sycra52

Some weird ideas now and there

[Lyman Stone 石來民 🦬🦬🦬 @lymanstoneky]

Unfortunately no, nobody has written it up. But the TL;DR is:

~doubling the wage replacement rate for maternity leave
~doubling cash bonuses for kids
~doubling child allowance
implementing new marriage bonus; nationally it's worth about $2,000, but local top-ups make it up to $10k or more in some areas, and people can obviously jurisdiction shop for marriage
~some other smaller tweaks

they did all this in 2022-2024

it's not just that fertility rose. fertility rose for first births to married women, and the marriage rate pivoted hard as well for the first time in decades

[Jack Morris @jxmnop]

my paper won an award at icml 😁

some thoughts:
• this work was rejected from NeurIPS. i cleaned it up a small amount and it got great reviews from ICML! don't give up
• ICML received 24k submissions and only gives out 7 awards, which is crazy. feeling grateful
• i distinctly remember sitting at my desk two winters ago wondering if i would ever finish this project. most of all this is the product of sitting down and forcing myself to keep working for several months straight. the results emerged from running the experiments over and over and fixing a long sequence of tiny details. eventually, the curves looked like that 👇
• also happy that the insights in this paper are becoming more widely accepted: 3.3 bits/param, thinking about capacity "LLM as flashdrive" mentality
• the method here is used successfully for selecting midtraining data at least one frontier lab, which is cool!
• i am grateful to my collaborators, but Meta is no longer a great place for academic research imo and this almost never got published for a number of reasons. i shall not elaborate further
• for future work, i think analyzing the implications of on-policy algorithms on capacity, as well as LoRA and things like it, are fruitful potential research directions
• sadly i'm not in Korea but am following the conference online from california and happy to chat!

a nice end to one phase of my research career :)

[sycra52 @sycramore]

@PhysRevE Combine that with that paper on entanglement and adiabatic passage time and you get an estimator for the hardness of a specific instance of the Ising model. Wondering when lattice models go down the drain?

[Physical Review E @PhysRevE]

Can #MachineLearning identify phase transitions without prior knowledge of the underlying #physics?

Using a 2D Ising model, this study shows that #NeuralNetworks can recover critical behavior and phase transitions directly from raw spin data.

🔗 https://t.co/Y68xPT9lMt

[Dom Italian Builder @dom_gag_96]

>Be Salvatore Sanfilippo.
>Born in Campobello di Licata, Sicily, in 1977.
>Start hacking on Unix systems and networks.
>Become known online as @antirez.
>Build hping, a tool used by security researchers around the world.
>Invent the idle scan, later implemented in Nmap.
>Build two Italian web products with your co-founder.
>Hit a scaling problem.
>Instead of complaining, create a new database.
>Redis.
>Open-source it in 2009 from Sicily.
>Lead it for more than a decade.
>Watch it become one of the invisible engines of the modern internet.
>Caches. Queues. Real-time apps. AI systems. Infrastructure everywhere.
>Step back in 2020.
>Come back years later because open source still matters.
>Be one of the reasons the internet is fast.
>Be Italian.

----

Join @italianbldrs and help build the next Renaissance.

[sycra52 @sycramore]

@SebJohnsonUK The guys from the IPP in Munich?

[sycra52 @sycramore]

@preskill @TeamOratomic Wishing you all the best and good luck!

[Andy Xu ✈️ ICML @4ndyXu]

For the past 1.5 years, I've worked on post-training language models for materials discovery. I believe we've gotten to the point where these models are starting to show the potential to be useful!

I've written up some of the thoughts and intuitions from open-source research I've done on PLaID++.

Check out the blog in 🧵for details & pretty figures/interactable components :)

[prathosh A P @prathoshap]

I taught a course on the mathematical nitty-gritty of GenAI, equation by equation (written). This covers all the SoTA algorithms, from VAEs to DDPMs to LLMs and state-space models (SSMs), in detail. The link for the entire playlist is in the next message.

[Elizabeth Wood 🧬🖥️🥼 @lizbwood]

We are hiring at @jura_bio both on the ML team (world) and the wetlab team (Boston). Please get in touch if you've been thinking about it. ⏩[email protected]

[Moritz Schularick @MSchularick]

👇🏻👇🏻

[sycra52 @sycramore]

@Jeyffre Can you recommend some papers on the subject everyone should read?

[sycra52 @sycramore]

@bilaltwovec probability distribution. Good luck :)

[sycra52 @sycramore]

@bilaltwovec You want to check out the probability distribution from the muons. If you want to use entropy for optimization normally optimization algorithms require a certain type of entropy. And the easiest way to classify an entropy source is to gather a bit of entropy and look at the

[sycra52 @sycramore]

@EasternDaylight I assume the potentiometer is used to set the initial conditions of that oscillator?

[sycra52 @sycramore]

@EasternDaylight Did you attend the ACDs?

[tender @tenderizzation]

6-7 decades of computer systems research in a nutshell

[Lillian Ma @lillian_ma_]

some AI4Science labs worth following :

@IsomorphicLabs (@demishassabis)
AlphaFold energy turned into AI drug discovery.

@FutureHouseSF (@SGRodriques, @andrewwhite01)
Building AI scientist agents for literature search, hypothesis generation, and discovery workflows.

@LilaSciences (@GVMaltzahn)
Autonomous labs + AI models for biology, chemistry, and materials.

@periodiclabs (@LiamFedus, @ekindogus)
Trying to connect AI scientists with real-world experiments, especially in materials.

@EvoscaleAI (@alexrives, @TomSercu)
Protein language models — basically scaling laws, but for biology.

@arcinstitute (@pdhsu, @SKonermann, @patrickc)
Digital biology, genome foundation models, virtual cells, and programmable biology.

@ProfluentBio (@thisismadani)
Generative AI for proteins and gene editors.

@Latent_Labs (@saakohl)
AI-designed binders and protein design from the AlphaFold orbit.

not exhaustive — add the ones I’m missing in the comments.

[Alex Pruden @apruden08]

https://t.co/KQyxdwpyri

[datavorous @datavorous_]

I reverse engineered Qualcomm's NPU compiler to find undocumented behaviour that affects every edge AI deployment.

Things nobody knew:

1. The compiler silently downgrades the precision of your model weights without telling you
2. Memory placement uses HiGHS which is an LP solver (not heuristics)
3. The same model on two different chips with identical reported VTCM can have 33x difference in DDR traffic
4. There's an undocumented internal simulator called Hextimate pricing ops without the hardware

Every NPU vendor be it Qualcomm, MediaTek or Apple NEVER tells you how to make the most use of their hardware.
I was very close to rage quitting before I finally lost all hope and thought of reverse engineering to understand how NPUs are handled.

Read the full write-up below: