Yiğit Demirağ
@yigit.ai
Research Scientist at Google. PhD from ETH Zürich. Exotic AI architectures and silicon.
👾 Zürich, Switzerland
👾 Zürich, Switzerland
“Playthings” is by far the best Black Mirror episode I’ve watched. Thronglets ❤️
May 8, 2025 at 9:33 PM
“Playthings” is by far the best Black Mirror episode I’ve watched. Thronglets ❤️
Research Scientist, Paradigms of Intelligence — Google Careers
www.google.com
May 8, 2025 at 9:23 PM
Reposted by Yiğit Demirağ
Why Turkey’s youth-led protests offer hope | Opinion
https://www.ft.com/content/8978c8b1-5bee-4398-a8fa-60403367b7d0
https://www.ft.com/content/8978c8b1-5bee-4398-a8fa-60403367b7d0
Why Turkey’s youth-led protests offer hope | Opinion
A new generation is finding brilliant and creative ways to object to the arrest of mayor Ekrem İmamoğlu
www.ft.com
April 5, 2025 at 5:21 AM
Why Turkey’s youth-led protests offer hope | Opinion
https://www.ft.com/content/8978c8b1-5bee-4398-a8fa-60403367b7d0
https://www.ft.com/content/8978c8b1-5bee-4398-a8fa-60403367b7d0
I appreciate @bsky.app once more when Elon starts blocking access to 100s of Twitter accounts engaged in expressing pro-democracy sentiments in Turkey.
April 5, 2025 at 4:15 AM
I appreciate @bsky.app once more when Elon starts blocking access to 100s of Twitter accounts engaged in expressing pro-democracy sentiments in Turkey.
Reposted by Yiğit Demirağ
Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n
February 4, 2025 at 6:54 PM
Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n
February 5, 2025 at 1:22 PM
If you’re passionate about brain-inspired algorithms/hardware and novel neural computation beyond current TPU/GPU stack please apply to the CapoCaccia Workshops for Neuromorphic Intelligence.
It values creativity, exploration and interdisciplinary collaboration🧪
capocaccia.cc/en/public/at...
It values creativity, exploration and interdisciplinary collaboration🧪
capocaccia.cc/en/public/at...
January 8, 2025 at 12:48 PM
If you’re passionate about brain-inspired algorithms/hardware and novel neural computation beyond current TPU/GPU stack please apply to the CapoCaccia Workshops for Neuromorphic Intelligence.
It values creativity, exploration and interdisciplinary collaboration🧪
capocaccia.cc/en/public/at...
It values creativity, exploration and interdisciplinary collaboration🧪
capocaccia.cc/en/public/at...
true but measuring compute is more fun.
January 7, 2025 at 9:21 PM
true but measuring compute is more fun.
Live ISS telemetry is interesting to watch. You can monitor critical sensors i.e, airlock or cabin pressures, or the urine tank percentage :)
ISS sensors in real time. iss-mimic.github.io/Mimic/
December 25, 2024 at 4:12 PM
Live ISS telemetry is interesting to watch. You can monitor critical sensors i.e, airlock or cabin pressures, or the urine tank percentage :)
I didn't properly practice winter sports during my 5-year PhD in Switzerland. This morning, I'm on the SBB train to LAAX to learn snowboarding in 5 days.
December 24, 2024 at 8:51 AM
I didn't properly practice winter sports during my 5-year PhD in Switzerland. This morning, I'm on the SBB train to LAAX to learn snowboarding in 5 days.
Reposted by Yiğit Demirağ
1/ Okay, one thing that has been revealed to me from the replies to this is that many people don't know (or refuse to recognize) the following fact:
The unts in ANN are actually not a terrible approximation of how real neurons work!
A tiny 🧵.
🧠📈 #NeuroAI #MLSky
The unts in ANN are actually not a terrible approximation of how real neurons work!
A tiny 🧵.
🧠📈 #NeuroAI #MLSky
Why does anyone have any issue with this?
I've seen people suggesting it's problematic, that neuroscientists won't like it, and so on.
But, I literally don't see why this is problematic...
I've seen people suggesting it's problematic, that neuroscientists won't like it, and so on.
But, I literally don't see why this is problematic...
This would be funny if it weren't sad...
