Scott H. Hawley

The preprint is here: arxiv.org/abs/2608.04378 The first 6 pages + refs are submitted to the NeurIPS Creative AI Track ("single-blind, preprints ok" 👍) Preprint adds 10 pages of Supplemental Materials: ablation studies, hyperparameter surveys, etc.

Helping Music Co-Creation Agents 'Listen' Well: Hierarchical Self-Supervised World Models for Understanding and Generation arxiv.org

Speaking of [collective noun] of agents, here's a way to check how/if your models are running on different machines. (Like running `nvtop` on each one manually...as I did before tonight.) github.com/drscotthawle...

GitHub - drscotthawley/remsysmon: Remote System Monitor: live GPU/VRAM/CPU/RAM sparklines and GPU process lists for several hosts over ssh, in one page github.com

This could be difficult. Typically instrumental recordings are done as work-for-hire. (e.g. "Standing in the Shadows of Motown") Still, one might argue for Neighboring Rights -- except the USA doesn't pay those. www.reuters.com/legal/litiga...

reuters.com

Woops! Copy paste error. Thanks for pointing that out. The project page has it right: drscotthawley-midi-rae-jepa-son.hf.space?__theme=dark

MIDI-RAE-JEPA-SON: scrawl me maybe drscotthawley-midi-rae-jepa-son.hf.space

What should be the collective term for a group of AI agents? A troop? A gaggle? A murder? A congress? I like we one we use for starlings: A murmuration!

Releasing "Opa" - "Own your context": I don't love how much of my work lives in platforms I don't control. When my email account died, Anthropic wouldn't let me change emails or export my data. Even on the new account, their search only matches chat titles, not content. Good luck finding anything...

I'm leaving academia, moving to NYC for a research role as Director of AI Alpha Lab at Hudson Bay Capital Management. The chance to work with CAIO Joe Miller (also former Hyperstate) was a big draw, and a lot of the science maps nicely onto my background.🚀 Last visit to Belmont:

Claude Code doing its best to bankrupt me. 😜 Reminder that Anthropic sells *tokens*

Re. falsely accusing outputs of being AI generated: I just got James Clear's "Atomic Habits" on Audible: "Read by the author." But it's delivered in such a robotic monotone that I figured it was rendered by a voice model. ..Nope, seems it was recorded in 2018; reviews say that's just how he talks!

NB you are not allowed to title anything "...in the Age of AI" anymore. That ended years ago.

I beg ML conference organizers to abandon author-year citation format when page limits are on the line. Numeric/IEEE is so much more concise and what most of us from physics and engineering are accustomed to. I'm losing nearly half a page to "(LastName1 & LastName2, XXXX)" over & over.

It still shocks me when people equate "generative AI" & LLMs. As someone active during the early heyday of text-to-image and text-to-audio models -- who taught GANs in class! -- my brain always goes to image gen first. "But that's just Machine Learning" Me: ?<always has been> "Oh! You mean LLMs..."

Why do UK universities keep cutting Humanities when it's software engineering (what most CS programs produce ) that's being gutted by AI? Or maybe they're cutting CS too and I'm just not seeing posts about it?

Seems Anthropic is saving money by just making our requests take 5x longer than before. :-/