Victor Escorcia

Also LoRAs transfer for the same arch between distilled versions. For example, a lora for Flux-dev1 works to a certain extent for Flux-Schnell (qualitative eyeballing).

(is that a subset of ur case?)

Susan Athey—pioneer of causal ML in econ & marketplace design, first woman Clark Medalist.
Different lane than 2024’s institutions Nobel; more about how platforms learn & allocate than why nations prosper?
Let’s see what 13 Oct brings 🧬🧠📈

Fair enough. What about the financial assets for your future self?

bsky.app/profile/esco...

bsky.app/profile/esco...

Diversify your investments (401k, (Roth)IRA, etc.) more aggressively than ppl outside the US)

Financial instruments is as powerful as votes in the modern days

The "Differentiable Systems and Scientific Machine Learning" workshop at the 1st @euripsconf.bsky.social is accepting submissions!💡

Submission deadline for the call for papers is Friday Oct 10th.

Find out more and submit your #EurIPS paper here: differentiable-systems.github.io/workshop-eur...

AI Capability Framework An open source framework helping universities to assess AI maturity, identify strengths, gaps, and improvement priorities.

Have you heard about this? Any opinion?

www.aicapability.org

JEPA 🤠🐴

connection being? use of SAE?

You could scale it to more authors. Then, feed a piece from a new author (discourse, news piece, tweet) & see where it lies in the embedding space of the trained authors.

If you're not into training, yet curious about AI/ML for humanities & news, check out @groundnews.bsky.social

#skyvision

As a scholar, I feel compelled to disclose that I heard about "hardware-driven curriculum learning" from Abhinav Gupta ~2016 😅. He became a father around that time & shared how that experience motivated/reinforced his interest in Robotics.

And yet, nature’s pre-training & self-replication 🧬 remain far ahead of any AI/ML model we’ve built.

Bio-inspired AI should lean toward hardware-aware, energy-efficient curriculum learning. Every computing device should grow its own intelligence & share it (for replication purposes)

3/3 🧵 (END)

Baby development is simply amazing. Our daughter went from quasi-blind 👁️, alert 👂, trembling, etc. to:
1. learn to control 👄, sucking 👩‍🍼🍼
2. locomotion for digesting, and hold the neck
3. sight & hear 👂
4. Short-term focus 👁️
5. locomotion for eating, rolling
...

2/3 🧵

@tkipf.bsky.social opened the discussion (🐦) about ways to go in AI development from his experience as a father 👪🧑‍🍼

Similar experience & I'd recommend anyone in AI/ML (especially those trained after the deep learning revolution) to observe & reflect on kids (or animals) early development 👶

1/3 🧵

Does it clean stuff like this? I don't mind the last two FP 😂
bsky.app/profile/artw...

bsky.app/profile/spic...

bsky.app/profile/spic...

Hi friends! I made a computer vision feed that was missing on Bluesky.

You can find it here: bsky.app/profile/did:.... Pin it to your profile if you like it. 😉

It filters relevant posts based on regular expressions. But to make sure that your post is included in the feed, add the #skyvision tag.

Plaud Note Plaud Intelligence: AI transcription in 112 languages with speaker labels and custom vocabulary. Generate summaries with 3,000+ templates, mind maps, and w...

🐇🤖 (CE device) - bs + practical use case
btw, w/o the device, there are a couple of apps targeting the same market.

...

I wonder if there is a place for selling software that distorts perception signals for AIs:
Watermarking & Consent in the age of AI

www.plaud.ai/products/pla...

skythings.lukeacl.com/bloomfie-bub...

lol some of the personality traits (that I think) of myself are represented in my emoji replies 😂😊

sharing the interactions' 🫧 as well for the curious ppl

the past few weekends here and there spread through a few months have been fun for me. i vibe-coded and along the way learned various services that enable one to rapidly deploy a trial version of a web service.

an awsome era for software development!

Familiar names among #ICCV2025 Outstanding Reviewers from our team 😇
Antoine Guédon @antoine-guedon.bsky.social
Sinisa Stekovic
Renaud Marlet
👏
@iccv.bsky.social
iccv.thecvf.com/Conferences/...

Noice!
Did you podcast a long time ago? I found you on tweets. Great to add a new dimension to your character 🤭🤓

A common confusion I'm seeing is people mixing levels of analysis wrt neural nets: we understand the implementation level well and the algorithmic level somewhat but not the computational level of "how does it internally compute things."

Multilevel Interpretability Of Artificial Neural Networks: Leveraging Framework And Methods From Neuroscience As deep learning systems are scaled up to many billions of parameters, relating their internal structure to external behaviors becomes very challenging. Although daunting, this problem is not new: Neu...

If people have interest in Marr's levels and AI, we're working on updating this: arxiv.org/abs/2408.12664

I gave a talk today at The Curve on the state of open models.
Here are the slides, recording soon.

Topics include: Chinese ecosystem, reflections on DeepSeek, the demise of Llama, who will fill the U.S. market, what local models do, ATOM project & ai2, and more topics
buff.ly/8BiC67C

The state of LLMs is messy: Some AI features (like vision) lag others (like tool use) while others have blind spots (imagegen and clocks). And the expensive "heavy thinking" models are now very far ahead of all the other AIs that most people use, capable of real work

None of this is well-documented

This lad also has fun. Now, isn't work fun?

Visiting Windsor 🏰🏴󠁧󠁢󠁥󠁮󠁧󠁿 & Reading

The author is a math 🐐 of my generation