Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Liz O’Sullivan Profile picture
Liz O’Sullivan
@lizjosullivan
CEO, @AskVeraIO. Committee Member, NAIAC. Previously: cofounder @itsArthurAI & 1st technology director @StopSpyingNY. Tweets in my own capacity only.
Jun 8, 2021 • 5 tweets • 2 min read
All the privacy news on Sidewalk says it’ll drain your battery, spend your data plan, expose you to hacking threats, all of this is true. But it misses the forest for the trees. Amazon in possession of a nationwide network gives them the power of a utility. With less oversight. People sometimes forget that the original tech villains were the telecoms, and rightfully so. Sidewalk can’t wiretap your conversations* like the telecoms can, but that’s not the only way police have used them to put innocent people in jail.
newyorker.com/news/news-desk…
May 7, 2021 • 11 tweets • 3 min read
Kinda excited and kinda sad that in my short career as a tech activist, I can now, for the first time, throw up my hands and say “I’ve been talking about this for years!” 🤦🏻‍♀️ One of my most popular tweets, in fact, was about this. I don’t love sharing it bc some mild factual errors appear due to haste and lack of available info at the time, but for posterity, here it is. Every now and then it resurfaces in my mentions bc (surprise!) people DO care.
Sep 21, 2020 • 4 tweets • 2 min read
Did you know that Amazon Sidewalk can track you, even if you’re not logged on?

Last year, people called me paranoid for worrying about probe requests and Amazon’s mesh network, but here we are a year later and location tracking is now more clearly at the heart of the operation. The more devices they sell into the world, the more complete a picture they’ll have of what’s in it; what we do, where we go, how we spend our time. Sidewalks talk to each other. And likely to all WiFi devices, with or w/o Sidewalk, in subtle ways.
theverge.com/2020/9/21/2144…
Jul 29, 2020 • 18 tweets • 7 min read
Me, getting ready to watch House anti-trust testimony: My whole Twitter feed:
Jun 11, 2020 • 8 tweets • 3 min read
🚨🚨This should feel like winning but it feels more like a thinly veiled threat. @BradSmi says MSFT won’t sell FRT without federal law governing use. IN FACT, Microsoft has WRITTEN such a law, and their lobbyists are pushing it from state to state.🚨🚨 /1
thehill.com/policy/technol… SURPRISE: their law sucks. Policy experts agree that, while there is benefit to requiring a warrant for the use of FRT, the MSFT-backed law has giant loopholes, allowing a vague definition of “emergency” to allow for continued rampant use of FRT unsupervised by the courts. /2
Jun 9, 2020 • 6 tweets • 1 min read
This is, without a doubt, my favorite algorithmic bias story of the year. There are so many errors here, I don’t even know where to begin... 1. In large scale automation cases like this one, you better be damn well sure your classifiers aren’t biased. Since this is an impossible task, (all models are biased) AT LEAST retain meaningful human oversight of the task at hand.
Jun 8, 2020 • 4 tweets • 1 min read
Memories aren’t real... OR ARE THEY???? Become God by eating worm poop.
Jun 3, 2020 • 7 tweets • 4 min read
Please bear with me while I pull all of my threads together in one of those annoying "I'm pulling all my threads together"... thread of threads (and links, I guess).

This is my Origin Story™: nytimes.com/2019/03/01/bus… Ok but what's the big deal about killer robots? LET ME TELL YOU

