The 'AR-Cloud' will change the way we view the physical environment.
Check out this world-scale AR scene, powered by
Pinned Tweet
0:46
30.6K views
19
193
519
Show this thread
Follow
Click to Follow TweetEdMiller
Edward Miller
@TweetEdMiller
LondonJoined June 2012
Edward Miller’s Tweets
LLMs like those used by ChatGPT are missing a crucial language for faithfully interpreting the physical world, without which, they have a ceiling for what questions can be answered/predicted.
Quote Tweet
LLMs have no physical intuition as they are trained exclusively on text.
They may correctly answer physical intuition questions *if* they can retrieve answers to similar questions from their vast associative memory.
But they may get the answer *completely* wrong
1/
Show this thread
2
3
One of the hardest challenges in developing AI for autonomous vehicles is evaluating the performance of our driving models. Why? (A short 🧵on our latest research on multi-agent RL).
0:33
46.6K views
9
86
467
Show this thread
It looks like Gordon Bell's 1998 vision of 'MyLifeBits' has found its way into the world. I believe this will be a significant theme over the next decade en.wikipedia.org/wiki/MyLifeBits
Quote Tweet
Introducing Rewind
Today we are unveiling Rewind: the search engine for your life. It’s a macOS app that enables you to find anything you’ve seen, said, or heard.
Here’s how it works:
(5/23)
youtube.com/watch?v=dIV0Zi
Show this thread
1
2
Authenticated identity is a necessity not a premium feature for many web3 applications. Seems Elon is in a rare situation where tying identity to account is actually desired by a lot of users. Will be interesting to see how this plays out.
1
4
Fun fact: in 2015 I registered the domain name MetaverseFestival.com as I thought there was going to be a future where immersive virtual worlds could be traversed with ease, and that one day it would be cool to host a global virtual festival to celebrate humanity's diversity 🙃
3
Massive kudos to and the team at Cognitive Assistance Lab for your work to enable audio wayfinding for people who are visually impaired, announced at Connect today.
Extended video of the partnership here, featuring yours truly 😊facebook.com/TechatMeta/vid
GIF
5
19
Should've called it Top Notch 👌
2
Excellent turnout for our tutorial on egocentric vision research using #projectaria at #CVPR2022. Learn about how Project Aria is being used to advance AI & ML research, including concretizing privacy challenges for machine perception in AR - agenda @ ariatutorial2022.github.io
2
7
40
For anybody lucky enough to be joining this year, join our team’s tutorial on June 19th to learn more about the dataset and how you can work with it (Details at ariatutorial2022.github.io) or learn more at about.facebook.com/realitylabs/pr
/end
1
2
19
Show this thread
In addition to “Daily Activities”, the dataset also contains 16 sequences of “Desktop Activities” captured with a multi-cam setup, for further accelerating research into human-object interactions.
7/n
GIF
1
2
2
Show this thread
… and speech-to-text annotation for research into audio/visual diarization.
6/n
1
1
2
Show this thread
…per-frame eye tracking to study human/object interactions…
5/n
GIF
1
1
4
Show this thread
…shared space-time information between multiple wearers…
4/n
GIF
1
1
3
Show this thread
In addition to sensor data, the Pilot Dataset also contains results from machine perception services which provide additional context to the spatial-temporal reference frames, such as:
Accurate 3D trajectories of users across multiple everyday activities in the same location
3/n
GIF
1
2
6
Show this thread
The Project Aria Pilot Dataset consists of 143 sequences of “Daily activities”, consisting of over 7 accumulated hours of data captured on Aria with additional face blurring to demonstrate our commitment to responsible innovation.
2/n
2
1
7
Show this thread
Ahead of #CVPR2022 , I’m excited to share the open dataset of Project Aria data from Meta Reality Labs, along with accompanying open research tools designed to accelerate AI and ML research. about.facebook.com/realitylabs/pr
A little about the dataset and why I think it’s so exciting…
1/n
GIF
4
62
204
Show this thread
"Any sufficiently advanced technology..." 🧙♂️🪄 😝🤟
This is wicked.
Quote Tweet
This @clipdropapp × @figma plugin is truly mindblowing
It can automatically decompose images into multiple layers. No more issues with images not fitting a layout 
Let me know if you'd like to try it
cc @zoink #FigmaCommunity #Config2022
Let me know if you'd like to try it
cc @zoink #FigmaCommunity #Config2022Show this thread
0:16
20.1K views
1
6
With our new measurement methodology, we’ve halved the testing time needed to obtain robust insight.
