Activity
-
🧑🎨 Our design team is growing! We’re looking for an experienced product designer to lead our design team and help us build the best AI assistant…
🧑🎨 Our design team is growing! We’re looking for an experienced product designer to lead our design team and help us build the best AI assistant…
Liked by Yann LeCun
-
I've been playing with SambaNova Systems's API serving fast Llama 3.1 405B tokens. Really cool to see leading model running at speed. Congrats to…
I've been playing with SambaNova Systems's API serving fast Llama 3.1 405B tokens. Really cool to see leading model running at speed. Congrats to…
Liked by Yann LeCun
-
Announced this morning and verified by Artificial Analysis, Cerebras Systems Inference is capable of serving Llama 3.1 70B at 450 tokens/sec and…
Announced this morning and verified by Artificial Analysis, Cerebras Systems Inference is capable of serving Llama 3.1 70B at 450 tokens/sec and…
Liked by Yann LeCun
Erleben Sie
Publications
-
Feature Learning and Deep Architectures: New Directions for Music Informatics
Journal of Intelligent Information Systems / Springer
-
Moving Beyond Feature Design: Deep Architectures and Automatic Feature Learning in Music Informatics
Proc. of the International Society of Music Information Retrieval (ISMIR) 2012
Projects
-
LAGR
-
LAGR was a DARPA competition in autonomous vision-based off-road robot navigation. Nine labs competed in many races through complex off-road terrain over the course of 3 years. In the first year, our team achieved the competition's 3-year goal of improving the baseline run-time by a factor of 2, forcing the government to up the goal to a factor of 3.
Using visual input for navigation is essential in unstructured outdoor environments, which can cause simpler sensors such as GPS…LAGR was a DARPA competition in autonomous vision-based off-road robot navigation. Nine labs competed in many races through complex off-road terrain over the course of 3 years. In the first year, our team achieved the competition's 3-year goal of improving the baseline run-time by a factor of 2, forcing the government to up the goal to a factor of 3.
Using visual input for navigation is essential in unstructured outdoor environments, which can cause simpler sensors such as GPS, odometry, and laser scanners to drift severely or become unusably noisy. As an alternative to such sensors, I contributed a novel, fast form of visual odometry. Our efficient “hybrid visual odometry” method used 5% of the CPU time, in contrast to full VO methods which require an entire CPU. Despite this, it maintained comparable accuracy to full VO over diverse terrain types. This makes it particularly suitable for consumer robots, extraterrestrial rovers, and other platforms with limited computing power.
This contribution was described in the ICRA 2009 paper Efficient Off-Road Localization Using Visually Corrected Odometry, and contributed to our JFR 2009 paper, A Multi-Range Architecture for Collision-Free Off-Road Robot Navigation.Other creatorsSee project -
DjVu
-
DjVu (pronounced "déjà vu") is a digital document format with advanced compression technology.
I initiated and directed the project.Other creatorsSee project -
Check Recognition with Deep Learning
-
The first practical system for reading the amount on bank checks.
It was deployed by NCR first in ATM machines in 1995, then in high-speed check reading machines in 1996. At some point in the late 90's these systems read an estimated 10 to 20% of all the checks in the US.
It used a convolutional neural net, combined with a so-called graph-transformer network, a form of non conditional random field in which the solution space can be represented by a treillis.
