Joined April 2008
- Tweets 6,799
- Following 1,412
- Followers 2,475
- Likes 10,966
351 Photos and videos
Jun 14
To quote my 7 year old, “We fucking did it!”
Knicks in 5! Greatest fucking city in the world!
2
112
Jun 8
A couple thoughts on CVPR 2026 as I fly home from Denver.
First, number of people and papers continue to explode. ~4k accepted papers/posters.
This is both because of expanded interest/investment in AI and also, I suspect, because AI itself is making it easier to write papers.
4
4
596
Jun 8
Finally, for a typical attendee, you’ll get much more out attending the workshops than out of the main conferences talks or posters.
While the trade show had an impressive number of robots and autonomous vehicles on display, it’s most interesting to students applying for jobs.
1
103
Jun 8
Having a paper accepted for a CVPR talk is a big deal, but the talks themselves are fast paced, given in massive conference halls, and entirely focused on the details of a somewhat arbitrary paper topic.
Also due to visa issue, many talks aren’t even given by the paper’s primary author.
In contrast workshops (of which there are dozens) tend to encompass a clear topic of interest and feature talented senior researchers giving a polished overview of their entire groups results.
On top of that, in a good workshop, you’ll get a real sense of how top researcher’s see the fields/topics being discussed. Also if you want, you can easily ask questions or engage with the workshop organizers.
In short, the workshops provide you with realtime information that is much harder to sniff out at the real conference.
1
1
109
Jun 7
I'm huge fan of @RichardSSutton's original Bitter Lesson Essay from 2019, but was late to the party on @vincesitzmann much more recent vision-focused version.
With many long time vision folks (like me!) moving into robotics, Vincent's observations couldn't be more timely.
Why?
Feb 16
In my recent blog post, I argue that "vision" is only well-defined as part of perception-action loops, and that the conventional view of computer vision - mapping imagery to intermediate representations (3D, flow, segmentation...) is about to go away.
vincentsitzmann.com/blog/bit…
6
6
484
Jun 7
So that's where we're headed.
Learned world models that accurately predict a robot's future observations from its current observations actions.
Learned policies (trained via reinforcement learning) that predict future actions from current and past observations.
If 3D representations end up in the mix, it'll likely be because today's robots often take 3D commands as inputs. Also, existing technology for rendering predicted observations sometimes benefits from 3D geometry. In the long run though, I'd be surprised if either of these constraints sticks around.
1
103
Jun 8
Oh also, for robotics specifically, we still do need the hardware to get better.
The most observations coming off the robot definitely can’t just be visual!
x.com/JitendraMalikCV/status…
I want to offer some unsolicited advice to computer vision researchers jumping into robotics. Don't focus too much on VLMs, VLAs etc. That's fine, but the real action is at the sensorimotor level. Most of the open problems in robotics are in manipulation, which is about hand-object interaction, and contacts and forces are central. Proprioception and tactile sensing are as important as vision. Don't get seduced by cherry-picked demos. You can't do robotics without doing robotics.
112
May 18
Hear me out.
Apple should release a new AirPods Case that also works as a wireless mouse.
Added bonus, it should support a sweet 360 spin gesture, that you perform fingerboard-style.
John Ternus, balls in your court.
170
May 15
3M, enabling robotics years before Nvidia ever showed up.
May 15
When applied robotics companies talk about the "deployment moat", this is what we're talking about. Good luck solving this with your next-gen robot/brain/data collection method.
(context: one of our field engineer's Slack message to a customer)
2
243
Mar 19
Costa Rica be Costa Rockin'
Mar 19
The English-speaking world is falling in rankings of life satisfaction (The Economist)
1
3
340
Mar 8
listing: "a working olive grove that produces oil"
me, after purchase: "so is there like a button I press?"
Mar 8
5.2 acres of land in Italy, a working olive grove that produces oil, private woodland, a renovated stone farmhouse and an 80m² cellar. €260k ($300k).
A lot of people are rethinking where they live and how they live. This is what that looks like in practice.
A courtyard with a wood oven and grape pergola. Three connected structures, 150m² (1,615 sq ft) of living space, 2 beds and 2 baths.
Get Starlink, run your business from anywhere, add solar panels, collect your water. You're largely off-grid.
One hour from Rome if you ever get bored and need the city.
Is this the kind of property that starts making more sense the more the world changes?
391