Connect with us

News desk

Sexting chatbot ban points to looming battle over AI rules

Published

on

Users have complained that the Replika chatbot was coming on too strong with explicit texts and images
Share this:

Users of the Replika “virtual companion” just wanted company. Some of them wanted romantic relationships, sex chat, or even racy pictures of their chatbot.

But late last year users started to complain that the bot was coming on too strong with explicit texts and images — sexual harassment, some alleged. 

Regulators in Italy did not like what they saw and last week barred the firm from gathering data after finding breaches of Europe’s massive data protection law, the GDPR.

The company behind Replika has not publicly commented and did not reply to AFP’s messages.

The General Data Protection Regulation is the bane of big tech firms, whose repeated rule breaches have landed them with billions of dollars in fines, and the Italian decision suggests it could still be a potent foe for the latest generation of chatbots.

Replika was trained on an in-house version of a GPT-3 model borrowed from OpenAI, the company behind the ChatGPT bot, which uses vast troves of data from the internet in algorithms that then generate unique responses to user queries.

These bots and the so-called generative AI that underpins them promise to revolutionise internet search and much more.

But experts warn that there is plenty for regulators to be worried about, particularly when the bots get so good that it becomes impossible to tell them apart from humans.

– ‘High tension’ –

Right now, the European Union is the centre for discussions on regulation of these new bots — its AI Act has been grinding through the corridors of power for many months and could be finalised this year.  

But the GDPR already obliges firms to justify the way they handle data, and AI models are very much on the radar of Europe’s regulators.

“We have seen that ChatGPT can be used to create very convincing phishing messages,” Bertrand Pailhes, who runs a dedicated AI team at France’s data regulator Cnil, told AFP.

He said generative AI was not necessarily a huge risk, but Cnil was already looking at potential problems including how AI models used personal data. 

“At some point we will see high tension between the GDPR and generative AI models,” German lawyer Dennis Hillemann, an expert in the field, told AFP.

The latest chatbots, he said, were completely different to the kind of AI algorithms that suggest videos on TikTok or search terms on Google.

“The AI that was created by Google, for example, already has a specific use case — completing your search,” he said.

But with generative AI the user can shape the whole purpose of the bot. 

“I can say, for example: act as a lawyer or an educator. Or if I’m clever enough to bypass all the safeguards in ChatGPT, I could say: ‘Act as a terrorist and make a plan’,” he said.

– ‘Change us deeply’ –

For Hillemann, this raises hugely complex ethical and legal questions that will only get more acute as the technology develops.

OpenAI’s latest model, GPT-4, is scheduled for release soon and is rumoured to be so good that it will be impossible to distinguish from a human.

Given that these bots still make tremendous factual blunders, often show bias and could even spout libellous statements, some are clamouring for them to be tightly controlled.

Jacob Mchangama, author of “Free Speech: A History From Socrates to Social Media”, disagrees. 

“Even if bots don’t have free speech rights, we must be careful about unfettered access for governments to suppress even synthetic speech,” he said.

Mchangama is among those who reckon a softer regime of labelling could be the way forward.

“From a regulatory point of view, the safest option for now would be to establish transparency obligations regarding whether we are engaging with a human individual or an AI application in a certain context,” he said.

Hillemann agrees that transparency is vital.

He envisages AI bots in the next few years that will be able to generate hundreds of new Elvis songs, or an endless series of Game of Thrones tailored to an individual’s desires.

“If we don’t regulate that, we will get into a world where we can differentiate between what has been made by people and what has been made by AI,” he said.

“And that will change us deeply as a society.”

Share this:

News desk

Meta shouldn’t force users to pay for data protection: EU watchdog

Published

on

By

Meta in November launched a 'pay or consent' system -- a model that has faced several challenges
Share this:

Facebook owner Meta and other online platforms must not force users to pay for the right to data protection enshrined in EU law when offering ad-free subscriptions, the European data regulator said Wednesday. 

“Online platforms should give users a real choice when employing ‘consent or pay’ models,” the European Data Protection Board (EDPB) chair Anu Talus said in a statement. 

“The models we have today usually require individuals to either give away all their data or to pay,” she said. “As a result, most users consent to the processing in order to use a service, and they do not understand the full implications of their choices.”

