Deep learning requires fundamentally new kinds of hardware | Jim Keller and Lex Fridman

Показать описание

Please support this podcast by checking out our sponsors:

GUEST BIO:
Jim Keller is a legendary microprocessor engineer, previously at AMD, Apple, Tesla, Intel, and now Tenstorrent.

PODCAST INFO:

SOCIAL:

Рекомендации по теме

Комментарии

Jim and Lex always have really great discussions.

johnkost

So this is the other way around. Classically coders have to optimize their code to specific hardware. But here they are optimizing hardware to run their (pytorch) programs most efficiently.

ramanmono

Excellent sound tangible. A tactical tactile sphere of coherence within symbiosis in biomorphic an molecular structures such as spectrums tone an form throughout (Singularity) Upgrade !

dougbaker

I don’t even know what just happened..

markjackson

Can someone explain like I'm 5: "the future of software is data networks..the programs"
Is he literally saying the future of software is basically pools of already functioning data networks, networked together to form a larger solution? Like, his example of pixels running their programs, but the pixels themselves have no idea what the other pixels are doing...and this will solve that.

Did I understand that correctly?

smudge

Jensen says you just need to buy more 3090s.

bobtoo

A fast Hadamard tranform chip could be made using a few thousand transistors. Even a neural network chip based on that would need less transistors (56000) than a Z80 CPU from the 1970s. Eg. Fast Transform fixed filter bank neural networks.

hoaxuan

The problem with not getting the basics down 1oo%. With not doing the scientific methodology 100% from the ground up is that very expensive, elaborate hardware gets built. When something far simpler would be better.

hoaxuan

i learn more listening to mr. keller than of my school.

ndx

My favourite topic AI Accelerators and ML on Chip

bhuvaneshs.k

Tenstorrent chip sounds like parallel processing running, interrelated, spatial threads. Getting closer to how the human mind works... it's only a matter of time now, bois. Tenstorerent --> realizing the capability for Software 2.0 --> Skynet.

brasidas

what hes trying to say is the Von neumann architecture will not natively work with neural networks.

HavokR

Anyway we already live in a world of talking sand. Magic eh?

hoaxuan

Fun fact:

Jim Keller is brother in law of Jordan B. Peterson.

miroslavdanilov

I feel dumber just from listening to him talk... the IQ of everyone listening just dropped 50 points.

Jk

drhilltube

Deep learning requires fundamentally new kinds of hardware | Jim Keller and Lex Fridman

Deep learning requires fundamentally new kinds of hardware | Jim Keller and Lex Fridman

Deep Learning | What is Deep Learning? | Deep Learning Tutorial For Beginners | 2023 | Simplilearn

Neural Networks Explained in 5 minutes

Universal Approximation Theorem - The Fundamental Building Block of Deep Learning

What are Convolutional Neural Networks (CNNs)?

Generative AI is not the real future! - New AI paradigm/brand

Day1 Session 1 | Deep Learning: Basics, history, and fundamental concepts. | #nrcm

Deep learning for fundamental sciences using high-performance computing (O’Reilly AI Conference)

Introduction to Model Evaluation in Deep Learning | AIML End-to-End Session 190

Fundamental Challenges in Deep Learning for Stiff Contact Dynamics

PROF. DR. DAGMAR MONETT - Overcoming fundamental problems | Rise of AI conference 2019

Fundamental cognitive units (Francois Chollet)

Neural Network In 5 Minutes | What Is A Neural Network? | How Neural Networks Work | Simplilearn

The fundamental theorem of calculus (fast AI lesson)

Coursera vs Simplilearn - 2025 | Which Should You Choose For Machine Learning?

Need for Speed: Machine Learning in the Era of Real-Time

Transformers Need Glasses!

(Part2 /???) Miss Circles Past AU || Fundamental Paper Education

Fundamental concepts of AI, machine learning, and deep learning

Helmut Bölcskei - Fundamental limits of deep neural network learning

Tensors: The Fundamental Atoms of PyTorch

Unlocking the power of AI: A fundamentally different approach to building intelligent systems

every tool fundamentally can get sidetracked

Forecasting Kafka Lag Issues with Machine Learning