Thomas Steinke

@steinke
352 Followers
62 Following
26 Posts
Thomas Steinke (@stein.ke)

Computer science, math, machine learning, (differential) privacy Researcher at Google DeepMind Kiwi🇳🇿 in California🇺🇸 http://stein.ke/

Bluesky Social
The most mundane superpower: Being able to remember everyone's name after the first introduction.
Physical SIM cards are an incredibly inefficient way of storing a digital identifier, but have you ever tried transferring an eSIM to a new device?

I wrote some expository material about DP algorithms that go beyond the standard clip-and-add-noise paradigm:

https://differentialprivacy.org/inverse-sensitivity/

https://differentialprivacy.org/down-sensitivity/

Beyond Global Sensitivity via Inverse Sensitivity

The most well-known and widely-used method for achieving differential privacy is to compute the true function value \(f(x)\) and then add Laplace or Gaussian noise scaled to the global sensitivity of \(f\). This may be overly conservative. In this post we’ll show how we can do better.

[20] sounds like it has some interesting results, I should read that paper.

*checks references*

oh, it's my paper

(Maybe I'm already like that. Who knows? In any case, I'm not doing good enough work that anyone would read my essays on sentience and the foundations of reality.)

Every year I feel a little bit more cynical, a little bit more exhausted with how the world works, a little bit more convinced that everyone else is insane.

If this trend continues for a few more decades, I'll become one of those senior researchers who now only studies "sentience" or something like that.

I used to wonder how that happens to people who used to do good work, but now I think I understand...

Pleased to announce the completion of my todo list for 2021.

For the holidays, here's one of my favourite #maths facts: Legendre's Constant! Legendre gave his name to a number!

This number is 1.

A tech company source used to make me turn off my phone before meetings because of concern that his employer would check to see if our phones were near each other. This was not a crazy concern it turns out! TikTok tried something like this to hunt down leakers: https://www.nytimes.com/2022/12/22/technology/byte-dance-tik-tok-internal-investigation.html
ByteDance Inquiry Finds Employees Obtained User Data of 2 Journalists

The company’s internal investigation showed that workers also obtained data on a small number of other U.S. users.

The New York Times