Tootfinder

"Dynamic electric vehicle charging incentives induce large shifts in both the timing and location of charging.
Incentives designed to shift charging toward peak solar generation periods result in a 34% increase in midday charging."
https://www.sciencedirect.com/science/arti

Space waste - an update of the anthropogenic matter injection into Earth’s atmosphere: #SpaceDebris

Part of the ability to detect bot-generated content is simply *good taste*. Humans maybe cannot describe what is “off” about the crud, yet still can see that it’s off.
Maybe it’s inconsistent perspective lines. Maybe it’s impossible focal aberrations. Maybe it’s the odd number of teeth. The redundant sentences. Excessive adjective use. Those who have developed good taste in text or visual art organically may not be aware of the specific technical issues, yet still recognize slop as slo…

NASA has announced that it will release a solicitation related to lunar surface power this month
and that it is considering sending a rover powered by a
"radioisotope thermoelectric generator" (RTG) to the moon.
In a press conference,
NASA Administrator Jared Isaacman and NASA Moon Base Program Manager Carlos García-Galšn discussed new award announcements
to Astrobotic, Firefly Aerospace, and Intuitive machines
to deliver NASA science payloads to…

NASA prepares to solicit moon surface power, teases RTG-powered moon rover

Shanghai-based AI chipmaker Biren raises ~$892.5M in a new share sale to boost GPU production; Biren's stock is up 150% since its January Hong Kong IPO (Ann Cao/South China Morning Post)
https://www.scmp.com/tech/tech-trends/arti

China’s Biren seeks US$900m to fund GPU push and challenge Nvidia
The company says 60 per cent of the fresh capital will be for commercialisation and mass production of next generation general-purpose GPUs.

The SMC Blind Spot: A Failure Mode Analysis of State-of-the-Art Beat Tracking
Jaehoon Ahn, Tae Gum Hwang, Moon-Ryul Jung
https://arxiv.org/abs/2605.12287 https://arxiv.org/pdf/2605.12287 https://arxiv.org/html/2605.12287
arXiv:2605.12287v1 Announce Type: new
Abstract: Over the past two decades, the task of musical beat tracking has transitioned from heuristic onset detection algorithms to highly capable deep neural networks (DNN). Although DNN-based beat tracking models achieve near-perfect performance on mainstream, percussive datasets, the SMC dataset has stubbornly yielded low F-measure scores. By testing how well state-of-the-art models detect beats on individual tracks in the SMC dataset, we identify three distinct failure modes: octave errors, continuity errors, and complete tracking failure where all metrics fall below 0.3. We reveal that state-of-the-art models tend to generate "confident-but-wrong" activations. Furthermore, we show that the standard DBN's default minimum tempo of 55 BPM prevents it from inferring the correct tempo for 21\% of SMC tracks, forcing double-tempo predictions on slow music. By exposing such fundamental oversights, we provide concrete directions for improving beat and downbeat detection, specifically emphasizing training data diversification and multi-hypothesis tempo estimation.
toXiv_bot_toot

Not entirely sure what my brain was up to, but it prompted Nano Banana to generate an “Overly Concerned Drafthorse,” and well, that’s that.

It's nice to see that a state of the art image generator like Midjourney 8.1 can still make really revolting meals.

Thinking about some things that @… said in another thread, and as someone who advocates against AI hype and against the use of most generative AI in most circumstances, I feel it's important to say: many of the ethical issues with using generative AI mirror almost directly the ethical issues with living/working on land stolen by colonists, except that they're less harmful.
Arguments like "well we don't really know whose work it's ripping off this time" and "artists that post their art online know it's going to be looked at; this is the same thing" and "well it's inevitable and everyone's doing it so it's unreasonable to make a big deal about it" directly echo arguments like "well now we don't know whose land it was any more exactly" (yes, we do; you can literally go look up the website of their descendants), or "the natives weren't really using the land anyways", or "it's all in the past now, and it's unavoidable." That unavoidable one is actually somewhat true of using stolen land, at least compared to LLM usage.
If you can see through those lies in the case of AI hype but choose not to do so in the case of colonialism, that says something about your priorities and allegiances.
This is not at all a call for people to talk less about AI; rather it's a call for those who take opposing AI hype seriously to look around and make some noise about other injustices too (I realize many of you already do this).
#AI #LLMs #LandBack #GenAI

SF-Flow: Sound field magnitude estimation via flow matching guided by sparse measurements
Ege Erdem, Shoichi Koyama, Tomohiko Nakamura, Orchisama Das, Zoran Cvetkovi\'c
https://arxiv.org/abs/2605.10398 https://arxiv.org/pdf/2605.10398 https://arxiv.org/html/2605.10398
arXiv:2605.10398v1 Announce Type: new
Abstract: Reconstructing a 3D sound field from sparse microphone measurements is a fundamental yet ill-posed problem, which we address through Acoustic Transfer Function (ATF) magnitude estimation. ATF magnitude encapsulates key perceptual and acoustic properties of a physical space with applications in room characterization and correction. Although recent generative paradigms such as Flow Matching (FM) have achieved state-of-the-art performance in speech and music generation, their potential in spatial audio remains underexplored. We propose a novel framework for 3D ATF magnitude reconstruction as a guided generation task, with a 3D U-Net conditioned by a permutation-invariant set encoder. This architecture enables reconstruction from an arbitrary number of sparse inputs while leveraging the stable and efficient training properties of FM. Experimental results demonstrate that SF-Flow achieves accurate reconstruction up to \SI{1}{kHz}, trains substantially faster than the autoencoder baseline, and improves significantly with dataset size.
toXiv_bot_toot