iiinova — where your digital self keeps evolving

Jon Barron @jonbarron.bsky.social · 08/04/2026

The trick is using the AI to write you smaller versions of the experiments that you can iterate on faster

Jon Barron @jonbarron.bsky.social · 08/04/2026

I would just ask them about their methodology and their interpretation in the normal human way that we currently do. It's very easy to tell the difference between a PhD student who is using an LLM to do science and someone who is just along for the ride as the LLM does science for them.

010

Jon Barron @jonbarron.bsky.social · 07/04/2026

They probably could if I was on their committee, which happens sometimes.

100

Reposted by Jon Barron

Alexander Kustov @akoustov.bsky.social · 10/03/2026

I see many folks are pledging not to use AI in their writing. I pledge the opposite: I will use the latest LLMs, and for that matter any other available tool, to best improve my research or the way I communicate it. That way, if my name is on it, you can be sure it reflects my own best judgment.

11565

Jon Barron @jonbarron.bsky.social · 12/03/2026

yeah I guess definitionally if someone makes new knowledge, you could say that they are learning the knowledge that they make by virtue of it being in their head? But that's not usually how people talk about new discoveries or findings being made.

000

Jon Barron @jonbarron.bsky.social · 12/03/2026

ah I guess my program fast-tracked me

000

Jon Barron @jonbarron.bsky.social · 12/03/2026

My understanding was that the goal of doing a PhD was to make new knowledge

430

Jon Barron @jonbarron.bsky.social · 11/03/2026

yeah that was an odd exchange, I didn't expect that concern. I guess I don't really think of a PhD has being primarily about the student learning stuff, but instead about the student accomplishing stuff (or learning how to accomplish stuff).

210

Reposted by Jon Barron

Zhenjun Zhao @ericzzj.bsky.social · 04/12/2025

Radiance Meshes for Volumetric Reconstruction Alexander Mai, Trevor Hedstrom, @grgkopanas.bsky.social, Janne Kontkanen, Falko Kuester, @jonbarron.bsky.social tl;dr: Delaunay tetrahedralization->constant density and linear color radian->radiance mesh->radiance field arxiv.org/abs/2512.04076

031

Jon Barron @jonbarron.bsky.social · 01/11/2025

This, combined with most fields outside of computer science being overly concerned with maintaining cultural and social solidarity (especially against encroaching technology brothers) seems like the most likely explanation.

010

Reposted by Jon Barron

Paul Gavrikov @paulgavrikov.bsky.social · 08/09/2025

Is basic image understanding solved in today’s SOTA VLMs? Not quite. We present VisualOverload, a VQA benchmark testing simple vision skills (like counting & OCR) in dense scenes. Even the best model (o3) only scores 19.8% on our hardest split.

2175

Reposted by Jon Barron

Hà Phan @hpdailyrant.bsky.social · 16/07/2025

Here’s what I’ve been working on for the past year. This is SkyTour, a 3D exterior tour utilizing Gaussian Splat. The UX is in the modeling of the “flight path.” I led the prototyping team that built the first POC. I was the sole designer and researcher on the project, one of the 1st inventors.

5665

Jon Barron @jonbarron.bsky.social · 04/07/2025

Ah cool, then why is that last bit true?

000

Jon Barron @jonbarron.bsky.social · 03/07/2025

I don't see how the last sentence follows logically from the two prior sentences.

100

Jon Barron @jonbarron.bsky.social · 03/07/2025

Be sure to do a dedication where you thank a ton of people, it's kind plus it feels good. Besides that I'd just do a staple job of your papers. Doing new stuff in a thesis is usually a mistake, unless you later submit it as a paper or post it online somewhere. Nobody reads past the dedication.

130

Reposted by Jon Barron

Matthias Niessner @niessner.bsky.social · 27/05/2025

🚀🚀🚀Announcing our $13M funding round to build the next generation of AI: 𝐒𝐩𝐚𝐭𝐢𝐚𝐥 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 that can generate entire 3D environments anchored in space & time. 🚀🚀🚀 Interested? Join our world-class team: 🌍 spaitial.ai youtu.be/FiGX82RUz8U

youtu.be

SpAItial AI: Building Spatial Foundation Models

YouTube video by SpAItial AI

5539

Jon Barron @jonbarron.bsky.social · 28/04/2025

Here's a recording of my 3DV keynote from a couple weeks ago. If you're already familiar with my research, I recommend skipping to ~22 minutes in where I get to the fun stuff (whether or not 3D has been bitter-lesson'ed by video generation models) www.youtube.com/watch?v=hFlF...

youtube.com

Radiance Fields and the Future of Generative Media

YouTube video by Jon Barron

25912

Jon Barron @jonbarron.bsky.social · 09/04/2025

yeah those fisher kernel models were surprisingly gnarly towards the end of their run.

000

Jon Barron @jonbarron.bsky.social · 08/04/2025

yep absolutely. Super hard to do, but absolutely the best approach if it works.

040

Jon Barron @jonbarron.bsky.social · 08/04/2025

If you want you can see the models that AlexNet beat in the 2012 imagenet competition, they were quite huge, here's one: www.image-net.org/static_files.... But I think the better though experiment is to imagine how large a shallow model would have to be to match AlexNet's capacity (very very huge)

image-net.org

120

Jon Barron @jonbarron.bsky.social · 08/04/2025

One pattern I like (used in DreamFusion and CAT3D) is to "go slow to go fast" --- generate something small and slow to harness all that AI goodness, and then bake that 3D generation into something that renders fast. Moving along this speed/size continuum is a powerful tool.

170

Jon Barron @jonbarron.bsky.social · 08/04/2025

It makes sense that radiance fields trended towards speed --- real-time performance is paramount in 3D graphics. But what we've seen in AI suggests that magical things can happen if you forgo speed and embrace compression. What else is in that lower left corner of this graph?

