There are 2 previous historical cases of countries destroying their science and universities, crippling them for decades: Lysenkoism in the USSR and Nazi Germany. The Trump administration will be the 3rd.
It's not just budgets but research, institutions, expertise, and training the next generation.

GTSAM 4.3a0 ! github.com/borglab/gtsa...

I know this look well. It’s the same one I get when I introduce the Fourier transform in my computer vision course to the CS crowd.

Wow, lucky you!

Trump Says He Won’t Rule Out Third Reich

Part 2 of SLAM handbook is out for public comments! let us know what you think :-) Issue tracker on GitHub awaits! Link: github.com/SLAM-Handboo...

And yet Joe Rogan puts guests on who say "Vaccines aren't actually responsible for the reduction in infectious diseases.”

Aria Gen 2 is very impressive, with fully onboard SLAM and various other perception all within a 75g device with an hours-long battery life. Processing on custom silicon. Congrats to the Reality Labs team.

MASt3R-SLAM code release!
github.com/rmurai0610/M...

Try it out on videos or with a live camera

Work with
@ericdexheimer.bsky.social*,
@ajdavison.bsky.social (*Equal Contribution)

CRA statement about NSF firings cra.org/cuts-to-nsf-...

Some personal news: as of January I am back full-time at Georgia Tech following a 2-year leave as Verdant Robotics’ CTO. I will continue to be involved with Verdant as part-time Chief AI Officer, thinking strategically about the role of AI in Robotics for Ag.

Gemini is good but too verbose :-)

The visual system of a jumping spider is fascinating. Look at those cones behind the fixed main lenses! The retinas are at the end of the cones. youtu.be/gvN_ex95IcE?...

We've built a simulated driving agent that we trained on 1.6 billion km of driving with no human data.
It is SOTA on every planning benchmark we tried.
In self-play, it goes 20 years between collisions.

Cool beans!

Video Depth Anything: Consistent Depth Estimation for Super-Long Videos

TL;DR: Long videos support; Depth Anything
V2 with efficient spatial-temporal head. Temporal
consistency loss -> depth gradient (no geometric priors)

Maybe a broom attached to the train :-)

That’s such a cool idea! (Thanks K for flagging this to me)

Done

Not sure I’m optimistic about that :-) I can see the number of humans involved decrease. Georgia Techs OMSCS is a good metaphor. More TAs, less teachers. TAs will be replaced by AI: one AI-TA per student per course.

I’m sure oxbridge tutors have a pretty good idea of what they have in front of them after a couple of sessions :-)

I don’t have the answers. But I’m optimistic we’ll arrive at a good outcome eventually. We will still need to find a way to somehow contribute value, and to document that fact in order to get hired for those skills. Maybe the key to certification lies in those 1-1 tutoring sessions themselves.

Sure. I think we’ll have to find a different way. At this point when I code, I use ChatGPT/Claude extensively, so maybe we should start thinking about this being a valid way to complete assignments :-)

I am more optimistic. It’s well known that one on one tutoring can be up to two standard deviations more effective than classroom teaching, but it’s prohibitively expensive. AI gives us cost-effective solution for just that. It’s *already* doing that for me :-)