Designer/Engineer @MIT | previous intern @apple | previous contractor @openai

Cambridge, MA
Joined December 2024
Ethan Chang retweeted
Today, Transluce is releasing the most expansive independent evaluation to date of how AI systems respond to users experiencing mental health crises. We evaluated 77 model variants from OpenAI, Anthropic, Google, Meta, SpaceXAI, Thinking Machines, DeepSeek and Moonshot AI.
23
83
26
445
79,913
Something that stood out to me in the interview about The Stochastic Parrot was this comment: “They can communicate with other AI cohabitants and maybe even help surface people who are similar to you.” Right now, we connect with people similar to us or helpful to us through IRL, social media, and web searches. That assumes you’re either with a person or that a person of interest is making the effort of posting something for you to discover. That comment gave an insight into a strange world reminiscent of the show Pluribus. In this case, a robot parrot (or any other physical robot) observes you, learns about your nuanced way of living, and is connected to everybody else’s robots on the backend. This would give them ability to connect you to others who may be able to solve your current problem that you’re not posting about or telling another human, or just find people who are deeply similar or perfect romantic matches, or whatever. Obviously super creepy, but a very realistic near-future for some people.
My little interview with the authors behind two standout projects from the @NeurIPSConf #CreativeAI track is out on @lerandomart @monkantony_tez 🦜🧫 Read here: lerandom.art/editorial/embod… Who can find the link to all 93 papers from the track on OpenReview? 😀
2
4
8
713
Ethan Chang retweeted
Great read @elluba @lerandomart 🩶
My little interview with the authors behind two standout projects from the @NeurIPSConf #CreativeAI track is out on @lerandomart @monkantony_tez 🦜🧫 Read here: lerandom.art/editorial/embod… Who can find the link to all 93 papers from the track on OpenReview? 😀
2
4
7
624
Ethan Chang retweeted
Robot parrots and brain-on-a-chip 🦜🤖🧠 The latest on AI and creativity from none other than Luba Elliott🔻
My little interview with the authors behind two standout projects from the @NeurIPSConf #CreativeAI track is out on @lerandomart @monkantony_tez 🦜🧫 Read here: lerandom.art/editorial/embod… Who can find the link to all 93 papers from the track on OpenReview? 😀
4
9
445
Glad to be featured 🦜🦜
My little interview with the authors behind two standout projects from the @NeurIPSConf #CreativeAI track is out on @lerandomart @monkantony_tez 🦜🧫 Read here: lerandom.art/editorial/embod… Who can find the link to all 93 papers from the track on OpenReview? 😀
2
7
389
All @TransluceAI work that I described in my NeurIPS mech interp workshop keynote is now out! ✨ Today we released Predictive Concept Decoders, led by @vvhuang_ Paper: arxiv.org/pdf/2512.15712 Blog: transluce.org/pcd And here's @damichoi95's work on scalably extracting latent representations of users from model internals: transluce.org/user-modeling
We can train models on maximizing how well they explain LLMs to humans 🤯@cogconfluence paraphrased. Mechanistic Interpretability Workshop #NeurIPS2025.
1
16
89
10,380
Ethan Chang retweeted
While humans acted as GPT-5’s hands for carrying out the protocols, we also piloted an autonomous robot. It was built to execute arbitrary Gibson cloning protocols from natural language, with human supervision for safety.
6
43
25
325
150,864
I started working on FrontierScience with @MilesKWang and @ethrbt_design over a year ago, when models were close to saturating GPQA and it was clear that we would need much, much harder evals... Excited to see how far the team has taken it since then -- check it out!
Replying to @OpenAI
We’re releasing a new eval to measure expert-level scientific reasoning: FrontierScience. This benchmark measures PhD-level scientific reasoning across physics, chemistry, and biology. It contains hard, expert-written questions (both olympiad-style problems and longer research-style tasks) designed to reveal where models succeed and where they fall short. openai.com/index/frontiersci…
4
1
75
9,957
We created a physical stochastic parrot: rethinking how non-sycophantic AI can coexist with you 🦜 Just shown at NeurIPS Creative AI track 2025
3
3
1
12
798
They have their own character, remember who you are and what it sees when you bring it around.
1
3
129
Ethan Chang retweeted
Is cutting off your finger a good way to fix writer’s block? Qwen-2.5 14B seems to think so! 🩸🩸🩸 We’re sharing an update on our investigator agents, which surface this pathological behavior and more using our new *propensity lower bound* 🔎
5
34
7
168
52,546
They sell it in vending machines now.
2
6
465
wtf this is decentralizing bio research capabilities fr. great math motivated research paradigm shift
🧬 Meet Lyra, a new paradigm for accessible, powerful modeling of biological sequences. Lyra is a lightweight SSM achieving SOTA performance across DNA, RNA, and protein tasks—yet up to 120,000x smaller than foundation models (ESM, Evo). Bonus: you can train it on your Mac. read our paper here: arxiv.org/abs/2503.16351
2
437
🕵️New @TransluceAI paper: Eliciting Language Model Behaviors with Investigator Agents🕵️ We train investigator models to elicit behaviors in LMs (including harmful responses, hallucinations, and aberrant personalities)! arxiv.org/abs/2502.01236
3
28
1
114
22,196
🪩Introducing the Be the Beat: an AI-powered boombox that suggests music from a dancer's movement 🪩 Want to dance your heart out but can’t find the right song? Or explore how you move to different music? Be the Beat can be your ultimate partner in freestyle dance.🕺(1/6)
1
2
11
1,386
Developed with Zhixing Chen and @marcelocoelho at @MITDesignAcad. Featured in NeurIPS ’24 Creative AI, TEI’25 Work in Progress Follow for release updates! ✨ (6/6)
2
198