@new_otters_raft@piefed.ca

The latest versions of Common Voice datasets, Scripted Speech v27.0 and Spontaneous Speech v5.0, are available for download on the Mozilla Data Collective! A giant thank you to all of our contributors for their hard work putting together these datasets.

What’s new in this release:

Version 27.0 of Scripted Speech includes datasets for 295 languages. These datasets contain 32,349,999 voice clips, making approximately 42,593 hours of speech data available for developing and improving speech technology.
Version 5.0 of Spontaneous Speech includes datasets in 80 languages. These datasets contain 89,754 voice clips of free-form answers to questions, of which 302 hours have been transcribed and validated.

Since the last release, the Common Voice community has welcomed 1 language to Scripted Speech - Pa’O (blk) - and 6 languages to Spontaneous Speech - Chinese (China, zh-CN), Swahili (sw), Palauan (pau), Sundanese (su), Bengali (bn) and Shan (shn).

In Warsaw, marching robots waved flags and chanted slogans outside the digital affairs ministry, calling on Poland’s government to regulate AI before it costs people their jobs.

SOCAN, Canada’s non-profit music rights organization, is suing AI music generation platform Suno, accusing it of infringing on the rights of more than 150 Canadian artists without their consent or providing proper compensation.

“SOCAN has a responsibility to act when the rights of music creators and publishers are put at risk. The evidence shows that the Suno platform has generated and streamed outputs that copy works in our repertoire, and that cannot go unchallenged,” SOCAN CEO Jennifer Brown said in a press release on Wednesday. “Innovation can’t come at the expense of human creativity. The future of music must belong to the people who make it.”

The release also said that the legal action is a “necessary and proactive step to ensure that human music creation is valued, respected, and compensated.” The organization added that it expects more violations to come to light as litigation moves forward.

Alongside the lawsuit, which was filed in Canadian federal court on Wednesday, SOCAN has launched a website comparing Suno outputs to songs by Canadian artists like Tom Cochrane and Avril Lavigne. The Suno clips play nearly note-for-note and word-for-word replicas of popular songs like “Life is a Highway.” The lawsuit’s claims have yet to be proven in court. BetaKit has reached out to Suno for comment.

Remember the original Quest headset that Meta (then Oculus) trumpeted back in 2019? Meta seems to hope you don’t, since the company officially stopped supporting the wireless VR headset in 2023 to focus on the more popular Quests 2 and 3. However, tinkerers haven’t abandoned the hardware and recently released a new root-access exploit and bootloader that gives “developers and enthusiasts full control over Quest 1 hardware.”

The QuestStack project integrates previously known vulnerabilities in the Quest’s Android fastboot process into a privilege escalation chain that leads easily to full root access on the device. The bootloading process is now streamlined enough that it can be completed without any downloads through a web interface after connecting the headset to a PC.

With this exploit, the original Quest hardware can now be officially divorced from any reliance on Meta’s servers or services to be useful. That means enterprising Quest owners should be able to sideload apps without needing to register for a Meta Developer account and activating Developer Mode through Meta’s mobile app. It also means users should be able to go through the initial setup and login steps for a fresh Quest headset even if and when Meta decides to shut down the servers that currently support this process.

FoodNeverComes is one of a growing number of fake online shopping sites where users can place a pretend food order. (CBC)

Rhia Catapano, an assistant professor at the University of Toronto who studies consumer behaviour and psychology, says online shopping simulators can activate “the same parts of our brain that doing the actual activity would.”

That’s particularly appealing for budget-conscious young people.

“Like playing a video game or dreaming about a vacation that you’re never actually going to take, we can imagine a thing,” she said. “In some way, it gives us some of the same psychological benefits.”

The sudden popularity of these sites, says Catapano, shows just how deeply consumerism has become integrated into our everyday lives, with people willing to spend time and effort browsing and “shopping” even when they know nothing will ever arrive.

But she sees a glimmer of resistance, too.

Some people, she says, are realizing through these sites that their shopping habits aren’t serving them — and are looking for ways to consume less and make more intentional choices.

“They recognize that what they’re doing is not good for them, so they’re trying to make changes.”

As journalists are shut out of conflict zones, investigators are using satellite imagery, social media and 3D modelling to document atrocities. But evidence alone cannot deliver accountability.

The article is long but comprehensive, please see if for the full discussion.