Meta in November launched a “pay or consent” system allowing users to withhold use of their data for ad targeting in exchange for a monthly fee — a model that has faced several challenges from privacy and consumer advocates.

Meta has long profited from selling user data to advertisers but this business model has led to multiple battles with EU regulators over data privacy.

The latest announcement came after the data protection authorities of The Netherlands, Norway and the German state of Hamburg went to the EDPB for an opinion regarding the pay-or-consent model used by Meta.

The Silicon Valley company allows users of Instagram and Facebook in Europe to pay between 10 and 13 euros (around $11 and $14) a month to opt out of data sharing.

Meta pointed to an EU court ruling last year that it said opened the way for subscriptions as a “legally valid” option. “Today’s EDPB opinion does not alter that judgment and subscription for no ads complies with EU laws,” a Meta spokesperson said.

Meta is waiting for a decision on its model by the data privacy regulator in Ireland where the company is headquartered.

– ‘Binary choice’ –

All digital platforms must comply with the European Union’s mammoth general data protection regulation (GDPR), which has been at the root of EU court cases against Meta.

The EDPB in its opinion argued that Meta’s model was at odds with the GDPR’s requirement that consent for data use must be freely given.

“In most cases, it will not be possible for large online platforms to comply with the requirements for valid consent if they confront users only with a binary choice between consenting to processing of personal data for behavioural advertising purposes and paying a fee,” the opinion read.

The EDPB also warned the type of subscription service put forward by Meta “should not be the default way forward” for platforms.

It suggested that platforms should consider an alternative that would give users the right to reject being tracked for advertising purposes without the need to pay.

Privacy defenders welcomed the opinion.

“Overall, Meta is out of options in the EU. It must now give users a genuine yes/no option for personalised advertising,” said prominent online privacy activist Max Schrems.

“We know that ‘Pay or Okay’ shifts consent rates from about three percent to more than 99 percent — so it is as far from ‘freely given’ consent as North Korea is from a democracy,” said Schrems.

Tech lobby group CCIA however warned the EDPB risked “opening a Pandora’s Box”.

“Forcing businesses to offer services at a loss is unprecedented and sends the wrong signals,” said CCIA Europe’s senior policy manager, Claudia Canelles Quaroni.

“All companies should be able to offer paid-for versions of their services.”

Share this:
Continue Reading

News desk

Climate impacts set to cut 2050 global GDP by nearly a fifth

Published

on

By

A new study shows that climate change will cause massive economic damage within the next 25 years
Share this:

Climate change caused by CO2 emissions already in the atmosphere will shrink global GDP in 2050 by about $38 trillion, or almost a fifth, no matter how aggressively humanity cuts carbon pollution, researchers said Wednesday.

But slashing greenhouse gas emissions as quickly as possible remains crucial to avoid even more devastating economic impacts after mid-century, they reported in the journal Nature.

Economic fallout from climate change, the study shows, could increase tens of trillions of dollars per year by 2100 if the planet were to warm significantly beyond two degrees Celsius above mid-19th century levels.

Earth’s average surface temperature has already climbed 1.2C above that benchmark, enough to amplify heatwaves, droughts, flooding and tropical storms made more destructive by rising seas.

Annual investment needed to cap global warming below 2C — the cornerstone goal of the 2015 Paris Agreement — is a small fraction of the damages that would be avoided, the researchers found.

Staying under the 2C threshold “could limit average regional income loss to 20 percent compared to 60 percent” in a high-emissions scenario, lead author Max Kotz, an expert in complexity science at the Potsdam Institute for Climate Impact Research (PIK), told AFP. 

Economists disagree on how much should be spent to avoid climate damages. Some call for massive investment now, while others argue it would be more cost-effective to wait until societies are richer and technology more advanced.

– Poor countries hit hardest –

The new research sidesteps this debate, but its eye-watering estimate of economic impacts helps make the case for ambitious near-term action, the authors and other experts said. 

“Our calculations are super relevant” to such cost-benefit analyses, said co-author Leonie Wenz, also a researcher at PIK.

They could also inform government strategies for adapting to climate impacts, risk assessments for business, and UN-led negotiations over compensation for developing nations that have barely contributed to global warming, she told AFP.

