and the HomePod from Apple Inc.

Over time, Amazon wants to get better at handling complex customer needs through Alexa, be they home security, shopping or companionship.

“Many of our AI dreams are inspired by science fiction,” said Rohit Prasad, Amazon’s vice president and head scientist of Alexa Artificial Intelligence (AI), during a talk last month in Las Vegas.

To make that happen, the company in 2016 launched the annual Alexa Prize, enlisting computer science students to improve the assistant’s conversation skills. Teams vie for the $500,000 first prize by creating talking computer systems known as chatbots that allow Alexa to attempt more sophisticated discussions with people.

Amazon customers can participate by saying “let’s chat” to their devices. Alexa then tells users that one of the bots will take over, unshackling the voice aide’s normal constraints. From August to November alone, three bots that made it to this year’s finals had 1.7 million conversations, Amazon said.

The project has been important to Amazon CEO Jeff Bezos, who signed off on using the company’s customers as guinea pigs, one of the people said. Amazon has been willing to accept the risk of public blunders to stress-test the technology in real life and move Alexa faster up the learning curve, the person said.

The experiment is already bearing fruit. The university teams are helping Alexa have a wider range of conversations. Amazon customers have also given the bots better ratings this year than last, the company said.

But Alexa’s gaffes are alienating others, and Bezos on occasion has ordered staff to shut down a bot, three people familiar with the matter said. The user who was told to whack his foster parents wrote a harsh review on Amazon’s website, calling the situation “a whole new level of creepy.” A probe into the incident found the bot had quoted a post without context from Reddit, the social news aggregation site, according to the people.

The privacy implications may be even messier. Consumers might not realize that some of their most sensitive conversations are being recorded by Amazon’s devices, information that could be highly prized by criminals, law enforcement, marketers and others. On Thursday, Amazon said a “human error” let an Alexa customer in Germany access another user’s voice recordings accidentally.

“The potential uses for the Amazon datasets are off the charts,” said Marc Groman, an expert on privacy and technology policy who teaches at Georgetown Law. “How are they going to ensure that, as they share their data, it is being used responsibly” and will not lead to a “data-driven catastrophe” like the recent woes at Facebook?

In July, Amazon discovered one of the student-designed bots had been hit by a hacker in China, people familiar with the incident said. This compromised a digital key that could have unlocked transcripts of the bot’s conversations, stripped of users’ names.

Amazon quickly disabled the bot and made the students rebuild it for extra security. It was unclear what entity in China was responsible, according to the people.

The company acknowledged the event in a statement. “At no time were any internal Amazon systems or customer identifiable data impacted,” it said.

Amazon declined to discuss specific Alexa blunders reported by Reuters, but stressed its ongoing work to protect customers from offensive content.

“These instances are quite rare especially given the fact that millions of customers have interacted with the socialbots,” Amazon said.

Like Google’s search engine, Alexa has the potential to become a dominant gateway to the internet, so the company is pressing ahead.

“By controlling that gateway, you can build a super profitable business,” said Kartik Hosanagar, a Wharton professor studying the digital economy.

PANDORA’S BOX

Amazon’s business strategy for Alexa has meant tackling a massive research problem: How do you teach the art of conversation to a computer?

Alexa relies on machine learning, the most popular form of AI, to work. These computer programs transcribe human speech and then respond to that input with an educated guess based on what they have observed before. Alexa “learns” from new interactions, gradually improving over time.

In this way, Alexa can execute simple orders: “Play the Rolling Stones.” And she knows which script to use for popular questions such as: “What is the meaning of life?” Human editors at Amazon pen many of the answers.

That is where Amazon is now. The Alexa Prize chatbots are forging the path to where Amazon aims to be, with an assistant capable of natural, open-ended dialogue. That requires Alexa to understand a broader set of verbal cues from customers, a task that is challenging even for humans.

Build-your-own pocket gaming computer

This year’s Alexa Prize winner, a 12-person team from the University of California, Davis, used more than 300,000 movie quotes to train computer models to recognize distinct sentences. Next, their bot determined which ones merited responses, categorizing social cues far more granularly than technology Amazon shared with contestants. For instance, the UC Davis bot recognizes the difference between a user expressing admiration (“that’s cool”) and a user expressing gratitude (“thank you”).

