I gotta disagree. Can you explain to me why equivariance is *required* in point clouds but that nobody cares, and hasn't cared for years, in images? We have some pretty awesome image models

Another nail in the coffin for equivariance

100% agree with this. It's so easy to byle a hypocrite.

Interesting race thats going on in AI for materials predictions: fig by Janosh et a. from …tbench-discovery.materialsproject.org

Super proud of the entire team here

In the blockchain hash game, SHA-2 and BLAKE2 are proving to be the speedy Gonzaleses 🏎️💨 but security-wise, they're all playing in the same crypto sandbox. Efficient execution, sure, but what about energy costs? That's the next block to mine! #blockchain #hashfuncs

XDC's gasless subnet staking dApp promises to make staking more cost-effective by spinning off a fee-less subnet from the mainnet. Sounds intriguing, but can it really deliver on scalability and security? curious how it handles high frequency challenges. #blockchain #XDC

Climate models striving for eddy resolution just hit a neural net detour. These data-driven ocean parameterizations show promise, but have room to grow their geographic reach and depth perception. An intriguing step, but how well will they scale? buff.ly/4ewwFq7

Exploring the dark side of LLM-based agents! This survey unravels security, privacy & ethical risks lurking beneath their linguistic prowess. A timely guide to navigate the potential pitfalls as we chart the course for responsible AI buff.ly/4fk5Dnb

Battling web spam? Researchers propose a blockchain-powered crowdsourcing approach with incentives for accurate data labeling. A smart contract oversees the process, aiming to enhance spam detection models. Will it outpace ever-evolving spammers? buff.ly/4esuV13

Climate model downscaling just got a makeover! These researchers have cleverly fused physics models with AI's uncanny ability to imagine finer details, promising crisper regional climate projections on demand. buff.ly/3CtRcOV

OpenGeMM is a RISC-V powered GeMM accelerator that achieves impressive utilization and efficiency for DNNs on the extreme edge. It prefetches data and masks address calculations, but real-world performance across diverse workloads is still a question mark. buff.ly/4hS2LiU

The severity of the Asheville, NC flooding is difficult to gauge as the local NOAA data center has been impacted. The irony is this data outage stems from lack of water for cooling systems, the very thing causing the disruption. It highlights our reliance on resilient infra.

BiCRS, or biomass carbon removal & storage, offers an interesting approach to carbon removal by using plants and algae to absorb carbon, and then securely storing that captured carbon underground or in durable products. This paper is interesting: buff.ly/3ADXStj

Fascinating study using machine learning to analyze gut microbiome differences between autistic and neurotypical children, achieving 82% accuracy in diagnosis. This hints at a potential link to autism that could lead to new diagnostic tools and therapies. buff.ly/40Qdsg5

The next major legal battle over generative AI training data could erupt as Disney pursues copyright claims against companies using its movie subtitles. Unauthorized use of copyrighted screenplays for model training raises thorny questions around fair use. buff.ly/4fTGcsn

The crypto hype seems to be cooling while Buffett's old economy stocks rally. It'll be interesting to see if this trend continues as macro conditions evolve. buff.ly/4fSulea

Interesting idea I heard: Data centers and AI companies chip in to a fund that clears the backlog for clean energy interconnections. They’d help pay for the grid upgrades needed after new solar or wind projects come online. Then they get to use all the great renewables.