iiinova — where your digital self keeps evolving

Reposted by @kyunghyuncho.bsky.social

Lightning AI ⚡️ @lightningai.bsky.social · 01/02/2026

Our CEO William Falcon joined NYU's Global AI Frontier Lab advisory council with @ylecun.bsky.social and @kyunghyuncho.bsky.social. Fireside chat on the PhD-to-startup journey and the Voltage Park merger. Thanks to NYU and Jon Krohn 🙏

041

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 22/01/2026

i was made aware of miscitations thanks to the GPTZero team (cc @alexcdot). ji won and i quickly checked them ourselves and have posted what happened on openreview: openreview.net/forum?id=IiE.... we have already notified NeurIPS'25 PC's about this issue. (1/2)

150

Reposted by @kyunghyuncho.bsky.social

Nicholas Lourie @nicholaslourie.bsky.social · 08/10/2025

📄🔈✨ Deep learning is an empirical science, but we rely on basic empirical methods. What might a better foundation—a simple theory—for empirical work look like? @kyunghyuncho.bsky.social, He He, and I move towards one in "Hyperparameter Loss Surfaces Are Simple Near their Optima" at #COLM2025! 🧵1/9

172

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 06/10/2025

i'll be at COLM on thursday (oct 9). DM me for a chat!

040

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 05/10/2025

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!

182

Reposted by @kyunghyuncho.bsky.social

Mila - Institut québécois d'IA @mila-quebec.bsky.social · 02/09/2025

Exciting news! We're thrilled to announce the appointment of Professor Hugo Larochelle as Mila's new Scientific Director! A deep learning pioneer and former head of Google's AI lab in Montreal, Hugo's leadership will be pivotal in advancing AI for the benefit of all. mila.quebec/en/news/hugo...

mila.quebec

Hugo Larochelle becomes the new Scientific Director of Mila | Mila

Montreal (Quebec), September 2, 2025 – Mila, the Quebec Artificial Intelligence Institute, announces the appointment of Hugo Larochelle—Adjunct Professor at the Université de Montréal and former head ...

2397

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 28/08/2025

amen those with savior and superiority complex, obsessed with sci fi, blinded by dollar signs, devoid of empathy, and severely gullible.

091

Reposted by @kyunghyuncho.bsky.social

FuriosaAI @furiosa.ai · 14/08/2025

Thank you to everyone in the AI community we met in NYC this summer and those who came to our happy hour at @nyu.edu’s Global AI Frontier Lab last week.

111

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 13/08/2025

everyone says google will lose search biz to llm, but in reality, all llm co's are so busy trying to reproduce google search. irony

071

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 12/08/2025

recently gave a talk on <Reality Checks> at two venues, and discussed (and rambled) about how leaderboard chasing is awesome (and we want it to continue) but that this isn't easy because everyone (me! me! me!) wants to write more papers. the link to the slide deck in the reply.

3245

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 12/08/2025

can’t really vibe code healthcare. a thoughtful and experience-rich talk by amazing Lavender Jiang at NYU Global AI Frontier Lab!

010

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 08/08/2025

vibe coding continues with a MS Edge extension: the world's most expensive, slow and inaccurate ad blocker, featuring Gemini!

330

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 07/08/2025

the perfect response from Google cannot be simpler: just halve the price of everything. please start with the Gemini Pro subscription. pretty please?

040

Reposted by @kyunghyuncho.bsky.social

Andrew Lampinen @lampinen.bsky.social · 05/08/2025

These biases can lead to dramatic downstream effects that cause unexpected conclusions from analyses. For example, RSA may identify two models computing the same, complex task as much less representationally-similar than either of them is to a model computing a much simpler task (right panel)!

RSA within and between different sets of models can give surprising results due to representation biases. This plot shows similarities within and between different models computing different types of features. Ideally the similarities would be highest in blocks on the diagonal (i.e. models computing the same features), and the blocks off the diagonal would show graded similarity corresponding to the functional overlap. However, that is not the case. (Left) When comparing a model trained to output both easy and hard features to ones that are trained on only one feature, the multi-task model appears very similar to the easy-task only model (cf. Hermann and Lampinen, 2020). In fact, the models trained only on the hard task do not even appear particularly similar to other models trained on the same exact task. (Right) When models are trained on multiple easy or multiple hard tasks, the models trained on only hard tasks appear less similar to other models trained on exactly the same tasks than they do to models trained on strictly easier tasks that use the same input units.

171

Reposted by @kyunghyuncho.bsky.social

FuriosaAI @furiosa.ai · 29/07/2025

🛬 Next week, we’re hosting our inaugural happy hour: Powered by Furiosa. 🍹 On August 8, we’re bringing together industry leaders and practitioners at the forefront of AI to discuss, share knowledge, and enjoy great food and drinks.

linkedin.com

Join us at our inaugural happy hour: Powered by Furiosa. This event will bring together industry practitioners and leaders at the forefront of AI inference. Connect with NYC researchers, engineers, a...

101

Reposted by @kyunghyuncho.bsky.social

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 25/07/2025

You know how RL is that whole big thing nowadays? Present your work at the first-ever New York Reinforcement Learning Workshop (NYRL), co-organized by Amazon, Columbia Business School & NYU Tandon School of Engineering. ny-rl.com!

