FANVince

@fxtrader @grok 总结一下，结合重点和新变化，推荐明天，一周内，和一个季度内需要关注的股票和行业板块

@VilleKuosmanen looks like you are using a RGB-D? or r u just used the RGB data?

Trained a new ACT model with 200 new episodes using a new (better) data collection method & more diversity. Resulting model handles different cup positions and performs better than model with 4x more data but less diversity.

My mental model of Sora is that it is the “GPT-2 moment” for video generation. GPT-2, which came out in 2018, could generate paragraphs of text that are coherent and grammatically correct. GPT-2 wasn’t able to write an entire essay without making mistakes like being inconsistent or hallucinating facts, but it spurred subsequent generations of models. In less than five years since GPT-2, GPT-4 is now able to grok skills like chain-of-thought or writing long essays without hallucinating. In the same way, Sora today can generate short videos that are artistic and realistic. Sora is currently not able to generate a 40-minute TV show with consistent characters and a compelling storyline. However, I believe that skills like maintaining long-term consistency, having near-perfect realism, and generating substantive storylines will emerge in the next generations of Sora and other video generation models. A few predictions about how this will play out: - Video is not as information-dense as text, and so it will take way more compute and data to learn skills like reasoning via video - As a result, leveraging other modalities as correlated information with video will be critical to bootstrapping the learning process - There will be massive competition for high-quality video data, just as there is for high-quality text datasets - AI researchers with experience in video will be in high demand, but they’ll have to adapt to new paradigms just as the traditional NLP researchers have had to adapt to the success of scaling language models - Disruption of the movie industry will play out similar to how GPT-4 has changed writing (as a tool and aid that surpasses average quality, but will still be far from the work of professionals)

@iainherd @haltor exactly...

@haltor By year-end someone will be building the first ski/bike/skate helmet designed for AVP

I may have invented a more comfortable Apple Vision Pro strap with just a baseball cap and a high tech rubber band. It's not touching my face at all. Higher fov, better weight distribution, open periphery and eye tracking still works. It's super comfy. The cap's brim is carrying all the weight. And no VR hair.

This is a helmet...

The launch ad for Apple Vision Pro vs the launch ad for the iPhone. Love to see this parallel. Great ad!

MoE, Mamba, beside Mobile Aloha and LangSplat... The speed of the revolution of AI has entered into another level~

Overfitting in embodied agents😅

Andrej Karpathy is a legendary researcher who helped start OpenAI and created Stanford's first deep learning class. @karpathy's advice on how to learn AI: (1) 10,000 hours of deliberate practice will make you an expert. You can iterate as you work. Only compare yourself to the past, not to others. (2) Don't worry about making mistakes. You build intuitions on what is useful vs. not useful- they are not dead work. (3) Teach to strengthen your understanding and find gaps in knowledge. Similar to "If you can't explain it to a six-year-old, then you don't understand it yourself" - Albert Einstein.

So exited to be part of the great partnership! We are opening a great entrance to the fantastic web 3.0 world🚀🚀 Much more is coming, right, let's wait and see 🔥🔥🔥

@MetaverseMoon announced #Whitelist 🏆250 winners 🏆 Allocation 500$ To participate complete form: gleam.io/HNphd/whitelis… The form will close on December 9 at 10.00 AM UTC

Nowadays, everyone is using the word non-fungible like breathing air... #NFTs

Decrypt this post with #mask_io ! mask.io/?PostData_v1=%…

this is most insane thing ive seen on tik tok

testtest

test

testtest

iOS 7 Photoshop and Sketch Template ios.robs.im #ios #icon #template #psd #sketch