Christian Szegedy
@ChrSzegedy
Followers
43K
Following
23K
Media
253
Statuses
8K
#deeplearning, #ai research scientist. Opinions are mine.
Sunnyvale, CA
Joined June 2015
I’ve worked in some of the top AI labs from Google to xAI. What @Jenny Shao and team have built at Robyn, an emotionally intelligent AI that remembers and grows with you—is truly remarkable. Even more impressive is Jenny’s relentless drive to build something that empowers real
Say hello to Robyn -- the first emotionally intelligent AI designed to be deeply human. 1000 free messages for accounts created in the next 24 hours. Download and comment below for access. 🧵 👉 Get Robyn: https://t.co/ZaMA1SqAgq
https://t.co/iPMN5hZxbn | available on iOS
5
30
110
Huge progress on user preference optimization. A frontier model you'd like to use.
Introducing Grok 4.1, a frontier model that sets a new standard for conversational intelligence, emotional understanding, and real-world helpfulness. Grok 4.1 is available for free on https://t.co/AnXpIEOPEb,
https://t.co/53pltyq3a4 and our mobile apps. https://t.co/Cdmv5CqSrb
21
23
346
Grok 4.1 just released. You should notice a significant increase in speed and quality.
Introducing Grok 4.1, a frontier model that sets a new standard for conversational intelligence, emotional understanding, and real-world helpfulness. Grok 4.1 is available for free on https://t.co/AnXpIEOPEb,
https://t.co/53pltyq3a4 and our mobile apps. https://t.co/Cdmv5CqSrb
8K
8K
51K
A new argument for our living in a simulation.
Hamiltonian Monte Carlo frames sampling from a probability distribution as a physics problem. By endowing "particles" with momentum and simulating their energy and motion through Hamilton's equations you can efficiently explore a distribution.
19
45
335
We disrupted a highly sophisticated AI-led espionage campaign. The attack targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We assess with high confidence that the threat actor was a Chinese state-sponsored group.
1K
3K
22K
Escher would like this.
Robot tiles that let you walk forever in VR Researchers at the University of Tsukuba built a tile that moves under your feet as you walk. It uses sensors to read your gait and predict where your next step will land. ✅ Each tile slides into place before your foot touches down.
3
39
108
Congrats Jenny, huge milestone!
We are live. Today we are launching Robyn, the world's most emotionally intelligent AI partner designed to help billions of people feel more alive. I'm thrilled to announce that Robyn is backed by $5.5M in seed funding led by M13, with support from visionary investors
0
19
38
I'm quite annoyed that I can't get on any Forbes list without one of my brothers also being on it.
4
26
111
100M Gaussians. Streamed on a single RTX 3090. Welcome to @NVIDIAAIDev 's fVDB Reality Capture — bringing 3D Gaussian Splatting, volumetric data, and OpenUSD into one pipeline. 🎥 Full demo → https://t.co/eJMI4lSXUD
#NVIDIA #GaussianSplatting #3D
14
71
563
LLMs memorize a lot of training data, but memorization is poorly understood. Where does it live inside models? How is it stored? How much is it involved in different tasks? @jack_merullo_ & @srihita_raju's new paper examines all of these questions using loss curvature! (1/7)
10
133
808
I don't quite agree with this, but I like Eliezer's writing style, so I will read it.
"If Anyone Builds It, Everyone Dies" is now out. Read it today if you want to see with fresh eyes what's truly there, before others try to prime your brain to see something else instead!
2
24
59
We mine "how to" reasoning from solving lots of math problems into a shared procedural memory (skills). It's like providing a model shared workspace to write to and read from. At test time, we read this memory in context for self-improvement. https://t.co/xLcwezJAEm
The thing that is exciting *to me* about @AnthropicAI's "agent skills" announcement is that it provides a step toward continual learning. - Rather than continuously updating model weights, agents interacting with the world can continuously add new skills. - Compute spent on
2
4
39
The paper shows that continued training on junk social posts makes LLMs think worse. 😨 This big concern here is that because many teams keep topping up models with web crawls. Feeding highly popular or shallow posts can cause lasting, hard to undo damage. They ran controlled
8
21
99
Very proud of my brother Balazs for pioneering the theory of graphons. While the naming was introduced in a later paper, Laci Lovasz and Balazs defined the notion in their paper "Limits of Dense Graph Sequences" in 2006 that received the Fulkerson prize in 2012.
Graphons are mathematical objects that model the structure of massive networks. In machine learning, they provide a powerful framework for analyzing and generating large graphs. They are used to estimate the underlying structure of a network, predict missing links, and understand
6
59
275
Equating AI with chatbots is like equating IT with web search.
5
52
137
The term “AGI” is currently a vague, moving goalpost. To ground the discussion, we propose a comprehensive, testable definition of AGI. Using it, we can quantify progress: GPT-4 (2023) was 27% of the way to AGI. GPT-5 (2025) is 58%. Here’s how we define and measure it: 🧵
201
404
2K
Grok trading!
41
166
1K