Burc Eryilmaz

Senior Deep Learning Computer Architect @NVIDIA #AI #AIhardware #AIalgorithms, previously PhD @stanford in EE and CS #neuromorphiccomputing #resistivememory

[Burc Eryilmaz @burceryilmaz]

Amazing work

[NVIDIA AI Developer @NVIDIAAIDev]

On the latest round of #MLPerf v1.0 training submissions NVIDIA improved up to 2.1x on a chip-to-chip basis and up to 3.5x at scale, setting 16 performance records. Learn how: https://t.co/fYGc4wZrFG

[NVIDIA GTC @NVIDIAGTC]

Today, we’re introducing a new kind of computer. The basic building block of the modern data center. Meet the NVIDIA Grace #CPU. #GTC21 https://t.co/VGIVv9vsVJ

[Hannah Ritchie @_HannahRitchie]

Most deforestation is driven by agriculture.

Beef is the leading cause – responsible for over 40%. Most of this in Brazil.

Soybeans and palm oil are also significant drivers.

Our latest article on what's causing deforestation: https://t.co/N3MHESUzDw

[Bryan Catanzaro @ctnzr]

The new 80GB A100 GPU has more than 2 TB/s memory bandwidth!
We’ve already started using these @NVIDIAAI to train Megatron language models. 🚀

https://t.co/E2aBB8hajh

[Ben Bartlett @bencbartlett]

Recursively rotating segments of an image rotates the image itself

🔗 Code by rvizzz on GitHub: https://t.co/HlzQKMH8I5

[Burc Eryilmaz @burceryilmaz]

@trengriffin @mims Very good point. Article, although has valid points, is a bit misleading. NVIDIA takes AI computing beyond what Moore's law offers through innovations in architecture, algorithms and software, but Moore's law isn't dead at all.

[Bryan Catanzaro @ctnzr]

Jensen Huang is a time machine who lives in the future. So much of @nvidia’s great success over the past 5 years happened because of the strategic decisions that he made 15 years ago along with his continuing investments to bring them from vision to reality.

[Burc Eryilmaz @burceryilmaz]

Innovations in AI hardware and algorithms that map on them efficiently are key enablers of large scale AI.

[hardmaru @hardmaru]

“DeepSpeed can train a language model with 1 trillion parameters using as few as 800 NVIDIA V100 GPUs”

Even “with a single V100, our users can run models of up to 13 billion parameters without running out of memory, while obtaining competitive throughput.”https://t.co/aAtrcF663A

[Burc Eryilmaz @burceryilmaz]

@hardmaru Brain works in a somewhat different way than modern HW: it is probabilistic, very sparse, asynchronous, slower 'clock', and we don't know everything about how it learns at large scale.
But now try imagining what we could achieve with a MW once we reach brain efficiency on chips.

[Kısaca Kimdir @kisacakimdir]

Dr. Burç Eryılmaz

Lisans Bilkent. Master ve doktorayı Stanford'da EE ve Bilgisayar Müh. alanında nöromorfik hesaplama üzerine yaptı. MIT'de araştırmacı olarak bulundu.

Şu an ABD, NVIDIA'da yapay zeka için verimli algoritmalar, çip ve süper bilgisayar mimarisi üzerine çalışıyor.

[Burc Eryilmaz @burceryilmaz]

@kisacakimdir Sayfanızda yer ayırdığınız için teşekkürler, umarım diğer gençlerimize de motivasyon kaynağı olur.

[Luigi Cruz @luigifcruz]

Nvidia is bringing one more technology from the datacenter to the desktop with RTX 3000 series. It’s called RTX IO and will be huge for Data Science! The GPU can directly access storage without latency from the CPU. I can't wait to see what folks @rapidsai can do with it!

[İlber Ortaylı @ILBERORTAYLIGSU]

Etrafındakiler “duralım” derken Atatürk “ilerliyoruz” dedi. #30AgustosBaskumandanZaferi

[Burc Eryilmaz @burceryilmaz]

Fun fact: Edges, angles and the area of a triangle can't be all rational numbers. This is one of the most interesting things I remember from high school since we were asked to prove this in Turkish Math Olympiad back in 2005.

[Burc Eryilmaz @burceryilmaz]

Black Panther was my favorite Marvel movie. RIP Chadwick Boseman.

[Marques Brownlee @MKBHD]

💔

[Burc Eryilmaz @burceryilmaz]

Türkçesi: COVID-19'u atlatanlarin çoğunda kalp hasarı meydana geldiği görülmüş ve kalp hücreleri kendini yenileyemediginden hasarın kalıcı olacağı öngörülüyor.

[Yaneer Bar-Yam @yaneerbaryam]

"Recent reports have indicated that aside from fatal cases of COVID-19, a majority of patients who recover from COVID-19 display significant levels of cardiac damage (~80%)[17,38,39] and nearly half of all mildly ill COVID-19 patients have echo abnormalities[17,40–42].

[Burc Eryilmaz @burceryilmaz]

Great work by NVIDIA research teams!

[NVIDIA AI Developer @NVIDIAAIDev]

NVIDIA is presenting over 20 papers at #ECCV2020. One of them, COCO-FUNIT, is a computer vision model that can translate an image of an animal such as a puppy, and convert it into a snow leopard. See the complete list of papers here: https://t.co/Yu1djP6zfi

[Sam Altman @sama]

Vector theory of impact:

[Jeff Dean @JeffDean]

Reed-Solomon codes are cool.

They are also used in distributed file systems to be robust to disk crashes & temporary machine availability. Break data block into, say, 5 pieces, generate 3 additional encoded pieces: can read if any 5 of 8 are available, at 60% space overhead.