Michelle Yi's Avatar

Michelle Yi

@michelle-yi.bsky.social

A technologist who specializes in AI and machine learning and is passionate about innovating for a better future. Views are my own. GH: https://github.com/yulleyi Art substack: yulle.substack.com LI: https://www.linkedin.com/in/michelleyulleyi

101 Followers  |  91 Following  |  54 Posts  |  Joined: 24.11.2024  |  2.2114

Latest posts by michelle-yi.bsky.social on Bluesky

Post image Post image

Incredible time at Data Connect West this week. Thanks to Women in Data and Women in Analytics for the collaboration!

#ai #diversityintech

23.05.2025 13:34 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
CrypTen Β· A research tool for secure machine learning in PyTorch A research tool for secure machine learning in PyTorch

Really enjoying the latest developments in privacy-first ML efforts.

crypten.ai

07.04.2025 22:20 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Was it really a war plan or just the concept of a plan πŸ€”

30.03.2025 22:26 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Collage with 20 trailblazing Women of AI- Anima Anandakumar, Ayanna Howard, Cynthia Breazeal, Cynthia Rudin, Daphne Koller, Devi Parikh, Doina Precup, Fei-Fei Li, Hanna Hajishirzi, Joelle Pineau, Joy Buolamwini, Latanya Sweeney, Leslie Kaelbling, Margaret Mitchell, Melanie Mitchell, Niki Parmar, Rana el Kaliouby, Regina Barzilay, Timnit Gebru, Yejin Choi

Collage with 20 trailblazing Women of AI- Anima Anandakumar, Ayanna Howard, Cynthia Breazeal, Cynthia Rudin, Daphne Koller, Devi Parikh, Doina Precup, Fei-Fei Li, Hanna Hajishirzi, Joelle Pineau, Joy Buolamwini, Latanya Sweeney, Leslie Kaelbling, Margaret Mitchell, Melanie Mitchell, Niki Parmar, Rana el Kaliouby, Regina Barzilay, Timnit Gebru, Yejin Choi

#WomensHistoryMonth: Honoring trailblazing #WomenOfAI whose research has made an impact on the current #AI/ML revolution incl. @anima-anandkumar.bsky.social @timnitgebru.bsky.social @mmitchell.bsky.social @deviparikh.bsky.social @ajlunited.bsky.social @yejinchoinka.bsky.social @drfeifei.bsky.social

30.03.2025 19:33 β€” πŸ‘ 43    πŸ” 16    πŸ’¬ 0    πŸ“Œ 0
Preview
IMPLEMENTING THE PRESIDENT’S β€œDEPARTMENT OF GOVERNMENT EFFICIENCY” COST EFFICIENCY INITIATIVE Continuing the "Executive Order Series"

Next in the art substack ❀️

Maybe wishful thinking?

yulle.substack.com/p/implementi...

#art

30.03.2025 21:57 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Representation Engineering: A Top-Down Approach to AI Transparency Understanding and Controlling the Inner Workings of Neural Networks

Definitely tracking this effort in the context of red-teaming. Their approach is a great example of the next evolution in this space as agents open up an entire new attack vector.

www.ai-transparency.org

#adversarialmodels #redteaming #agents

30.03.2025 21:56 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Generationship AI Founders Survey 2025SurveyMonkey logo with text in primary

AI/ML Founders of Bluesky! Please take a couple minutes to complete our survey so VCs can better understand needs! We will share our findings toward the end of the year.

www.surveymonkey.com/r/8HP6S8K

#venturecapital #startups #ai

11.03.2025 16:50 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Elon Musk labels Geoffrey Hinton β€˜carelessly ignorant, cruel’ after β€˜godfather of AI’ calls for him to be banished from Royal Society Geoffrey Hinton wants Musk to be banished form the Royal Society, but note for alleged "Nazi salutes".

So proud of Hinton for speaking up. Science and funding for science matters. Many amazing developments in our own field, computer science, and advances in AI are thanks to grants.

www.yahoo.com/news/elon-mu...

05.03.2025 00:39 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

You're the best!

28.02.2025 16:58 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
A Team of Female Founders Is Launching Cloud Security Tech That Could Overhaul AI Protection Cloud β€œcontainer” defenses have inconsistencies that can give attackers too much access. A new company, Edera, is taking on that challenge and the problem of the male-dominated startup world.

We at Generationship are so happy to back @edera.dev! Let's celebrate their incredible series A 🌞

www.wired.com/story/edera-...

#venturecapital #startups

26.02.2025 05:08 β€” πŸ‘ 6    πŸ” 2    πŸ’¬ 1    πŸ“Œ 0
Preview
Portrait: A Search for Equal Opportunity Inspired by: "ENDING ILLEGAL DISCRIMINATION AND RESTORING MERIT-BASED OPPORTUNITY"

My latest piece in the executive order series.

open.substack.com/pub/yulle/p/...

