Skip to content
Main Site News Console

NVIDIA’s New AI Is an Efficiency Monster

· Two Minute Papers Translated
视频论文

返回

搜索

Use your voice to search

Sign in

Video 1

NVIDIA’s New AI Is an Efficiency Beast

Tap to unmute

2x

NVIDIA’s New AI Is an Efficiency Beast

Two Minute Papers 49,290 views 7 days ago

Copy link

Info

Shopping

Image 4

If the video doesn’t start playing right away, try restarting your device.

•

You’re signed out

Videos you watch may be added to the TV’s watch history and influence TV recommendations. To avoid this, cancel and sign in to YouTube on your computer.

Cancel Confirm

- [x] Include playlist

Error retrieving sharing information. Please try again later.

Image 6

0:00

0:00 / 0:00

Live

•Watch full video

•

•

Sign in to confirm you’re not a robot. This helps protect our community. Learn more

Sign in

NVIDIA’s New AI Is an Efficiency Beast

Image 7: Two Minute Papers

Two Minute Papers

Two Minute Papers

1.81M subscribers

Join

Subscribed

1.7K

Save

Download

Download

49K views 7 days ago#nvidia

49,290 views • May 13, 2026 • #nvidia

❤️ Check out Lambda and sign up for their GPU cloud service here: https://lambda.ai/papers 📝 Paper here: ……more

……more

Transcript

Follow along with the transcript.

Show transcript

Show less

NVIDIA’s New AI Is an Efficiency Beast

49,290 views 49K views

May 13, 2026

1.7K

Save

Download

Download

120 comments

Sort comments

Sort by

Image 12: Default profile photo

Add a comment…

Image 13

Pinned by @TwoMinutePapers

@fateless_shadow

7 days ago

What a wonderful time to be alive!

Show less Read more

Like

102

Dislike

Image 14

❤ Liked by @TwoMinutePapers

Reply

3 replies

Hide replies

3 replies

Image 15

@CognitiveAxiom

7 days ago

Every Two Minute Papers video feels like a patch notes update for the real world

Show less Read more

Like

44

Dislike

Reply

1 reply

Hide replies

1 reply

Image 16

@nyny

7 days ago

Thanks for being so supportive of open source and open science!

Show less Read more

Like

73

Dislike

Image 17

❤ Liked by @TwoMinutePapers

Reply

Image 18

·

1 reply

Image 19

·

Hide replies

Image 20: Two Minute Papers

·

1 reply

Image 21

@pandoraeeris7860

7 days ago

What an AI era!

Show less Read more

Like

33

Dislike

Reply

3 replies

Hide replies

3 replies

Image 22

@SteveL-T100

15 hours ago

I think I’ve been watching your videos since at least 2010. What a great time to be alive, right? Thanks for everything you do.

Show less Read more

Like

Dislike

Reply

Image 23

@clownsalad

6 days ago

That “Hold on to your papers” game looks awesome. I want to play it

Show less Read more

Like

1

Dislike

Reply

Image 24

@havocthehobbit

7 days ago

Local models have really gotten to the point where I’ve enjoyed using them over the past month. I remember trying when Meta first opened up Llama and thinking, this is terrible; but I’ve tried Qwen 3.6 and Gemma 4 now, and it’s made me realize that these massive data centers being built right now may end up being a little wasteful, because in a few years they may only be used as spillover capacity for models, and the priority might shift to local-first. Inference and quantization have made even smartphone- and Raspberry Pi-sized models viable. I’m actually enjoying agent development on local models now; before, I mostly used large cloud models as a Stack Overflow. As Carlo always says: “What a wonderful time to be alive!”

Show less Read more

Like

15

Dislike

Image 25

❤ Liked by @TwoMinutePapers

Reply

Image 26

@radeksparowski7174

2 days ago

It needs to be plug-and-play to go mainstream. At this point, I thought we’d already have Jarvis-like AI assistants that are as out-of-the-box as Linux distributions: voice input and output, basic browser tasks, permanent memory, all in one package; AI can self-diagnose and propose solutions to bottlenecks, and can download and install upgrades while running or by rebooting itself… Technically, this is already possible, but it takes a lot of tuning for specific hardware. We may get there by the end of the year, but this really should already be mainstream on GB10 / Strix Halo / Snapdragon X or Elite / M4-5, and NVIDIA’s upcoming N1X.