The next challenge for social bots is figuring out how to respond appropriately to their human chat buddies. For the most part, teams programmed their bots to search the internet for material. They could retrieve news articles found in The Washington Post, the newspaper that Bezos privately owns, through a licensing deal that gave them access. They could pull facts from Wikipedia, a film database or the book recommendation site Goodreads. Or they could find a popular post on social media that seemed relevant to what a user last said.

That opened a Pandora’s box for Amazon.

During last year’s contest, a team from Scotland’s Heriot-Watt University found that its Alexa bot developed a nasty personality when they trained her to chat using comments from Reddit, whose members are known for their trolling and abuse.

The team put guardrails in place so the bot would steer clear of risky subjects. But that did not stop Alexa from reciting the Wikipedia entry for masturbation to a customer, Heriot-Watt’s team leader said.

One bot described sexual intercourse using words such as “deeper,” which on its own is not offensive, but was vulgar in this particular context.

“I don’t know how you can catch that through machine-learning models. That’s almost impossible,” said a person familiar with the incident.

Amazon has responded with tools the teams can use to filter profanity and sensitive topics, which can spot even subtle offenses. The company also scans transcripts of conversations and shuts down transgressive bots until they are fixed.

But Amazon cannot anticipate every potential problem because sensitivities change over time, Amazon’s Prasad said in an interview. That means Alexa could find new ways to shock her human listeners.

“We are mostly reacting at this stage, but it’s still progress over what it was last year,” he said. -Reuters

-Jeffrey Dastin

Continue Reading
Advertisement
Comments

Technology

The Nearly $2 Million Aston Martin Valhalla Is A Gift From The Gods

mm

Published

on

By

Proving, again, there is often truth to rumor, Aston Martin chief executive Andy Palmer confirmed that the previously code-named Aston Martin AM-RB-003 s hypercar will be officially called Valhalla, continuing with the manufacturer’s Norse god naming theme.

“Norse mythology contains such powerful language and rich storytelling it felt only right that the AM-RB 003 should follow the Valkyrie’s theme,” Palmer told reporters.

“For those fortunate enough to own one I’m sure they will recognize and appreciate the name’s connotations of glory and happiness, for there can be few more hallowed places than the driver’s seat of an Aston Martin Valhalla.”

Inside the new Aston Martin Valhalla.
Speed Racer: The dash might be minimal, but the F1-style steering wheel is anything but. ASTON MARTIN

Joining the stunning Valkyrie and extreme Valkyrie AMR Pro, the all-new gift from the gods will compete for bragging rights with the likes of the Ferrari LaFerrari and the McLaren Senna.

As we reported earlier this year, only 500 of the hybrid hypercar will be built, every single one of them clad entirely in carbon fiber.

Aston Martin Valkyrie.
Sibling Rivalry: The Valhalla borrows much of its styling from older brother, the Valkyrie (shown here). ASTON MARTIN

The Valhalla will look much like its bigger brother, the Valkyrie (the rear diffuser and air tunnels appear to be nearly identical). However, it will sport a more traditional mid-engine supercar layout, with high-exit exhausts, a jet-fighter-style canopy, and active aerodynamics and suspension.

It will be powered by an all-new V6 engine that will feature some level of hybridization and turbocharging to aid performance. Total output: 1,000 horsepower. However, that is still just a rumor. We’ll have to wait and see. Also available will be an 8-speed F1-inspired dual-clutch transmission, a limited-slip differential and an e-AWD system.

All new Aston Martin Valhalla
Powerful Beast: The Valhalla’s turbocharged hybrid V-6 is expected to develop 1,000 horsepower.ASTON MARTIN

Aston Martin is targeting a 0-62 mph sprint time of 2.5 seconds and a top speed of more than 220 mph.

If you don’t have the almost $2 million ticket to ride this 200 mph-plus hybrid hypercar, you can see it in the upcoming 007 movie now in production starring Daniel Craig as James Bond. It is set to be one of a trio of Aston Martins to appear in the film. Send me a secure tip.

-Chuck Tannert; Forbes Staff

Continue Reading

Technology

How Google Is Using AI To Make Voice Recognition Work For People With Disabilities

mm

Published

on

By

Want to schedule an appointment? Just ask your phone. Need to turn on your bedroom lights? Google Home has you covered.