#art

23.02.2025 19:14 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Intuitive physics understanding emerges from self-supervised pretraining on natural videos We investigate the emergence of intuitive physics understanding in general-purpose deep neural network models trained to predict masked regions in natural videos. Leveraging the violation-of-expectati...

World model capabilities will increase drastically this year.

arxiv.org/abs/2502.11831

#ai #worldmodels #research

19.02.2025 05:51 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
PARTNR Meta FAIR PARTNR Benchmark

Very useful reasoning benchmark

aihabitat.org/partnr/

#ai #benchmarks #reasoning

18.02.2025 06:15 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Portrait: Erasure When sexual assault data, research, and women's health guidelines are erased.

Starting an art series based on the executive orders

open.substack.com/pub/yulle/p/...

#art

16.02.2025 22:09 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
OpenAI o3-mini System Card This report outlines the safety work carried out for the OpenAI o3-mini model, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

At a minimum, organizations should be documenting Model Cards like this (first introduced by @mmitchell.bsky.social 's paper several years ago)

openai.com/index/o3-min...

#safety #reasoning #alignment

11.02.2025 15:55 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Generationship | Ep. #28, Collective Intelligence with Emily Mackevicius | Heavybit Rachel Chalmers and Emily Mackevicius unpack intelligence in all its formsβ€”from songbird learning to group cognition in subway rats and humans.

Check out our latest episode with Dr. Emily Mackevicius on her research in collective intelligence and what it means for artificial intelligence.

www.heavybit.com/library/podc...

#agents #artificialintelligence #neuroscience

07.02.2025 02:16 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

I had a fascinating conversation with a Christian MAGA supporter on my flight back from a conference. We had a lot in common as big sci-fi nerds.

But he also said the world is flat (where does it end then?) and in the nicest way possible told me that I'm going to hell, along with the "homosexuals"

02.02.2025 20:52 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 2    πŸ“Œ 0

Yes! Check out the repo here

github.com/yulleyi/ddt2...

02.02.2025 18:00 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Faded Memories: Graduation in Greenville Continuation of this style's series

A fun art style I tried exploring further in an attempt to convey the fading nature of memory. No AI here - all errors are my own.

yulle.substack.com/p/faded-memo...

#art #sketching

28.01.2025 03:29 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Post image

We are off to the races at @datadaytexas.bsky.social !!!

25.01.2025 16:34 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
The Data Day Texas 2025 Schedule | Data Day Texas

Excited to be at @datadaytexas.bsky.social! Let's connect!

Full schedule here: datadaytexas.com/2025/schedule

#artificialintelligence #data #genai #adversarialmodels

24.01.2025 02:54 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
2025: The Simple Things A Year to Appreciate the Fundamentals and Authenticity

And also, a little pencil sketch to ring in the new year. Happy 2025 all.

Check out my art substack if you want to see some badly drawn, but authentically human, art:

open.substack.com/pub/yulle/p/...

07.01.2025 00:47 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Event details HBA Logo

Excited to join Life Sciences expert, friend, and colleague Kamayani Gupta to share thoughts on JPM Healthcare Conference and NeurIPS with Healthcare Businesswomen's Association!

my.hbanet.org/EVENTS-PROGR...

#healthcare #lifesciences #ai

06.01.2025 19:27 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Still processing everything from @neuripsconf.bsky.social 2024!

One award-winning paper I liked was "Not All Tokens Are What You Need for Pretraining". Their approach demonstrates the potential for more data-efficient LLM training.

Link: arxiv.org/pdf/2404.07965

03.01.2025 20:34 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Hi everyone!

24.12.2024 01:10 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Extremely disappointed by the keynote here at @neuripsconf.bsky.social. When given a chance graciously by the audience member during Q&A if the reason for calling out the Chinese was because this was an exception, the speaker said no it's because the Chinese don't learn morals and values in schools.

15.12.2024 14:42 β€” πŸ‘ 4    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Great talk by Ilya Sutskevar at @neuripsconf.bsky.social 2024. Pre-training as we know it will end.

Compute is growing: better hardware, algorithms, and larger clusters.

Data is not growing: there's only one Internet, and data is the fossil fuel of AI.

#artificialintelligence #neurips2024

14.12.2024 03:40 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Thank you for putting this together!

12.12.2024 16:59 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
System 2 Reasoning At Scale System 2 Reasoning At Scale Workshop @ NeurIPS ---

Come check out our Basis at the System 2 Reasoning at Scale workshop at @neuripsconf.bsky.social on the 15th!

s2r-at-scale-workshop.github.io

#neurips #artificialintelligence #reasoning

12.12.2024 16:56 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Post image

An AI Clone For Any X Profile

AI personas can be designed to mimic your style / knowledge. Here, we use long-term memory and RAG to build a lightweight, simple, fully customizable clone for any Twitter/X user.

12.12.2024 16:33 β€” πŸ‘ 5    πŸ” 1    πŸ’¬ 2    πŸ“Œ 0

@michelle-yi is following 20 prominent accounts