
now @google - Before: [email protected] / Consultant & Lecturer @ETH / Machine Learning for Archives
Tokyo, Japan
Joined July 2016
- Tweets340
- Following175
- Followers322
- Likes457
Benoit Seguin retweeted
How can we address the scarcity of data required for specialized AI?
Learn about Simula, a framework that reframes synthetic data generation as dataset-level mechanism design. By using reasoning to architect datasets from first principles, Simula enables fine-grained control over coverage, complexity, and quality.
More →goo.gle/4cf15iV
Benoit Seguin retweeted
We’ve identified a “Collaboration Gap” in today’s top AI models.
Testing 32 leading LMs on our novel maze-solving benchmark, we found that models that excel solo can see their performance *collapse* when required to collaborate – even with an identical copy of themselves.
A \🧵
Benoit Seguin retweeted
We recently published our early paper on Simula, an internal framework we built at Google that has powered >20 launches using #SyntheticData across products over the previous year.
Simula empowers our teams to generate data at scale, offering fine-grained control over diversity, complexity, & quality. ✨
Curious about Simula's applications? Here’s a glimpse:
-📱 On-device models (Pixel scam call detection, Android messages scam protection, Gemini Nano safety classifiers, etc.)
- ☁️ Specialized server-side models (Gemini AI moderation, ShieldGemma 1 & 2, etc.)
- 🛡️ Pre-launch evaluations & crucial safety testing for new products
- ...and much more!
At its core, Simula leverages LLM reasoning at every stage: from strategizing diversification factors and building deep domain-specific taxonomies, to advanced meta-prompting and data complexification.
It then deploys the best tools for executing data generation based on this reasoning, plus robust mechanisms for critiquing & revising as needed. 🛠️
📄 Early Paper openreview.net/pdf?id=VOoeog…
While this paper focuses on our seedless generation approach, Simula's capabilities extend far beyond – covering all kinds of data augmentation and the multi-modal, multi-step nature of modern data generation.
We have more coming in the full paper!
Huge thanks for the great co-authors who made this paper possible: @im_td, @Seguin_Be, @enricobacis, & @cesar_ilharco.
Building Simula, with its flexible abstractions for diverse use cases, was a significant challenge. @Seguin_Be and I bootstrapped this journey last year. The paper tells part of the story, but the effort to build usable software extends far beyond theory to the smallest details.
Still, the potential we've unlocked continues to amaze us!
#ICLR2025 #SyntheticData #AI #LLMs
Benoit Seguin retweeted
synthetic data is the future. but — generating explainable and controllable synth data at Google-scale is **really** hard.
stop by Peridot 202-203 at 1130-1230 today to chat w/ @hamzaharkous and me about “Orchestrating Synthetic Data w/ Reasoning”! 🦾🧠
openreview.net/pdf?id=VOoeog…
Benoit Seguin retweeted
🚨 @BiblHertz offers Pre- and Postdoctoral Fellowships for the academic year 2022/23 in the department of @TristanWeddigen. Deadline: July 24.
Find out more here:
Department 👉https://teapawt.net/t.co/kB0qDcSeK9
Application 👉https://teapawt.net/t.co/0zmrwCcZge
Benoit Seguin retweeted
[📅 TODAY]
From 12 to 2pm we are very pleased to welcome two speakers:
🗣️ @v_christlein: « Machine Learning for the Analysis of Artworks and Historical Documents »
🗣️ @Seguin_Be: « Beyond the Proof-of-Concept — DHAI Methods in the Real World »
🔽🔽🔽
📢 #DHAI_seminar
Next session of the #DHAI seminar will take place on May 10th, from 12 to 2pm (UTC+2):
🗣️ @v_christlein: « Machine Learning for the Analysis of Artworks and Historical Documents »
🗣️ @Seguin_Be: « Beyond the Proof-of-Concept — DHAI Methods in the Real World »
Benoit Seguin retweeted
A detailed neoclassicism painting depicting the frustration of being put on hold during a phone call #dalle
Benoit Seguin retweeted
📢 #DHAI_seminar
Next session of the #DHAI seminar will take place on May 10th, from 12 to 2pm (UTC+2):
🗣️ @v_christlein: « Machine Learning for the Analysis of Artworks and Historical Documents »
🗣️ @Seguin_Be: « Beyond the Proof-of-Concept — DHAI Methods in the Real World »
Benoit Seguin retweeted
Totally not suspicious and totally organic activity of totally real people on social media totally not seeking to influence French Presidential campaign.
I have been working a lot recently on a project where we tackle the art market pricing problem, providing explanable valuations with the most advanced ML technologies. It is taking shape and I am so excited 😁
Btw, I will be at #ArtBasel this weekend if people want to meet.
Benoit Seguin retweeted
Notes from my @uOttawa workshop talk on why increased DH data complexity means increased labor - and not just the work of data entry: matthewlincoln.net/2021/06/0…
Benoit Seguin retweeted
"Hosting SQLite databases on Github Pages" is absolutely brilliant: it adds a virtual filesystem to SQLite-compiled-to-WebAssembly in order to fetch pages from the database using HTTP range requests phiresky.github.io/blog/2021…
Benoit Seguin retweeted
I couldn't resist. With contributions from @maxhkw
Benoit Seguin retweeted
Good old #dhSegment by Sofia Ares Oliveira and @Seguin_Be helping us to look at what's important in the digitised images of #BilderDerSchweiz @swissartresearX github.com/dhlab-epfl/dhSegm…
Benoit Seguin retweeted
All 1.8M papers in arXiv - title + abstract embedded with SPECTER (@ai2_allennlp) -> UMAP (@RAPIDSai). Colored by the year of the "update" date, which I think roughly maps to the original upload date.
To anyone who, like me, has sometimes issues with conda being stuck in dependencies solving for large environments. Mamba is an amazing drop-in replacement. Instant favorite! github.com/mamba-org/mamba @mambaorg
1
Benoit Seguin retweeted
We asked our art history students @artskuleuven to create and photograph Still Lifes #lockdownchallenge no.1 — results are 🤩







