aclu.org/blog/national-…
May 29, 2020 • 4 tweets • 1 min read
Military technology always seems to find its way back home
vice.com/en_us/article/… What is CBP even doing in Minnesota, anyway?
May 20, 2020 • 4 tweets • 2 min read
Turns out Google is still very interested in defense work... /1
prnewswire.com/news-releases/… One day before this news breaks, Google delivers what seems to be a big win for PR and their activist staff, but continues to do “non-custom” AI work for the same industry. So... 🤷🏻‍♀️ /2
forbes.com/sites/rachelsa…
May 10, 2020 • 6 tweets • 2 min read
I've been talking with @SethColaner at @VentureBeat a bit about COVID-19 and AI, but mainly I've been quiet about things publicly. I have loved ones who are immunocompromised and we live in the love of my life, NYC, aka the epicenter of the outbreak. /1 venturebeat.com/2020/05/08/ai-… We must get the virus under control. It's important that, as surveillance activists and sometimes-AI skeptics, we understand that the top priority for folks (ourselves included) is worrying about survival in a non-abstract way. But we must also remember that huge changes are /2
Apr 28, 2020 • 4 tweets • 1 min read
🚨 Here's what @RicCantrell, AG of Utah, said of Banjo, the ex-white-nationalist/Neo-Nazi-run surveillance company "in Utah we have a fighting chance to invent the system in a way that will not subvert the historic liberties of citizens... I don’t trust other states to do it." 🚨 Dear Ric,

Civil liberties advocates think that live, real-time camera surveillance powered by AI is incompatible with a free democracy. We certainly don't think Banjo is a company we can "trust".
Feb 11, 2020 • 4 tweets • 2 min read
For the love of god, can the press PLEASE stop legitimizing these one-off anecdotal experiments? Like, FFS, no, you did not @CNN. Image
Jan 29, 2020 • 7 tweets • 2 min read
Next up excited for @KLdivergence with a risk assessment tool analysis on “overbooking”. This is when someone is arrested on charges that are more serious than they should be. First off: there’s very little accountability so this is gonna be a hard problem to address. #FAT2020 Guess what? There is a lot of racial disparity in this. I’ll say... But again, this is hard to address because there’s no real “ground truth”. It’s complicated and we can attack the issue by looking at actual convictions vs. accusations.
Jan 29, 2020 • 7 tweets • 2 min read
Next up, something I’m very interested in: bias in hiring algorithms. Can it be made better? #FAT2020 with @manish_raghavan we’re gonna find out. Looking at vendors like HireVue and others, specifically all their publicly available information on steps they’ve taken to “de-bias”.
Jan 28, 2020 • 8 tweets • 2 min read
Quote of the day at #FAT2020: Image So far my key takeaway from this conference is that anthropologists are absolutely vital to this process. A call to “study up” is to contextualize a process by studying its situation within systems. Image
Jan 28, 2020 • 5 tweets • 2 min read
Ok now how do we make it more fair? Kit Rodolfa at #FAT2020: recidivism algorithms. Image Too true: anything you optimize for will have a down side. In this case, let’s look at recall optimization. In other words, equal rates of finding those who will actually re-offend, regardless of race (I think? Might have that backwards...)
Jan 28, 2020 • 5 tweets • 2 min read
Fairness now at #FAT2020 starting off with a really important question: what do you do if you don’t collect demographic data? Image Proxy methods like BISG can be used to infer the category, but that means that all auditing is vulnerable to the particularities/problems in the proxy method.
Jan 28, 2020 • 5 tweets • 2 min read
Explainability up next (!) at #FAT2020 with @JcMalgieri and @MargotKaminski paper on GDPR and impact assessments. Image No public consensus on the Right to Explanation but lawyers wanted automated decision making to be justified based on fairness. Image
Jan 28, 2020 • 5 tweets • 2 min read
Next up: the brilliant @rajiinio!! #FAT2020 Image Data sheets and model cards aren’t enough. Design decisions must be examined. Especially in large, complex systems with multiple algorithms. Image
Jan 12, 2020 • 4 tweets • 2 min read
I had a few words to say about the new White House guidance for agencies seeking to regulate AI. Tl;dr — it’s not good enough. /1 Yes, bias and explainability are table stakes; I’m glad to see our government require that any rulemaking in AI address those points. But what about social harms at scale? Automation? Impact assessments? Documentation? All missing, nowhere to be found. /2