Coming from the "giants" of AI.
Or maybe this was posted out of context? Please clarify.
I can't process this...
Coming from the "giants" of AI.
Or maybe this was posted out of context? Please clarify.
I can't process this...
December 16, 2024 at 8:03 PM
Nobel lecture in economic sciences from Daren Acemoglu is about to start
YouTube
Share your videos with friends, family, and the world
www.youtube.com
December 8, 2024 at 1:01 PM
Nobel lecture in economic sciences from Daren Acemoglu is about to start
Today is Gemini's 1st birthday 🎂, and the new experimental model, gemini-exp-1206 is #1 across the board in LMSYS Chatbot Arena.
December 6, 2024 at 8:26 PM
Today is Gemini's 1st birthday 🎂, and the new experimental model, gemini-exp-1206 is #1 across the board in LMSYS Chatbot Arena.
if you see this, post a knight
December 5, 2024 at 2:06 AM
if you see this, post a knight
ASML in Europe builds one of the most complex and precious engineering artifact, EUV lithography machines, and sit at the root of modern tech tree.
Computational lithography: Driving nanometer precision in microchip manufacturing | ASML
YouTube video by ASML
youtu.be
November 30, 2024 at 10:38 AM
ASML in Europe builds one of the most complex and precious engineering artifact, EUV lithography machines, and sit at the root of modern tech tree.
Reposted by Yiğit Demirağ
Why does #compneuro need new learning methods? ANN models are usually trained with Gradient Descent (GD), which violates biological realities like Dale’s law and log-normal weights. Here we describe a superior learning algorithm for comp neuro: Exponentiated Gradients (EG)! 1/12 #neuroscience 🧪
October 28, 2024 at 5:18 PM
Why does #compneuro need new learning methods? ANN models are usually trained with Gradient Descent (GD), which violates biological realities like Dale’s law and log-normal weights. Here we describe a superior learning algorithm for comp neuro: Exponentiated Gradients (EG)! 1/12 #neuroscience 🧪
My morning routine now includes practicing latte art with my flat white at the Google MKs.
November 28, 2024 at 8:45 AM
My morning routine now includes practicing latte art with my flat white at the Google MKs.
A good loss landscape: Switzerland 🇨🇭
November 24, 2024 at 8:42 PM
A good loss landscape: Switzerland 🇨🇭
After a PhD with Python, returning to C++ for a good reason :)
November 24, 2024 at 7:15 PM
After a PhD with Python, returning to C++ for a good reason :)
Google Research Zürich is a magical place quite like Hogwarts. Every wizard I meet works on powerful spells and potions.
August 19, 2023 at 1:56 AM
Google Research Zürich is a magical place quite like Hogwarts. Every wizard I meet works on powerful spells and potions.
Quite a candy for my optimization appetite :) As majority of inference are still on CPUs on mobile/edge accelerators, Mojo will be interesting to watch closely. I wonder how well it will support Triton or CUDA
https://youtu.be/6GvB5lZJqcE
https://youtu.be/6GvB5lZJqcE
youtu.be
May 4, 2023 at 1:08 AM
Quite a candy for my optimization appetite :) As majority of inference are still on CPUs on mobile/edge accelerators, Mojo will be interesting to watch closely. I wonder how well it will support Triton or CUDA
https://youtu.be/6GvB5lZJqcE
https://youtu.be/6GvB5lZJqcE
@vedatmilor.bsky.social Hosgeldiniz Vedat Bey :)
April 20, 2023 at 9:27 PM
@vedatmilor.bsky.social Hosgeldiniz Vedat Bey :)
I release a minimal (<150 lines) JAX implementation of "Gradients without Backpropagation" paper. It proposes a simple addition to forward AD to estimate unbiased gradients during single inference pass (quick project, might be further optimized)
https://github.com/YigitDemirag/forward-gradients
https://github.com/YigitDemirag/forward-gradients
github.com
April 18, 2023 at 2:49 PM
I release a minimal (<150 lines) JAX implementation of "Gradients without Backpropagation" paper. It proposes a simple addition to forward AD to estimate unbiased gradients during single inference pass (quick project, might be further optimized)
https://github.com/YigitDemirag/forward-gradients
https://github.com/YigitDemirag/forward-gradients