Find out how we're using the latest methods in data science and ML to accelerate the speed at which we can test and evaluate driving models on the road.
9
20
Interesting provocation about how AR can be used as the interface to drive and visualize the future intention of robotic devices.
Quote Tweet
When we re-think interaction and use AR as a human-machine interface, a lot of new possibilities emerge. Get in touch if you have a use case you'd like to explore. Full video: youtu.be/2zxIPkpYseo #construction #manufacturing #factory #automation #robots #augmentedreality
2:16
352 views
1
5
How long before we have a truly live World Map to explore? How long after that can we jump into it virtually?
How long before we can use that map to code against reality, for AVs, for AR, for AI?
This idea will never cease to fascinate me! 🤯
Quote Tweet
Fixed the stuttering in the FMV by interpolating frame locations between metadata captures. Now, pretty smooth! 
twitter.com/dkwiens/status…
twitter.com/dkwiens/status…
0:30
14.2K views
4
3
38
I've just appreciated for the first time that we now have a universal hand gesture for "VR". 📞🎥📷📺😎
Quote Tweet
.@KevinJames' son convinced him to get an Oculus in order to be more active and then Kevin found his son standing there ringing people up at a virtual convenience store. #FallonTonight
1:56
55.5K views
3
1
19
As a…
– user
I want…
– an immersive VR shopping experience
so that…
– I can enjoy the inconvenience of physical shopping from the convenience of my home
read image description
ALT
17
397
2,425
Computer vision folks: this is the only thread you need for images that illustrate the challenge of long-term localization. ✌️ twitter.com/buitengebieden
This Tweet is unavailable.
1
2
38
Is this real life? 🧐
Quote Tweet
1
1
4
I just tried ClipDrop... it's incredibly good!
Quote Tweet
We raised $1,5M to reinvent AI-first photo editing! link.medium.com/zJedBf9lpjb
Join us!
1
1
18
Read more on medium (medium.com/@clipdropapp/c), try the product (clipdrop.co) and let them know if you know any great mobile and front-end engineers (notion.so/We-re-hiring-e) 🚀🚀🚀 4/4
6
Show this thread
Smashing progress so far from & the team (11k paid subscribers grown organically) and I'm truly excited to see as the team roll out their API to turbo-charge the value to creators in the coming months. 3/4
1
5
Show this thread
. will be a major player in that ecosystem, and I'm thrilled to be participating in their journey as the team strives to dramatically lower the barriers for creating high quality product visuals, powered by a touch of AI. 2/4
2
4
Show this thread
The next few years will truly be owned the makers and creators. From synthetic video ( ) and audio (@sonanticlabs), to NFTs and AR, to Community-driven marketplaces () and 3d object scanning (). 1/4
Quote Tweet
We raised $1,5M to reinvent AI-first photo editing! link.medium.com/zJedBf9lpjb
Join us!
3
9
28
Show this thread
Guess the film!
1
Show this thread
Annoyingly I've got to stay in hospital overnight (nothing serious I hope). Brought a Quest with me so I can get through a few levels of
3
14
Show this thread
#2: Berlin is the place to be in Europe for AR creators! Would very much love to meetup in person some time and experience some of your AR demos again in reality!
7
Show this thread
#1: Spaces is super compelling. Different to video call, different to group text. I didn't use clubhouse but definitely see that group voice broadcasts warrant their own category.
1
3
Show this thread
Just came off my first Spaces meetup, courtesy of , and the Berlin AR massive. I have two thoughts -
1
9
Show this thread
Back then in 2016, was already pursuing his vision for autonomous mobility, and was convinced in the redundancy of expensive lidar sensors or HD-maps.
1
2
Show this thread
The complexity of this environment for a self-driving car is off the charts! Awesome demo from
Wayve's CEO, was one of our technical advisors at Scape, when and I met him his old lab at Cambridge University.
GIF
1
2
10
Show this thread
Amazing results from @sonanticlabs, creating a text-to-speech model to revive Val Kilmer's voice for him, after he lost it to cancer. Could be an interesting partner to
Quote Tweet
Replying to @matthewclifford @valkilmer and @join_ef
2
6