OK! let’s talk about That Op-ed. The one that insisted not only that privacy is dangerous, but that not affirmatively building surveillance into communication tools is a radical ideological position. 1/ web.archive.org/web/2023010119…
But what’s going on here isn’t substance. And that’s what I want to focus on. Those of us invested in defending privacy need to understand that this op-ed wasn’t written for people with expertise, and its purpose won’t be perturbed by expert rebuttal. We’re not the audience. 3/
The op-ed works to create the appearance of a “debate” on more or less settled issues. This is a powerful function, bolstered by the NYT imprimatur, which allows it serve as a “Potemkin citation” -- a seemingly credible reference in support of bad privacy laws and platforms. 4/
What laws? What political platforms? I don’t know. But the age ID requirement passed in CA this week, and the regulations that would require communications apps to proactively scan and police content that are currently moving forward in the EU and the UK give us some clues. 5/
Particularly because these laws would, in effect, prevent people developing tech from NOT building mass surveillance and censorship capabilities. Which, while extremely poorly argued, is effectively the main thrust of the op-ed. 6/
In short, we are right, our arguments are robust, and we have done the reading. But if we want to defend privacy, we’ll need to be coordinated and bold, and not make the mistake of assuming that being correct is in itself a strategy. There's a lot of work ahead in 2023! 7/
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Thanks for raising this. We did disable some languages recently. In choosing which to keep, we opted for the official language of a given region (in the case of HK, Traditional Chinese). We did this for a couple reasons... 1/
1. the most basic reason: good translation, which is especially important around privacy/security, is expensive and we're a nonprofit. 2/
2. supporting the large number of languages we were previously including made the Signal client file size extremely big, which caused problems for people in low-bandwidth environments & for those with devices that didn't have a lot of storage 3/
📢New paper! In which I work through a lot of my uncomfortable observations since joining academia, examining the alarming-but-quiet capture of academic AI research by big tech, what this means for how and what we know about AI, and how we can resist.
In it I trace the history of the recent turn to AI, which was less about algorithmic advances and more about the concentrated data and compute resources controlled by a consolidated tech industry. Who still gatekeepers these resources.
As "AI" became the hot new thing, universities wanted a piece. And getting one was reliant on access to corporate compute and data resources. W/o these you could do work, but you're not gonna beat the benchmark, and tenure, prestige, big NSF grants would be farther out of reach.
Jen, thank you and solidarity. You're not alone. For the last years AI Now -- the team, myself and other leadership -- struggled to remove Kate and Jason Schultz from the organization and to recover from the toxic pattern of extraction and harm.
I stand with the team. It cost us significantly, in terms of our individual work and mental health, and in terms of our collective access to and standing w/in the academic prestige networks where these two hold power.
I learned a lot about how academic & non-profit institutions are (way too often) structured in ways that celebrate + reward racialized extraction and abuse while propping up white celebrity & power. Solidarity now and forever with everyone working to change this.
[Modest Thread Alert] Late, but thrilled to be a coauthored on this short piece outlining ethical priorities for neurotech and AI. An important step: AI models provide the foundation on which technologies like brain-computer interfaces increasingly rely. nature.com/news/four-ethi…
With that, some additional comments, clarifying my personal position since some of my more nuanced edits aren't reflected in the published version...
First, I think it's important that the call for regulation extend beyond the transfer and reuse of neural data. To be effective, it must also apply to models trained with this data.