Zikang Jiang
@ZikangJiang
Followers
873
Following
722
Media
3
Statuses
202
19 | m&t dropout | https://t.co/BCr5lxJiNR
Joined October 2024
Win or die
8
2
43
Research Intimacy: Internalizing obscure knowledge, equations, relationships, and ways of thinking related to a research topic.
4
53
303
Our world is created with human priors. To truly leverage commoditized intelligence, we must enable it to interface, experiment, and build upon truths (the physics of the world, the hard constraints) uncorrupted by our thinking.
1
0
10
wow, that was fast! @k7agar
playing around with Egocentric-10K by @eddybuild. got the hands reconstruction done w HaMeR. hand kinematics is sorted and solvable, next i will try to extract contact and trajectories. factory robots soon
1
0
9
going through the dataset, i realized that there are plenty of tasks currently done by humans that should be automated ASAP. no human being deserves to sit in front of a machine 8 hrs a day. human creativity can be used for far better
playing around with Egocentric-10K by @eddybuild. got the hands reconstruction done w HaMeR. hand kinematics is sorted and solvable, next i will try to extract contact and trajectories. factory robots soon
14
9
146
above average
You can just open-source things! If you think you don’t have enough experience, support or ressources to release useful open-source, remind yourself that @eddybuild is 18 and released the currently number one trending dataset on Hugging Face, ahead of Nvidia and Meta
0
0
3
a lot of mosquito bites later
today, we’re open sourcing the largest egocentric dataset in history. - 10,000 hours - 2,153 factory workers - 1,080,000,000 frames the era of data scaling in robotics is here. (thread)
0
0
13
If a problem seems intractable, it's almost always because your specification of it is vague or incomplete. The solution doesn't appear when you "think harder". It appears when you describe the problem in a sufficiently precise and explicit fashion -- until you see its true
59
158
1K
Your default cognitive environment is hostile to deep thought. If you consume a minute-by-minute feed optimized for engagement, you are training your mind to be reactive, shallow, short-term. The most potent way to think about the world is the opposite: analytical, long-term,
71
202
2K
there are essays, and then there are "violently shift your entire perception of reality" essays
145
2K
12K
The particular thing you do is luck, but that you do something is not. - dick hamming
0
0
3
learn like a diffusion model. denoise the problem space. explore different solutions in parallel.
1
0
7
Tinker provides an abstraction layer that is the right one for post-training R&D -- it's the infrastructure I've always wanted. I'm excited to see what people build with it. "Civilization advances by extending the number of important operations which we can perform without
Introducing Tinker: a flexible API for fine-tuning language models. Write training loops in Python on your laptop; we'll run them on distributed GPUs. Private beta starts today. We can't wait to see what researchers and developers build with cutting-edge open models!
49
115
1K
It is very difficult to properly convey “zero-shot generalization”, it’s a very overloaded and overabused term these days! But let me try to add some color to why so many of my colleagues have been so shocked at what the Gemini Robotics 1.5 VLA has been doing… 🎨 The whole
You have to watch this! For years now, I've been looking for signs of nontrivial zero-shot transfer across seen embodiments. When I saw the Alohas unhang tools from a wall used only on our Frankas I knew we had it! Gemini Robotics 1.5 is the first VLA to achieve such transfer!!
6
14
114
rlhf is model calibration. many downstream/orthogonal insights
0
0
1