Other creatorsSee project
More activity by Yann
-
Right on the heels of the update I shared on adoption of Llama yesterday, we’re also seeing FAST growth in usage of Meta AI across our family of apps…
Right on the heels of the update I shared on adoption of Llama yesterday, we’re also seeing FAST growth in usage of Meta AI across our family of apps…
Liked by Yann LeCun
-
one thing I've repeatedly seen and learned at Meta was that the world is large and my tech bubble can't accurately predict traction that spans…
one thing I've repeatedly seen and learned at Meta was that the world is large and my tech bubble can't accurately predict traction that spans…
Liked by Yann LeCun
-
Notebooks 💻 for Lecture 6 on Graph-based Visualization https://lnkd.in/g4Nxhjhj Lab1: PCA Lab2: Robust PCA Lab3: LLE Lab4: Laplacian…
Notebooks 💻 for Lecture 6 on Graph-based Visualization https://lnkd.in/g4Nxhjhj Lab1: PCA Lab2: Robust PCA Lab3: LLE Lab4: Laplacian…
Liked by Yann LeCun
-
Proud to see Nabla named as a top "boring AI" startup 😂
Proud to see Nabla named as a top "boring AI" startup 😂
Liked by Yann LeCun
-
Deployed on a local cluster, Infosys incorporated Llama 3 as part of an in-house legal assistant app to keep their most sensitive data secure. Beyond…
Deployed on a local cluster, Infosys incorporated Llama 3 as part of an in-house legal assistant app to keep their most sensitive data secure. Beyond…
Liked by Yann LeCun
-
there's a bunch of phenomenal distributed / scaling talks and posters lining up at the PyTorch Conference in September! the conference is getting…
there's a bunch of phenomenal distributed / scaling talks and posters lining up at the PyTorch Conference in September! the conference is getting…
Liked by Yann LeCun
-
Pioneering a new era of AI, Llama is at the forefront of openness, customizability, and affordability. We're dedicated to developing in the open…
Pioneering a new era of AI, Llama is at the forefront of openness, customizability, and affordability. We're dedicated to developing in the open…
Liked by Yann LeCun
-
Ready to channel your AI skills into a meaningful cause? Join us for the AI for Ukraine Recovery Hackathon happening simultaneously in 🇺🇦 Ukraine…
Ready to channel your AI skills into a meaningful cause? Join us for the AI for Ukraine Recovery Hackathon happening simultaneously in 🇺🇦 Ukraine…
Liked by Yann LeCun
-
La France peut-elle vraiment devenir un leader de l’IA ? Il y a 1 an, la Première ministre a mis en place une commission IA chargée de définir des…
La France peut-elle vraiment devenir un leader de l’IA ? Il y a 1 an, la Première ministre a mis en place une commission IA chargée de définir des…
Liked by Yann LeCun
-
Exciting news: Andrew Ng and Yann LeCun have joined Korea's National AI Committee as advisors! I had the privilege of joining Andrew for the meeting…
Exciting news: Andrew Ng and Yann LeCun have joined Korea's National AI Committee as advisors! I had the privilege of joining Andrew for the meeting…
Liked by Yann LeCun
-
Meta's Llama🦙 has become the dominant platform in the AI ecosystem. An exploding number of companies large and small, startups, governments, and…
Meta's Llama🦙 has become the dominant platform in the AI ecosystem. An exploding number of companies large and small, startups, governments, and…
Shared by Yann LeCun
-
Open source AI is the way forward and today we're sharing a snapshot of how that's going with the adoption and use of Llama models. Read the full…
Open source AI is the way forward and today we're sharing a snapshot of how that's going with the adoption and use of Llama models. Read the full…
Liked by Yann LeCun
-
Excited to share a great profile of our VP, Tehmina Masud, in BioPharma-Reporter, as part of their “Women in Science” series. As head of Target…
Excited to share a great profile of our VP, Tehmina Masud, in BioPharma-Reporter, as part of their “Women in Science” series. As head of Target…
Liked by Yann LeCun
-
Introducing our new webinar series: PyTorch Expert Exchange 💡 Join Muyang Li live on August 30 at 10am PT to discuss efficient generative…
Introducing our new webinar series: PyTorch Expert Exchange 💡 Join Muyang Li live on August 30 at 10am PT to discuss efficient generative…
Liked by Yann LeCun
-
If each step of an ai agent is 95% accurate. None of the 30 step work flows will work. Going from 95-> 99.9 is a similar last mile problem as with…
If each step of an ai agent is 95% accurate. None of the 30 step work flows will work. Going from 95-> 99.9 is a similar last mile problem as with…
Liked by Yann LeCun
Other similar profiles
Gemeinsame Artikel erkunden
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Yann LeCun
2 others named Yann LeCun are on LinkedIn
See others named Yann LeCun