240

Jon Barron @jonbarron.bsky.social · 08/04/2025

And this gets a bit hand-wavy, but NLP also started with shallow+fast+big n-gram models, then moved to parse trees etc, and then on to transformers. And yes, I know, transformers aren't actually small, but they are insanely compressed! "Compression is intelligence", as they say.

130

Jon Barron @jonbarron.bsky.social · 08/04/2025

In fact, it's the *opposite* of what we saw in object recognition. There we started with shallow+fast+big models like mixtures of Gaussians on color, then moved to more compact and hierarchical models using trees and features, and finally to highly compressed CNNs and VITs.

240

Jon Barron @jonbarron.bsky.social · 08/04/2025

Let's plot the trajectory of these three generations, with speed on the x-axis and model size on the y-axis. Over time, we've been steadily moving to bigger and faster models, up and to the right. This is sensible, but it's not the trend that other AI fields have been on...

150

Jon Barron @jonbarron.bsky.social · 08/04/2025

Generation three swapped out those voxel grids for a bag of particles, with 3DGS getting the most adoption (shout out to 2021's pulsar though). These models are larger than grids, and can be tricky to optimize, but the upside for rendering speed is so huge that it's worth it.

170

Jon Barron @jonbarron.bsky.social · 08/04/2025

The second generation was all about swapping out MLPs for a giant voxel grid of some kind, usually with some hierarchy/aliasing (NGP) or low-rank (TensoRF) trick for dealing with OOMs. These grids are much bigger than MLPs, but they're easy to train and fast to render.

140

Jon Barron @jonbarron.bsky.social · 08/04/2025

A thread of thoughts on radiance fields, from my keynote at 3DV: Radiance fields have had 3 distinct generations. First was NeRF: just posenc and a tiny MLP. This was slow to train but worked really well, and it was unusually compressed --- The NeRF was smaller than the images.

29321

Jon Barron @jonbarron.bsky.social · 19/03/2025

Here's Bolt3D: fast feed-forward 3D generation from one or many input images. Diffusion means that generated scenes contain lots of interesting structure in unobserved regions. ~6 seconds to generate, renders in real time. Project page: szymanowiczs.github.io/bolt3d Arxiv: arxiv.org/abs/2503.14445

1356

Jon Barron @jonbarron.bsky.social · 19/03/2025

I made this handy cheat sheet for the jargon that 6DOF math maps to for cameras and vehicles. Worth learning if you, like me, are worried about embarrassing yourself in front of a cinematographer or naval admiral.

0335

Jon Barron @jonbarron.bsky.social · 11/03/2025

It's certainly a shocking result, but I think concluding that "Sora learned 3D consistency" isn't a totally well-founded claim. We have no real idea what any models learn under the hood, and it should be possible for models to produce plausible videos without actually doing anything "in 3D".

040

Jon Barron @jonbarron.bsky.social · 03/03/2025

But if there's an actual physical conference with talks and posters, there would need to be an actual cutoff date for submissions to be presented at this year's conference. Wouldn't that become the de facto deadline, even in a theoretically rolling system?

120

Jon Barron @jonbarron.bsky.social · 27/02/2025

Good tweet! Hadn't seen it, wasn't subtweeting it

040

Jon Barron @jonbarron.bsky.social · 27/02/2025

(no, sorry, what post? Can you link me?)

100

Jon Barron @jonbarron.bsky.social · 27/02/2025

hah so meta, by letting you see my internal thought process I have undermined the conference's perceived legitimacy. See?!

000

Jon Barron @jonbarron.bsky.social · 26/02/2025

The upsides are 1) people have fewer concrete things to complain about when they don't have visibility into the process, and 2) it will make the organization look more competent. Companies and governments understand this.

200

Jon Barron @jonbarron.bsky.social · 26/02/2025

I would strongly support submission fees, especially if submitters are required to review. Bonus points if you get some or all of your fee back if your paper gets accepted.

030

Jon Barron @jonbarron.bsky.social · 26/02/2025

just pay reviewers and ACs iff they are prompt. There is currently no incentive structure for encouraging timeliness.

120

Jon Barron @jonbarron.bsky.social · 26/02/2025

I think this would go more smoothly if the internal CVPR review process was more opaque to the broader community. What is the value of letting people know what is happening? Do submitters really need to know about ACs and SACs and cross-checking and AC-internal deadlines?

100

Jon Barron @jonbarron.bsky.social · 24/02/2025

Can someone point me to a video of renderings of some 3DGS-like algorithm *as it is being optimized*? I want to see all those little Gaussians wobbling around.

240

Jon Barron @jonbarron.bsky.social · 24/02/2025

I think I like the title, though I do generally take issue with intensifying adverbs in scientific publications

100

Jon Barron @jonbarron.bsky.social · 24/02/2025

"Our analysis conclusively determines that Sora's geometric consistency is lit."

2110

Jon Barron @jonbarron.bsky.social · 23/02/2025

yeah same. Too many grad student reviewers, not enough industry veterans.

130

Jon Barron @jonbarron.bsky.social · 23/02/2025

Next week is the one year anniversary of this paper showing that videos generated from Sora are nearly 3D-consistent. I'm surprised we never saw any follow-up papers in this line evaluating other videos models this way, it would be helpful to track these metrics over time. arxiv.org/abs/2402.17403

3432

Jon Barron @jonbarron.bsky.social · 22/02/2025

Veo 2 now has a public price point: $0.50 per second. Very important number to keep in mind when considering the future of generative and non-generative media. Taken from cloud.google.com/vertex-ai/ge...

0110

Jon Barron @jonbarron.bsky.social · 19/02/2025

Thanks Bart! That means a lot coming from you.

010