Jonathan Scholz
@JonathanScholz2
Followers
227
Following
231
Media
3
Statuses
120
Joined September 2018
Ooo finally a use for all that blue TPU I bought 🤩.
We have developed a new tactile sensor, called e-Flesh, with a simple working principle: measure deformations in 3D printable microstructures. Now all you need to make tactile sensors is a 3D printer, magnets, and magnetometers! 🧵
0
0
4
I’ve been looking forward to seeing what @peteflorence and @andyzeng_ would cook up ever since they left DM last spring, and they did not disappoint!. These are some of the prettiest behaviours I’ve seen from a robot yet.
Today we're excited to share a glimpse of what we're building at Generalist. As a first step towards our mission of making general-purpose robots a reality, we're pushing the frontiers of what end-to-end AI models can achieve in the real world. Here's a preview of our early
0
0
6
Been out 8 months but DeepMind finally asked for my laptops back. I can only assume they’re training something BIG.
0
0
8
These guys are crushing it. If anyone gets the ethos of a proper picks-and-shovels startup, they do.
🚀 Rerun 0.23 is out!. The highlights:. 🔸Backwards-compatible .rrd – open today’s recordings with tomorrow’s viewer. 🔸 Viewer callbacks – wire the viewer into your notebooks or web apps and react to selections and time-scrubs in real-time, e.g., to build annotation tools. 👇
1
3
15
This is *so* cool. Does anyone not believe at this point that general-purpose robotics is coming?.
We got a robot to clean up homes that were never seen in its training data! Our new model, π-0.5, aims to tackle open-world generalization. We took our robot into homes that were not in the training data and asked it to clean kitchens and bedrooms. More below⤵️
0
0
5
Great new take on the Track-Anything problem from some of my fav former colleagues at DM! .It's been amazing to see this work mature from bespoke architectures to such an elegant and general formulation.
We're very excited to introduce TAPNext: a model that sets a new state-of-art for Tracking Any Point in videos, by formulating the task as Next Token Prediction. For more, see: 🧵
0
0
3
RT @CarlDoersch: We're very excited to introduce TAPNext: a model that sets a new state-of-art for Tracking Any Point in videos, by formula….
0
57
0
[2/2] .Once that's solved and it's a business problem, people that can't rapidly switch to being problem-first rather than solution-first will get left behind by the next generation of Jobs and Gates types who know how to dream big and steal good ideas.
0
0
0
[1/2] Interesting read -- my take:.I think they're right, but *slightly* too soon. Robotics is in the "Engelbart" phase now -- we're inventing the UI of the future, and you need deeply technical people to do it.
1
0
1
Huge congrats to my former DeepMind colleagues! Big step towards interactive robotics, and no feet required đź‘Ź.
Super excited to finally be able to share some of the really exciting work we have been cooking up @GoogleDeepMind! . Interactivity, Dexterity, Generalization and Multi-Embodiment seem far less far-fetched than before!. Reach out if this excites you too!.
1
0
4
I've been fond of saying the term "foundation model" won't age well (bc it'll be fractionated by post-training for specific markets), but i'll update that to "we'll soon see a foundation model foundry that, like TSMC, doesn't compete with its customers".
0
0
1
I can personally relate to his words on the double-edged sword of ambition, selfishness, and the way the field has charged. I’ve been going through my own version of this struggle, thankfully far less severe, and I’m grateful to Felix for having the bravery to communicate his.
0
0
3
This was heartbreaking. I didn’t know Felix well, but we crossed paths many times and he always struck me as smart, happy, and engaged. That this was his internal world is terrifying, and makes me worry for all the others who are struggling with depression.
I worked with Felix on a paper in 2021. I remember Felix as a consistently kind person, and as a brilliant thinker. I'm sharing his farewell letter, as I think it's what he would have wanted. Please be mindful, it's not an easy read: Rest in peace.
1
0
6
RT @rodneyabrooks: I recommend being kind to everyone rather than demonizing them. I try to be kind to people I disagree with, even if they….
0
2
0
UNROOTED teams up with Disney's Mufasa: The Lion King ❤️❤️. You should buy Unrooted shots, and not *just* because it's my wife's company. They're actually amazing 🤤.
0
0
0
great article from @svlevine about the real-world being the ultimate place to harvest data for robotics . I've argued for this for years at Google, not particularly successfully. Hopefully articles like this can help shift the attitude.
2
0
2
I'm with David on this one. Great paper!. The key bit is "Having to attack all resolutions & all abstractions at once leads naturally to human-interpretable attacks" . All the adversarial generation stuff flows from there.
I think this is possibly the biggest step forward in adversarial robustness in years. So much churn in the ML space and it's hard to tell from one twitter thread what's worth paying attention to, but this one deserves a close read.
0
0
0
[6/n] It'll be like a loom for weaving music which creates space in between the composer and the DJ. Then we'll hear them say it gives us the nice bright colors, momma don't take my kodachrome away.
0
0
0
[5/n] This is a blip though. Very soon the big models will be commoditized, the interfaces will mature enough to express the nuances people want, and society as a whole will have become familiar and enough to use these tools for creation at a global scale.
1
0
0
[4/n] It's easy to get hung up on how derivative AI art it is right now, especially since the economics all seem to favor stealing from artists to put money in the pockets of big companies.
1
0
0