This week: after the US Supreme Court’s landmark overturning of Roe v Wade, what measures can people take to ensure the privacy of their reproductive choices? Plus, Meta’s new AI Rosetta Stone, and the latest from Elon’s Twitter takeover. It’s The Roundup
News
It’s the story that won’t die; the news-cycle cockroach. Elon Musk’s takeover of Twitter seems less and less likely by the minute as its prospective head argues about the state of the platform, with the incumbent, via that selfsame platform. It’s all very meta.
In new developments, related by Dan Milmo at The Guardian, Twitter has doubled-up on a May Tweet from current CEO Parag Agrawal, claiming that the platform suspends 500,000 spam accounts a day. Now that figure has been revised up to a colossal 1 million.
Spare a thought for those people manually checking thousands of these accounts a month, and pray that we can all move on from this nonsense soon.
If you’ve been following The Roundup, you’ll have seen that the bigwigs at Meta are desperately engaged in trying to steal back their own lunch from TikTok. Now comes reports that Instagram is experimenting with automatically converting videos into Reels.
As Aisha Malik of TechCrunch notes, Meta revealed earlier this year, in its Q1 earnings, that Reels now accounted for 20% of time spent on Instagram, leading to a pivot away from photos and videos.
Still, it’s unclear how this development, if implemented platform-wide, would affect videos already on Instagram, nor how videos shot in a landscape orientation would fare. At the very least, though, this is yet another piece of evidence of Meta’s intention to move quickly and aggressively to regain its lost market share.
Analysis
As many feared, the US Supreme Court announced its decision to end constitutional protections for abortion, leaving individual states to decide their own stance on the issue. As this chart from The Guardian shows, this means that laws denying reproductive rights are in place, or in the works, in more than half of the 50 US states.
The immense ramifications of this decision cross into online behavior, too. Earlier this month, Google announced that it would delete the location histories of those visiting “sensitive” sites such as abortion or fertility clinics, to minimize the possibility that such data could be used in prosecutions. But experts are recommending other precautions as well. As this guide from researchers Julia Slupska and Laura Shipp makes clear, even such seemingly innocuous tools as period trackers could be weaponized as instruments of “criminalisation and surveillance.”
Ultimately, they argue, without an American equivalent of the EU’s GDPR laws, the reversal of Roe v. Wade is not only a blow to the sanctity of bodily autonomy, but also the notion of individual privacy itself.
The Indian government is clamping down on social media companies over content that the government deems harmful or illegal.
In recent week, the Indian government ordered Twitter to remove numerous posts, to which Twitter mostly acquiesced. Now, however, Twitter has taken legal action against the government in a bid to overturn some of its orders.
The difficulty of operating under increasingly repressive speech laws is highlighted in this article by Rest of World’s Andrew Deck, who notes the increase in so-called “hostage-taking laws” around the world. These laws mandate that tech companies hire compliance officers, stationed locally, who are made criminally liable for failures to comply with government orders.
AI
/cloudfront-us-east-2.images.arcpublishing.com/reuters/HZO2LUX5XJLSTKN6HGKFIPZYYU.jpg)
News last month that a Google engineer had been suspended for claiming that the LaMDA AI with which he had been working had become sentient, has spurred renewed interest in the ethics human/AI interaction.
Is there anything wrong with people believing that their AI has become “alive”? Is there harm in people developing emotional attachments to a sophisticated mimic?
This enlightening article from Reuters delves into this confusing world.
And finally: Meta has open-sourced a single AI model that is able to translate 200 different languages, many of which have so far been beyond the capabilities of existing tools.
According to this article by James Vincent at The Verge, one of the major achievements of this model is its focus on languages of that are not widely spoken, “with fewer than 1 million publicly available translated sentence-pairs.”
Of the 700 languages which we know have gone “extinct,” over 30% have disappeared in the last 60 years, whilst of the 7000 known languages currently in use, over 40% are believed to be endangered.
There’s hope, then, that these languages may be functionally preserved for posterity, even if they fall silent.
