Shlomi Hod @hodthoughts

Responsible AI. Previously, @BUCompSci @opendp_org @Columbia @Twitter. shlomi.hod.xyz Berlin, Germany Joined March 2022

Tweets

138
Followers

240
Following

2K
Likes

171

Mahdi Haghifam @HaghifamMahdi

a week ago

Very happy to share our paper with Adam Smith and Jon Ullman (@thejonullman) on the “sample complexity” of membership inference (MIA). We ask: Is the number of data typically used in practice enough to build strongest possible MIAs? 🧵 link: arxiv.org/abs/2508.19458

9 12 90 8K 33

Download Image

Hayden Field @haydenfield

a week ago

NEW: Anthropic will start training its AI models on user data, including new chat transcripts & coding sessions, unless users choose to opt out by 9/28 (it's a pop-up window that will give you the choice). It’s also extending its data retention to 5 years. theverge.com/anthropic/7675…

38 194 1K 180K 397

François Chollet @fchollet

2 weeks ago

LLM adoption among US workers is closing in on 50%. Meanwhile labor productivity growth is lower than in 2020. Many counter-arguments can be made here, e.g. "they don't know yet how to be productive with it, they've only been using for 1-2 years", "50% is still too low to see…

Oyvind Bjerke @BjerkeOy

3 weeks ago

17 40 211 710K 110

Download Image

337 591 5K 919K 1K

OpenDP @opendp_org

2 months ago

The OpenDP Community Meeting registration site will be available soon! In the meantime, big THANKS to our newly added sponsors @CapitalOne , @Google , @Meta , and @Apple in addition to @Microsoft and @oblivious_AI for your generous support! sites.harvard.edu/opendp/events/…

0 3 4 1K 1

Simon Willison @simonw

3 months ago

The biggest scandal in AI energy right now deserves to be xAI's data center running 35 methane gas turbines that don't need air permits because "temporary" and don't have catalytic reduction pollution controls installed because... they just didn't bother? simonwillison.net/2025/Jun/12/xa…

32 146 1K 94K 271

Deb Raji @rajiinio

4 months ago

In the absence of federal action, the states have led the way for years on pushing forward common sense AI policy. This is really not good.

4 6 27 3K 3

Download Image

Meredith Whittaker @mer__edith

4 months ago

📣 The rushed integration of AI into everything is a major threat to privacy, as we've been saying. Microsoft Recall is a great (not great!) example of exactly why--it screenshots everything on your desktop, to give AI the data to tell you...what you were doing on desktop.…

Signal @signalapp

4 months ago

120 531 4K 199K 326

17 177 642 50K 85

SaTML Conference @satml_conf

4 months ago

🌍 Help shape the future of IEEE SaTML! We are on the hunt for a 2026 host city - and you could lead the way. Submit a bid to become General Chair of the conference: forms.gle/vozsaXjCoPzcbt…

0 9 14 8K 4

Nathan Lambert @natolambert

4 months ago

transparency in AI is dying: no evals, no release notes, just vibes and more bad naming

Sam Altman @sama

4 months ago

transparency in AI is dying: no evals, no release notes, just vibes and more bad naming

2K 1K 23K 3.6M 1K

56 122 3K 235K 144

Daniel Björkegren @danbjork

6 months ago

New working paper: Could AI leapfrog the web? Only 37% of sub-Saharan Africans use the internet. Cost is the #1 constraint. In a study with 469 teachers in Sierra Leone, we find AI works better than the web--and is 87% cheaper. arxiv.org/abs/2502.12397 Let's dive in... 🧵(1/X)

2 22 63 10K 21

Download Image

Shayne Longpre @ShayneRedford

6 months ago

What are 3 concrete steps that can improve AI safety in 2025? 🤖⚠️ Our new paper, “In House Evaluation is Not Enough” has 3 calls-to-action to empower independent evaluators: 1️⃣ Standardized AI flaw reports 2️⃣ AI flaw disclosure programs + safe harbors. 3️⃣ A coordination…

6 31 61 12K 34

Download Image

Thomas Wolf @Thom_Wolf

7 months ago

After 6+ months in the making and burning over a year of GPU compute time, we're super excited to finally release the "Ultra-Scale Playbook" Check it out here: hf.co/spaces/nanotro… A free, open-source, book to learn everything about 5D parallelism, ZeRO, fast CUDA kernels,…

110 706 4K 363K 4K

Download Image

Shashwat Goel @ShashwatGoel7

7 months ago

🚨Great Models Think Alike and this Undermines AI Oversight🚨 New paper quantifies LM similarity (1) LLM-as-a-judge favor more similar models🤥 (2) Complementary knowledge benefits Weak-to-Strong Generalization☯️ (3) More capable models have more correlated failures 📈🙀 🧵👇

6 29 141 21K 88

Download Image

Gary Marcus @GaryMarcus

8 months ago

Scientific alert: OpenAI’s apparently impressive results with o3 on FrontierMath should be taken with a huge grain of salt. - OpenAI had access to the problems and solutions. - We don’t know what they trained on. - We don’t know what problem-tailored validation or data…

Elliot Glazer @ElliotGlazer

8 months ago

3 3 33 34K 8

19 41 252 28K 60

Association for Computing Machinery @TheOfficialACM

8 months ago

Congratulations to #ACMFellow Cynthia Dwork on receiving the prestigious National Medal of Science! The honor recognizes her renowned contributions to the science of cryptography, distributed computing, algorithmic fairness, and differential privacy: bit.ly/3BXsnuZ

1 9 73 4K 2

Download Image

anton @abacaj

8 months ago

using LLMs in predefined workflows makes a lot of sense. using them to discover the workflow on its own doesn’t work (agent). had to build a way to fetch manuals (pdfs) from websites, some websites required you to fill in an input field which came from another source, this often…

38 38 443 40K 224

Jakob Mökander @jakobmokander

8 months ago

New Paper Alert! How can AI innovation improve social outcomes? @ds_wats0n Luciano Floridi and I @oiioxford introduce “Sociotechnical Pragmatism” - a robust stance for policymakers seeking to reap the benefits of AI while managing the risks link.springer.com/article/10.100… 🧵 1/10

1 11 32 5K 18

Niloofar (✈️ ACL) @niloofar_mire

8 months ago

I've been thinking about Privacy & LLMs work for 2025 - here are 5 research directions and some key papers on privacy/memorization to get started: 🧵

11 54 336 32K 326

Download Image

Shreya Shankar @sh_reya

8 months ago

how come nobody is talking about how much shittier eng on-calls are thanks to blind integrations of AI-generated code? LLMs are great coders but horrible engineers. no, the solution is not “prompt the LLM to write more documentation and tests” (cont.)

80 129 2K 331K 852

Digital EU 🇪🇺 @DigitalEU

9 months ago

📢 Exciting Job Opportunities at the EU AI Office! The new 🇪🇺 European AI Office is looking for professionals with legal and policy backgrounds to join their team & support the implementation of the #AIAct. Learn more and apply now ⤵️!