1264

Reposted by @kyunghyuncho.bsky.social

NYU Center for Data Science @nyudatascience.bsky.social · 25/07/2025

CDS professors have launched the Global AI Frontier Lab in Brooklyn with Korean partners like KAIST. Profs @kyunghyuncho.bsky.social and @yann-lecun.bsky.social lead work on fundamental AI, responsible AI, and healthcare, and a biweekly seminar series. nyudatascience.medium.com/building-bri...

nyudatascience.medium.com

Building Bridges Across the Pacific: How CDS Launched a New Model for International AI Research…

CDS has launched a new research lab in Brooklyn in partnership with Korean institutions, leading cross‑Pacific collaborations in AI.

011

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 21/07/2025

😂 i think this “challenging problem” may have been finally solved after five years. === Understanding and creating mathematics using natural mathematical language … used by humans is a challenging and important problem for driving progress in machine learning. ===

3161

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 19/07/2025

thanks all for the interest! i have put the postdoc in touch with many amazing folks.

010

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 19/07/2025

japan’s let-them-eat-cake moment: which japanese politician will say “let them eat cake”?

020

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 14/07/2025

ok truly done with my two weeks of vibe-coding (catching up with the trend): the final day ended with local inference using gemma3 via ollama for the medical advisor app for Mac OS. it's slow (bc my laptop) but it works to some extent. local models, local inference and privacy will win!

171

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 13/07/2025

EPIC never sent me any offer, not to mention a zuck-level one, and thus, out of spite, i've tried to build a Mac native app of my personal medical advisor. still waiting for a call with a zuck-level offer from EPIC 🤑

140

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 10/07/2025

some people are way too comfortable with unfalsifiable assumptions over empirical results or simulation based results. sad

040

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 08/07/2025

yet another day of vibe-coding: people use LLM as their bespoke medical advisor. what if this advisor can be super-charged with your own medical records? download your medical records from Epic App, ingest them into db, and have Gemini answer your question based on your own medical records for you!

190

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 06/07/2025

to cap the weekend of vibe coding, i asked Gemini to create a web app version, and it did it perfectly in one go.

100

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 06/07/2025

vibe-coding day 2: i decided to rewrite the app from yesterday using a more native app platform. never used xcode nor swift before ... and voila. two hours and here's a full-blown app. am i now a swift developer?

120

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 06/07/2025

since vibe coding is all the rage, i vibe coded an absolutely useless app today with Gemini. `gemini-vision` lets you capture a screen and query Gemini to answer your question about the captured screen. vibe coding is fun, but i infinitely prefer bbq over vibe coding

260

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 05/07/2025

some educational systems are exceptionally good at producing cheaters.

020

Reposted by @kyunghyuncho.bsky.social

Ben Recht @beenwrekt.bsky.social · 04/07/2025

Pearson's RNGs: www.argmin.net/p/sounds-lik... That said, I could make the case that computers today wouldn't look the way they do without the Statistics revolution of the 1930s.

argmin.net

Sounds Like Noise to Me

Computational randomness in the 1920s

021

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 04/07/2025

"But I’m particularly fascinated by the weird obsession with statistics" amen

120

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 03/07/2025

really, apple? this is the most important thing to say about this gigantic update?

230

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 02/07/2025

i was asked by a few (inc. @yuanqingwang.bsky.social) what i meant by this earlier tweet, and since i'm pretty busy, i decided to write a blog post to answer the question 🤦

230

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 23/06/2025

i have the perfect solution to the nuclear power plant debate: build permanent high-grade nuclear waste facilities in Aspen and Park City. both have low pop density, solid foundation, good road access and are relatively remote. you’re welcome

150

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 21/06/2025

wow … we could’ve collectively saved so much time of reviewers only if this paper from 2005 was more widely known.

150

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 21/06/2025

i was finally convinced by sungmin and sangheum to work on unlearning. the first we did (or i learned) together with sungjin and dason was to learn how people evaluate unlearning, and horror 😱 here is our short writeup and report on what we found and what we propose as a remedy.

291

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 15/06/2025

the new administration in Korea has managed to bring onboard Jungwoo as the secretary of AI and future technologies to directly report to and advise the President. some real AI researcher, developer and director! a great job!

1170

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 12/06/2025

finally, wind is changing its direction: causal inference becomes easier if we give up on designing a new estimation algorithm ourselves (i don't think we've evolved to do so ourselves well.) let learning find one for you!

1100

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 12/06/2025

“every bar is local” - Henri Dwyer (2025)

020

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 11/06/2025

ai for drug discovery looks a lot like machine translation research/development during the cold war.

130

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 10/06/2025

is consulting as a job finally at its end? 🥳 for some uninteresting reasons, i thought it would be interesting to find all AI-related courses at NYU and create a report on AI education at NYU, but i'm lazy. (1/3)

130

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 03/06/2025

you can say the authors of <BLEND: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages> were dedicated by looking at Table 1.

191

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 31/05/2025

society has decided we need to invest in long-term science for the benefit of all not just shareholders or rich patrons, and universities have agreed to step up beyond education to provide venues for conducting science on the society’s behalf via federal funding based on tax.

2293

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 24/05/2025

a thought provoking and also much better grounded interview of Mina Lee at U Chicago about AI and its impact (in korean): www.newstheai.com/news/article...

040

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 23/05/2025

BREAKING: single-issue voters continue to be punched in their faces over and over again.

130

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 20/05/2025

it's been more than a decade since KD was proposed, and i've been using it all along .. but why does it work? too many speculations but no simple explanation. Sungmin and i decided to see if we can come up with the simplest working description of KD in this work. (1/4)

2406

kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 16/05/2025

don't try to make a consensus about the future. future is uncertain.

021