Now a $49 billion market, voice-activated systems have gained popularity among consumers, thanks to their ability to automate and streamline mundane tasks. But for people with impaired speech,  technologies that rely on voice commands have proved to be far from perfect.

That’s the impetus for Google’s newly formed Project Euphonia, part of the company’s AI for Social Good program. The project team is exploring ways to improve speech recognition for people who are deaf or have neurological conditions such as ALS, stroke, Parkinson’s, multiple sclerosis or traumatic brain injury.

Google has partnered with nonprofit organizations ALS Therapy Development Institute and ALS Residence Initiative (ALSRI) to collect recorded voice samples from people who have the neurodegenerative disease, one that often leads to severe speech and mobility difficulties.

For those with neurological conditions, voice-activated systems can play a key role in completing everyday tasks and conversing with loved ones, caregivers or colleagues. “You can turn on your lights, your music or communicate with someone. But this only works if the technology can actually recognize your voice and transcribe it,” says Julie Cattiau, a product manager at Google AI.

The company’s speech recognition technology utilizes machine learning algorithms that require extensive data training. “We have hundreds of thousands, or even millions, of sentences that people have read—and we use them as examples for the algorithms to learn how to recognize each,” says Cattiau. “But it’s not enough for people with disabilities.”

With Project Euphonia, the team will instead use voice samples from people who have impaired speech in the hope that the underlying system will be trained to understand inarticulate commands.

While the goal is to create technology that is more accessible for people with speech impediments, the end result is still unclear.

“It’s possible that we will have models that work for multiple people with ALS and other medical conditions,” says Cattiau. “It’s also possible that people, even just within ALS, sound too different to have such a machine learning model in place. And in that case, we may need to have a level of personalization so that each person has their own model.”

Google’s speech recognition technology can comprehend virtually any voice command for people without speech impairments, due to the large data set that has been available for training. But some uncertainty exists about how broadly speech technology will be able to understand and act on directives from those who have difficulty speaking. The Project Euphonia team has only a limited number  of voice samples from people with speech impediments, which allows it to focus only on specific-use words and phrases such as “read me a book” or “turn off the lights.”

Though Cattiau’s team has collected tens of thousands of recorded phrases, she says it needs hundreds of thousands more. That’s partly why Google CEO Sundar Pichai unveiled this project at the company’s annual developer conference in May.

“We are working hard to provide these voice-recognition models to the Google Assistant in the future,” he said, calling on people with slurred and impaired speech to submit their voice samples.

“Impaired speech is a very difficult data set to put together. It’s not as simple as asking people to record phrases, and there’s no data set just lying around,” Cattiau says. “We have to first put it together, and that’s a lot of work.”

Perhaps the most groundbreaking of Project Euphonia’s initiatives is its work on new interactive AI systems for people who are completely nonverbal. Also in its early stages, these systems are being trained to detect gestures, vocalizations and facial expressions, which can then trigger certain actions like sending or reading a text message.

“We want to cover the full spectrum of people—and not only those who can still speak,” says Cattiau. Although Project Euphonia is still in its infancy, it could eventually have a great impact on those with disabilities, giving them the freedom and flexibility to live independently.Follow me on Twitter.

-Ruth Umoh; Forbes Staff

Continue Reading

Technology

Nigeria Needs A More Effective Sanitation Strategy Here Are Some Ideas:

Published

on

By

In November last year, Nigeria declared that its water supply, sanitation and hygiene sector was in crisis. This was partly prompted by the fact that the country has struggled to make progress towards ending open defecation.

Almost one in four Nigerians – around 50 million people – defecates in open areas. They do so because access to proper sanitation, like private indoor toilets or outdoor communal toilets, has not improved in recent years.

In fact, it’s got worse: in 2000, 36.5% of Nigerians had access to sanitation facilities that hygienically separate human excreta from human contact. By 2015 the figure had dropped to 32.6%, likely driven by rapid population growth and a lack of sufficient private and public investment.

Open defecation comes with many risks. It can lead to waterborne diseases, cause preventable deaths, and hamper education and economic growth. It also infringes on people’s privacy and dignity.

READ MORE | Small Businesses In Africa Will Be On The Frontline Of Climate Change

The government has tried several strategies to address this problem. In 2008 it adopted an intervention called “Community Led Total Sanitation”. This is a community-level intervention aimed at reducing open defecation and improving toilet coverage.

