Good morning. If you’re reading this from the U.K., then you’re more likely to be a Reddit user than an X user.
That’s according to the British media regulator Ofcom, whose latest survey shows Reddit to be the country’s fastest-growing large social media platform, now used by most online adults in the U.K. X, meanwhile, has been shedding Brits for a couple years.
Other findings include ChatGPT being used by a third of U.K. internet users at least once during the last year…and a measly 18% of over-16s believing that generative AI’s output is reliable. —David Meyer
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
Australian social media restrictions outrage
Australia’s lawmakers last week took the drastic step of banning (as of late next year) all social media use by under-16s, parental consent be damned.
The legislative process, which was reportedly the result of a politician’s wife reading Jonathan Haidt’s bestseller The Anxious Generation, took all of one week. Social media companies are unsurprisingly really unhappy with how this all went.
A Meta spokesperson pointed out that the Australian Parliament had admitted a lack of demonstratable, clear links between social media and negative impacts on kids’ mental health, adding: “This demonstrates the lack of evidence underpinning the legislation and suggests this was a predetermined process.”
TikTok and Snap are similarly irked, with one of their big concerns being the fact that nobody knows how this new law is supposed to be enforced when it comes into effect.
Even with milder laws allowing access with parental consent, experience has shown that enforcement is the big issue—it’s tricky to get a mechanism that works while still respecting privacy and other rights.
Musk targets OpenAI transformation
Elon Musk has asked a Californian district court to stop OpenAI doing a bunch of things he doesn’t like.
The tycoon, who co-founded OpenAI nine years ago before leaving after he lost a power struggle with CEO Sam Altman, claims the AI leader is discouraging investors from backing rivals such as his own xAI.
He also alleges that OpenAI “wrongfully obtained competitively sensitive information” through its Microsoft connections, and wants the court to stop OpenAI from doing business with companies in which the defendants have material financial interests.
Those defendants include Altman and president Greg Brockman, as well as Microsoft board member Reid Hoffman and Microsoft VP Dee Templeton, as well as Microsoft itself.
Crucially, Musk wants the court to stop OpenAI’s conversion from non-profit to for-profit.
The Tesla boss sued OpenAI and Altman earlier this year over their abandonment of the firm’s non-profit status, but withdrew the suit in June. Now it’s sort of back, in expanded form.
Canada goes for Google’s jugular
The Canadian antitrust authority has sued Google to “put a decisive end” to what it claims are the company's anticompetitive practices in online advertising—which still represents the majority of parent company Alphabet’s revenues.
“Through a series of calculated decisions, taken over the course of multiple years, Google has excluded competitors and entrenched itself at the center of online advertising,” the Competition Bureau said.
The watchdog wants Google to divest its AdX ad exchange and DFP ad server, plus a fine running as high as 3% of global revenues (so, just north of $9 billion, based on fiscal 2023.)
Google ad VP Dan Taylor responded by insisting that the complaint “ignores the intense competition where ad buyers and sellers have plenty of choice.” The company now has a month and a half to file its official response.
New AI models
Amazon has reportedly developed a new multimodal large language model, which is to say it can handle images and video as well as text.
According to The Information, the “Olympus” AI model will help Amazon avoid over-reliance on Anthropic, the OpenAI rival that it’s been heavily funding.
And with Amazon holding its annual re:Invent AWS conference over this week, Olympus could get its debut imminently.
Meanwhile, the Qwen AI team at China’s Alibaba has produced a new “reasoning” model that will compete against OpenAI’s o1 model.
The team claims that QwQ-32B-Preview outsmarts o1 on some benchmarks, though they also note certain bad habits, like the model suddenly switching languages for no reason. Being Chinese, it also avoids touchy political subjects like the Tiananmen Square massacre.
Meta’s cable plan
Meta wants to wrap the entire world in a new fiber-optic subsea cable, TechCrunch reports.
What’s particularly novel about this plan is that Meta would be the sole owner and sole user of the data pipeline.
The initial budget for the project, which the Facebook and Instagram proprietor will apparently announce next year, would reportedly be $2 billion, with $10 billion being the more likely cost once all is said and done.
Data Sheet’s South African author today is delighted to note that the project is apparently being led by Meta’s operation in that country—which would play host to two interconnection points, as would India.
The U.S. would get two on each coastline. Australia just gets one.
More data
Intel's $7.86 billion subsidy deal restricts sale of its manufacturing unit. Strings were attached.
Italian watchdog warns publisher GEDI against sharing data with OpenAI. Privacy rules are the sticking point.
Amazon employees are ‘rage-applying’ as its return-to-office mandate looms—experts say that’s exactly what CEO Andy Jassy wants. Is it a sneaky layoff ploy?
Fully autonomous planes could be coming to a flight near you—and that’s not a good thing. Your captain is speaking.
TSMC founder Morris Chang offered top job to Jensen Huang, memoir shows. “I already have a job,” said Nvidia’s CEO.
