Johan Edstedt

IPS 2025 for which you are serving as reviewer. We have received a mandatory acknowledgement on a submission to NeurIPS 2025 for which you are serving as reviewer. We have received a mandatory acknowledgement on a submission to NeurIPS 2025 for which you are serving as reviewer. We have received a m

We have received a mandatory acknowledgement on a submission to NeurIPS 2025 for which you are serving as reviewer. We have received a mandatory acknowledgement on a submission to NeurIPS 2025 for which you are serving as reviewer. We have received a mandatory acknowledgement on a submission to Neur

Gotta practice my whistling and egg-laying to remain hireable

3D model meaning neural nets in this context.
Reconstruction method meaning (?)

Neural networks leave you wondering how they work, despite your obviously wrong implementation.

Handcrafted methods leave you wondering how they break, despite your obviously correct implementation.

Im gonna ask clarifying questions when appropriate so hard now!

Sometime after that there will hopefully be some new detectors + features + matchers as usual. Right now getting into self-supervised lit.

These days im mostly finishing up my thesis, but also doing some side-quests. Righy now thinking about nicer apis/libs for 3d. Poselib is getting some updates to python bindings soon :)

Why is Angjoo Kanazawas lab so good? Big fan.

Might also mean that the way humans are 'special' is different from 'intelligence' (in the current understanding of the word)

Will caveat that it depends on the field and school, but yes. It also touches both points 1 and 2 of your list.

Well we can disagree, but I think the freedom of a PhD is worth just as much as the buckets you brought up.

Yep, I'm very conflicted about it.

Everyone asks where is the theory and not how is the theory :'(

Besides the obvious implication, I'm concerned about these tools becoming needed.

If you start addressing obvious LLM misconceptions to increase acceptance rate it leads to bad things.

LLM reviews need to be punished much more severely.

review.cspaper.org

Somehow more impressed that france is so much lower than US cmp china.

Ah, got it.

Wouldn't be surprised if people started paying them to burn oil to own the libs.

Alright 80% if we include seminatural. My point was that we already have extremely large inpact on the environment by how we use land.

Washing machine is a laundry robot, just not very mobile.

On the flip-side, converting 90% of earth to cities and farmland is also extreme geoengineering, and seems less controversial.

If the solution is to disrupt reproduction of specific species associated with malaria, that seems good. If the "solution" is to carpet bomb with DDT, not so much.

In a similar way, the potential risks and uncertainties of geoengineering matters.

Could you expand on this? Differently as in not the best solution, or differently as in "doesn't work at all"?

The best solution to climate change is finding alternatives to oil/gas/coal.

Arguably geoengineering is useful to mitigate the effects of existing increase in CO2, but you cannot "solve" climate change without stopping fossil fuel use.

Kernel panic is how I usually wake up in the morning, don't worry.

This might be more true in fields like CS than e.g. wet lab science.

Infinite things that could be put on this list, but I think community/collaboration is an important one.

The freedom academics have to collaborate with basically anyone on earth is unmatched.

Moe

When no one likes my definitely very funny RANSAC memes.