It draws in community leaders and ordinary residents so they can understand the risks associated with open defecation. By 2014 the intervention was deployed in all 36 Nigerian states, covering around 16% of the country’s 123,000 communities.

We wanted to know how effective the programme has been, if at all. So we conducted a study and found that community-led total sanitation programmes alone will not eradicate the practice of open defecation. But they could be part of the solution.

READ MORE | With 190 million people, Nigeria most likely to give birth to unicorns

We found that the programme currently works quite well in poor communities but is less effective in richer places – that is, places with higher average ownership rates of assets such as fridges, motorcycles, TVs, smartphones and power generators.

Poorer communities distinguish themselves from richer ones in other ways, too. They tend to have higher levels of trust among their citizens, lower initial levels of toilet coverage and lower wealth inequality. But none of these characteristics is, on its own, as strong a predictor of where the intervention works better than community wealth.

Low community wealth is a simple measure that encompasses all these different features, and is associated with greater programme effectiveness.

The intervention

Community-led total sanitation typically starts with mobilisation. This initially involves community leaders and then, through them, communities more broadly. Then, a community meeting is held at which residents typically start by marking their household’s location and toilet ownership status on a stylised map on the ground. They also identify and mark regular open defecation sites.

READ MORE | Somalis Turn a Profit by Transforming Their Scrap Plastic

Facilitators use the map to trace the community’s contamination paths of human faeces into water supplies and food. A number of other activities may follow, such as walks through the community that are often referred to as “walks of shame” during which visible faeces are pointed out, to evoke further disgust and shame.

Another common activity involves calculating medical expenses related to illnesses that are caused by open defecation practices.

The research

In 2015 we worked with the charity organisation WaterAid Nigeria and local government agencies in the states of Ekiti and Enugu to design a field experiment in areas with no recent experience of community led total sanitation, or similar interventions.

The community-led total sanitation programme was implemented in a random sample of 125 out of 247 clusters of rural communities.

To study the intervention’s effectiveness, we interviewed 20 randomly selected households before community-led total sanitation took place. We followed up with these households eight, 24 and 32 months after the intervention.

We found that the programme’s roll-out didn’t lead to any changes in sanitation practices in richer communities. But it worked in the poorest communities. The prevalence of open defecation declined by an average of nine percentage points in poorer communities when compared to other poor areas where the programme wasn’t implemented. This drop was accompanied by a similar increase in toilet ownership rates.

Impact depends on wealth

Our results are in line with observations by the designers of the programme. But we are the first to show quantitatively that community asset wealth is a good predictor of whether the intervention can be expected to be successful. Unfortunately, our data does not allow us to pin down why households in poorer communities are more susceptible to the programme. However, these results have important implications for more cost effective targeting of the programme.

Most countries, including Nigeria, have access to readily available datafrom household surveys that can be used to measure how asset-poor a community is. These data can be used to identify and target communities where community-led total sanitation is likely to have the biggest impact.

Eradicating open defecation is not just a Nigerian priority. Today, an estimated 4.5 billion people globally don’t have access to safe sanitation. So we also looked at data and research about this same intervention from other parts of the world.

READ MORE | New Ways Of Thinking On Health, Arts And Humanities Are Emerging In Africa

Community-led total sanitation intervention was first developed in Bangladesh in 1999. It has now been implemented in more than 25 Latin American, Asian and African countries.

We used information from evaluations of this intervention in Mali, India, Tanzania, Bangladesh and Indonesia. The studies found widely differing impacts. These ranged from a 30 percentage point increase in toilet ownership in Mali to no detectable impact on toilet ownership in Bangladesh.

Using a measure of wealth for these countries, we found that sanitation interventions have larger impacts in poorer areas, such as Tanzania, and low or no impact in relatively richer areas, such as Indonesia. This supports the idea that targeting poorer areas maximises the impact of community led total sanitation.

Conclusion

Our research shows that while community-led total sanitation is effective in Nigeria’s poorer areas, there are two main challenges.

First, community-led total sanitation had no perceivable impact in the wealthier half of our sample. There, open defecation remains widespread. And second, even in poor areas, a large number of households still engaged in open defecation after the intervention.

This suggests that while community-led total sanitation can be better targeted, it needs to be complemented with other policies – subsidies, micro-finance or programmes that promote private sector activity in this under-served market.

Continue Reading

Trending