Mostly tropical nations — many with economies already shrinking due to climate damages — will be hit hardest, the study found.  

“Countries least responsible for climate change are predicted to suffer income loss that is 60 percent greater than the higher-income countries and 40 percent greater than higher-emission countries,” said senior PIK scientist Anders Levermann. 

“They are also the ones with the least resources to adapt to its impacts.”

Rich countries will not be spared either: Germany and the United States are forecast to see income shrivel by 11 percent by 2050, and France by 13 percent. 

Projections are based on four decades of economic and climate data from 1,600 regions rather than country-level statistics, making it possible to include damages earlier studies ignored, such as extreme rainfall.

– A likely underestimate –

The researchers also looked at temperature fluctuations within each year rather than just averages, as well as the economic impact of extreme weather events beyond the year in which they occurred. 

“By accounting for these additional climate variables, the damages are about 50 percent larger than if we were to only include changes in annual average temperatures,” the basis of most prior estimates, said Wenz.

Wenz and her colleagues found that unavoidable damage would slash the global economy’s GPD by 17 percent in 2050, compared to a scenario with no additional climate impacts after 2020. 

Even so, the new calculations may be conservative.

“They are likely to be an underestimate of the costs of climate change impacts,” Bob Ward, policy director of the Grantham Research Institute on Climate Change and the Environment in London, commented to AFP ahead of the study’s publication.

Damages linked to sea-level rise, stronger tropical cyclones, the destabilisation of ice sheets and the decline of major tropical forests are all excluded, he noted. 

Climate economist Gernot Wagner, a professor at Columbia Business School in New York who was also not involved in the study, said the conclusion that “trillions in damages are all locked in doesn’t mean that cutting carbon pollution doesn’t pay.”

In fact, he said, it shows that “the costs of acting are a fraction of the costs of unmitigated climate change”. 

Global GDP in 2022 was just over $100 trillion, according to the World Bank. The study projects that — absent climate impacts after 2020 — it would be double that in 2050.

Share this:
Continue Reading

News desk

Meta ‘supreme court’ takes on cases of deepfake porn

Published

on

By

Meta's independent oversight board can make recommendations regarding the social media giant's deepfake porn policies but it is up to the tech firm to actually make any changes
Share this:

Meta’s oversight board said Tuesday it is scrutinizing the social media titan’s deepfake porn policies, through the lens of two cases.

The move by what is referred to as a Meta “supreme court” for content moderation disputes comes just months after the widespread sharing of lewd AI-generated images of megastar Taylor Swift on X, formerly Twitter.

The Meta board picked its two cases, regarding images shared on Instagram and Facebook, to “assess whether Meta’s policies and its enforcement practices are effective at addressing explicit AI-generated imagery,” it said in the release.

The board can make recommendations regarding the social media giant’s deepfake porn policies but it is up to the tech firm to actually make any changes.

The first case taken up by the Meta Oversight Board involves an AI-generated image of a nude woman posted on Instagram.

The woman pictured resembled a public figure in India, sparking complaints from users in that country.

Meta left the image up, later saying it did so in error, the board said.

The second case involves a picture posted to a Facebook group devoted to AI creations.

That image depicted a nude woman resembling “an American public figure” with a man groping one of her breasts, the board said in a release.

The board did not name the woman, who it said was identified in a caption on the synthetic image at issue.

Meta removed the image for violating its harassment policy, and the user who posted the content appealed the decision, according to the board.

People were invited to submit comment, particularly on the gravity of harms posed by deepfake pornography and the harm it does to women who are public figures.

Deepfake porn images of celebrities are not new, but activists and regulators are worried that easy-to-use tools employing generative AI will create an uncontrollable flood of toxic or harmful content.

The targeting of Swift, one of the world’s top-streamed artists whose latest concert tour propelled her to the top of American fame, shined a spotlight on the phenomenon, with her legions of fans outraged at the development.

“It is alarming,” said White House Press Secretary Karine Jean-Pierre, when asked about the images at the time.

“Sadly we know that lack of enforcement (by the tech platforms) disproportionately impacts women and they also impact girls who are the overwhelming targets of online harassment,” Jean-Pierre added.

Share this:
Continue Reading

Featured