Kyle McDonald
@kcimc
Followers
36K
Following
17K
Media
4K
Statuses
22K
are you are thinking about tech, nature, the ocean, sailing, or ✨mysterious luminous phenomena✨? then you may enjoy this talk 💙 it’s my first time speaking about my work with the traditional polynesian voyagers from taumako, solomon islands
7
9
53
i'm still figuring out editing and cross-posting, and the upload here has totally wrong colors. check any of these other sites: https://t.co/VdKsHnlw4e
https://t.co/LNs3WTF8ry
https://t.co/A3oT6GDo0b
https://t.co/irEQuKmo6f
0
0
1
i cannot even begin to explain my rage at how glitchy the instagram reels captions tool is
0
0
1
i’m on day 19 of 50 days in the south pacific, helping upgrade power and internet for two traditional voyaging organizations—and trying to capture a rare flash of light called “te lapa”. i’m posting daily on instagram https://t.co/Jw2eOtDLCR
0
0
21
future games will be scripted in natural language, and an LLM will interface with a world model like this one as both game state and rendering later.
Harder, Better, Faster, Stronger, Real-time! We are excited to reveal Genie 3, our most capable real-time foundational world model. Fantastic cross-team effort led by @jparkerholder and @shlomifruchter. Below some interactive worlds and capabilities that were highlights for me
0
0
0
Tomorrow will be 10 years since we shut down the Free Art & Technology Lab aka F.A.T. #RIP
https://t.co/DEiRj6ETtm
@fffffat 👋 @jamiew @danzeeeeman @wheresaddie @theowatson @EvanRoth_ @gleuch @jamespowderly @bekathwia @arambartholl @golan @kcimc @LM4K
2
5
17
and the fourth leap is these generative video models. they represent a completely different and insultingly unprincipled approach to virtual cameras. but it works, in its own way.
imagine this demo, but instead of navigating a virtual world you are clicking around an app. the end-to-end will continue until morale improves. https://t.co/4oqxfJ6qV1
0
0
3
the third leap was nerfs, and later gaussian splats
this is wild. why do photogrammetry when you can optimize for a 5D neural radiance field? this isn't a 3D model of a car: it's a series of renders from a fully connected (non-convolutional) network. the scene is stored in the weights of the network. https://t.co/2Mv148IuzM
2
0
4
the second big leap for me was learning about the plenoptic function, which is a way of describing how a scene looks from every possible angle. it's one conceptual step beyond the more familiar 3d modeling, voxels, point clouds. https://t.co/A5z593OmiE
1
0
2
i think the first time i had this vision for a future of imaging with virtual cameras was watching the bullet time scene in the matrix, and then the burly brawl in matrix reloaded
1
0
0
only a few people are starting to put this together. maybe because a lot of the in-between states can be kinda wonky. but when it works, it works
Finally got the chance to try Runway Aleph. Definitely some interesting stuff here. Still very much figuring out the prompt structure and task type that works best and will update as things progress.
1
0
0
video models like @runwayml and veo 3 can do more than render alternative perspectives—they can render everything in-between and extrapolate beyond. they are implicitly/associatively representing the entire scene.
Filmmaking just changed forever. @runwayml just dropped Aleph, and the implications are massive! Everything you need to know in this thread 🧵👇
1
2
15
i feel like people still aren't putting it together that this is probably the future of most computation—a shared, simulated hallucination, loosely structured by a story/script instead of code https://t.co/gWKSzJ45Xd
1
0
5
these realtime implementations will eventually converge with the most cutting-edge offline techniques we have today
Introducing Runway Aleph, a new way to edit, transform and generate video. Aleph is a state-of-the-art in-context video model, setting a new frontier for multi-task visual generation, with the ability to perform a wide range of edits on an input video such as adding, removing
0
0
2
if anyone is interested in re-scraping linkedin and updating the database please get in touch.
25
4
16
i originally launched this project in 2018 using the microsoft face recognition API, because microsoft was contracting for ICE and i wanted to invert the tech. microsoft finally restricted access to their API in 2024, disabling the site. i just rebooted it to run fully on-device.
47
9
33
https://t.co/Dazi7u8z7O shows the identity of ICE employees using face recognition against hundreds of scraped linkedin profiles 🤗
icespy.org
Find ICE employees using face recognition.
222
73
142