Agencies tout very high success rates, but some experts compare it to “crime control theatre”

On Sunday, March 22 of this year, a large swath of the population in Quebec was woken up at 4:25 a.m. as cellphones lit up and screamed. An Amber alert had been broadcast. Less than four hours later, the two missing children were thankfully found, unharmed, and the alert was cancelled.

Amber alerts cast wide nets in an effort to find abducted children and teenagers, but each time one is transmitted in Canada, we go through a predictable cycle. Many complain that these alerts cannot be turned off or muted, even in sleep mode, and that waking people up in the middle of the night to help them find a specific car that was last spotted 200 kilometres away is unnecessarily disruptive—with some even calling 911 to complain. Meanwhile, law enforcement officials remind us through the media that these alerts are only used when necessary and that they save lives.

Any subject tied to the safety of children is bound to evoke the deepest emotions. In fact, the crime that led to the creation of Amber alerts—and to their name—was monstrous enough that it’s no wonder it rallied influential people into building a system that could potentially prevent this kind of crime.

But warranted emotions aside, do Amber alerts even work? What little data we do have point to a system that has likely been oversold to us.

French President Emmanuel Macron has been the global leader in promoting a ban on social media for kids, consistently welcoming countries that propose such measures with a “thanks for joining the movement” tweet. That includes Canada, which received its congratulatory tweet after a ban was included in Bill C-34. As I argued last month, however, I believe the Canadian ban is vulnerable to constitutional challenge on at least two fronts: freedom of expression, since blocking an entire cohort of Canadians from accessing information and expressing themselves on social media infringes Section 2(b) of the Charter, and privacy, since enforcing the ban will require millions of Canadians to verify their age to use social media services with the promised safeguards nowhere in place when the ban takes effect. Last week, France became the first major jurisdiction to have such a ban reviewed by its highest constitutional court, and it proceeded to strike down the law on substantially the same two constitutional concerns.

The French Parliament adopted a social media ban for kids under 15 on July 21st, with the law scheduled to take effect in September. But France’s system of constitutional review led the ban to be reviewed by the constitutional court, resulting in it being assessed and rejected before ever taking effect. Macron responded to the ruling by instructing Prime Minister Sébastien Lecornu to rework the legislation to address the court’s concerns, with the government now targeting spring 2027 for a revised law.

TLDR on the history side of the article

In 1804, a French weaver named Joseph-Marie Jacquard patented a loom that could weave intricate patterns using punch cards — paper cards with encoded holes.

Once they pass through the loom, the presence or absence of a hole tells the loom which warp thread to raise or lower, acting as the earliest form of binary code programming. Weaving that once required a master weaver and an assistant could now be done by craft hobbyists.

This feels like bad design

Smart pet-feeder company Petlibro suffered an outage on Tuesday, giving users a harsh reminder of the risks of relying on smart devices.

Petlibro makes pet feeders that connect to Wi-Fi and, for setup with Petlibro’s mobile companion app, Bluetooth. Users can set the feeders to automatically dispense food at scheduled times, which is handy for pet care when the owner is out of the home.

In a statement to The Verge, Petlibro CEO York Wu said this week’s outage was “related to how the Petlibro app communicates with devices through its online servers.”

Some user reports have also conflicted with Petlibro’s claims about offline functionality. As The Verge pointed out, various users online claimed that during the outage, their devices did not perform any scheduled feedings, performed some feedings but not others, or performed feedings late (examples here, here, here, here, and here).

Terabytes worth of credentials, many belonging to the world’s biggest and most sensitive organizations, have been exposed in a supply-chain attack on LiteLLM, an open source tool that streamlines A- driven software development. Microsoft, Amazon, Cisco, Samsung, and Salesforce are only a handful of the entities whose access secrets were exposed.

The revelation was posted on Tuesday and Wednesday by security firms CloudSEK and Hudson Rock. CloudSEK said it found cloud keys, repository tokens, SSH keys, Kubernetes secrets, package publishing credentials, environment variables, and AlI provider keys that could allow attackers to gain access to more than 2,500 organizations.

The credentials were extracted during a 40-minute window in March while the victims used compromised versions of LiteLLM downloaded from the package’s official location in the Python Package Index repository. Hudson Rock said it made the discovery after analyzing a 195TB file that it obtained. Neither firm identified the source of the information.