i'm baffled why everybody is concentrating on computer vision because of robotics, but they're not concentrating on the power of reasoning and making it possible with AI. Because there's no point in having a robot if it is as dumb as a door knob. LLM is the interface for the robotics hardware, so get stage one right.. the LLM
I have a jetson yet when LLMs hit the market I stopped playing around with it over night as LLMs changed the way AI is thought of. I would love to have a new version of Jetson that had a smaller LLM locally!! Jetbot version 2? I can't wait!
What would it even mean to use an LLM with motion sensors? That's pretty interesting - like could you feed it IMU data and tell the LLM that you wanted it to be alerted if it was dropped or if someone was running vs walking with the device?
video object detection comparison between nvidia T4 16GB (320 Tensor Cores) vs Orin 64GB (64 Tensor cores) would be really good, as T4 is also low power consumption GPU
i'm still looking all over RUclips to find out if the Nvidia AGX orin jetson 64 gb devkit can be used for running local language models but they're only thing I've seen is somebody using the nano and it's the same speed as a raspberry pi 5 which is ridiculous because it's five times the price at least. i'm also doubting the claims made by the CAO in the Keynote, ever since I saw that large companies ended up with the new Nvidia chip overheating I mean the Blackwell. If you put 100 of them together there's a problem they overheat. So nVidia has been announcing things a little bit early before the storm
![Noob To Max With DRAGON REWORK In Blox Fruits [FULL MOVIE]](http://i.ytimg.com/vi/LnBMOinoOvA/mqdefault.jpg)
Big kudos to Dustin for his brilliant communicative skills! You're special, man!
i'm baffled why everybody is concentrating on computer vision because of robotics, but they're not concentrating on the power of reasoning and making it possible with AI. Because there's no point in having a robot if it is as dumb as a door knob. LLM is the interface for the robotics hardware, so get stage one right.. the LLM
What a fantastic overview of what's possible! This tracks a lot of components I've been working on and I can't wait to merge in some of these modules.
that's incredible! amazing presentation thank you
Does jetpack 6 support the older Xavier platform?
This is incredible. Absolutely futuristic.
Amazing, great video, love the demo. Can I also run these demos on my Jetson Orin?
Awesome job Dusty!
Would be far more impressive if jetpack 6 also supported the jetson nano platform as well.
I have a jetson yet when LLMs hit the market I stopped playing around with it over night as LLMs changed the way AI is thought of. I would love to have a new version of Jetson that had a smaller LLM locally!! Jetbot version 2? I can't wait!
Will all the examples he showed run on Jetson Orin Nano Developer Kit?
This is awesome! thanks for the information and great examples!!!!
Thanks for the video, great developments! Is it possible to run those on Xavier NX ? Has anyone tried?
What would it even mean to use an LLM with motion sensors? That's pretty interesting - like could you feed it IMU data and tell the LLM that you wanted it to be alerted if it was dropped or if someone was running vs walking with the device?
Nice post with regards By Adv T E Barat Bushan Senior Advocate Member of MHAA Chennai
This is amazing, I think its really wonderful and If we could talke to videos too this could be of great help
video object detection comparison between nvidia T4 16GB (320 Tensor Cores) vs Orin 64GB (64 Tensor cores) would be really good, as T4 is also low power consumption GPU
Wow, that's impressive!
cool, im interested in autonomous virtual humans 🤔
A great talk.
i'm still looking all over RUclips to find out if the Nvidia AGX orin jetson 64 gb devkit can be used for running local language models but they're only thing I've seen is somebody using the nano and it's the same speed as a raspberry pi 5 which is ridiculous because it's five times the price at least. i'm also doubting the claims made by the CAO in the Keynote, ever since I saw that large companies ended up with the new Nvidia chip overheating I mean the Blackwell. If you put 100 of them together there's a problem they overheat. So nVidia has been announcing things a little bit early before the storm
😍
How in the hell do you expect to do any real development without conda support ? I picked one in hopes to this is depressing.
There’s always virtualenv.