Harry Zhao (@theharryzhao) 's Twitter Profile
Harry Zhao

@theharryzhao

PhD Candidate at @mcgillu (Brain-Inspired Computational Reasoning), advised by Doina Precup & Yoshua Bengio

A true friend who roasts you and learns with you.

ID: 1265747761594540043

linkhttps://scholar.google.com/citations?user=TZh189YAAAAJ&hl=en calendar_today27-05-2020 20:52:46

115 Tweet

709 Followers

1,1K Following

Harry Zhao (@theharryzhao) 's Twitter Profile Photo

#Delusions AREN'T #hallucinations! Don't be confused! When agents can't reject some obviously wrong beliefs, despite conflicting evidence, delulus emerge; While, hallucination describes the formation of false beliefs. Delulus involve BOTH belief formation💡& evaluation🤔systems

Danijar Hafner (@danijarh) 's Twitter Profile Photo

Excited to introduce Shortcut Models! Shortcut models generalize diffusion/flow models by conditioning not just on the time step but also on the desired step size. This allows for high-quality generations in a single forward pass. Learn more below 👇

Excited to introduce Shortcut Models!

Shortcut models generalize diffusion/flow models by conditioning not just on the time step but also on the desired step size.

This allows for high-quality generations in a single forward pass. Learn more below 👇
Eric Elmoznino (@ericelmoznino) 's Twitter Profile Photo

1/5 Very excited to announce our new paper that formally defines compositionality, giving us theoretical insight into what it actually is and helping us build it into AI models. This was work with Thomas Jiralerspong, Yoshua Bengio, and Guillaume Lajoie. arxiv.org/abs/2410.14817…

Sitao Luan (@sitaoluan) 's Twitter Profile Photo

Watch out your graph before learning on it, because heterophilic graph is everywhere and it can destroy your model badly. Come to learn about heterophilic graph learning with Will Hua, Qincheng Lu and me at Learning on Graphs Conference 2024 . Check out sites.google.com/mila.quebec/he… for updates.

Watch out your graph before learning on it, because heterophilic graph is everywhere and it can destroy your model badly. Come to learn about heterophilic graph learning with <a href="/WilliamCQHua/">Will Hua</a>, Qincheng Lu and me at <a href="/LogConference/">Learning on Graphs Conference 2024</a> . Check out sites.google.com/mila.quebec/he… for updates.
Raymond Chua (@raymondrchua) 's Twitter Profile Photo

It’s been a long time coming, but I’m thrilled to share my first research paper with Arna Ghosh , Christos Kaplanis, Blake Richards, and Doina! Just accepted to NeurIPS 2024 (see u in Vancouver! 🇨🇦). This will be a longer thread—thanks for following along! arxiv.org/abs/2410.22133 1/11

Lex Fridman (@lexfridman) 's Twitter Profile Photo

Scientific American editor-in-chief is an embarrassment to science. Get politics out of science, engineering, STEM education. Science and engineering is about building cool shit, and exploring the mysteries of physics, chemistry, biology, etc.

CTV News (@ctvnews) 's Twitter Profile Photo

Peanut, the social media star squirrel euthanized after being taken from home, tests negative for rabies ctvnews.ca/world/peanut-t…

Yoshua Bengio (@yoshua_bengio) 's Twitter Profile Photo

Mila - Institut québécois d'IA launches its annual Impact Report! This is a pivotal moment in AI with repercussions across all sectors, and I'm extremely proud of our 1550+ members who contribute to the development of this science for the benefit of all.

Richard Sutton (@richardssutton) 's Twitter Profile Photo

"In a world of change, the learners shall inherit the earth, while the learned shall find themselves perfectly suited for a world that no longer exists." - Eric Hoffer, philosopher and author

Wenhu Chen (@wenhuchen) 's Twitter Profile Photo

It seems that lots of people don't know the backstory of this. The very brief version is that: A large amount (probably more than half) of Chinese-nationality STEM MS/PHD applicants accept offers from Canadian universities but couldn't get their visa on time. Canadian

Harry Zhao (@theharryzhao) 's Twitter Profile Photo

Hi everyone, I have a new paper about how to make planning agents avoid delusional subgoals, which will be discussed tomorrow at the NeurIPS SafeGenAI workshop (3-5 PM, in East Exhibition Hall A) Please drop by! #NeurIPS2024 #NeurIPS #AIsafety

Tongtian Zhu (@tongtian_zhu) 's Twitter Profile Photo

ICML 2025's rebuttal process be like🤣: 👨‍💻 Authors: spend a whole week writing a careful rebuttal ✅ Reviewer: clicks "acknowledge" without reading 🚫 Author: not allowed to reply anymore So what does acknowledge mean here? "You speak. I pretend to listen. Conversation over."🙃

Andrei Lupu (@_andreilupu) 's Twitter Profile Photo

#ICML reviewer clicked a button saying they read my rebuttal, so it must be true. Nevermind that they visibly didn't read entire sections of the paper before writing their review.

Richard Sutton (@richardssutton) 's Twitter Profile Photo

The PhD thesis of my _first_ PhD student, Doina Precup, is at-long-last available in digital form. Title: Temporal Abstraction in Reinforcement Learning Url: incompleteideas.net/papers/Precup_… Abstract: Decision making usually involves choosing among different courses of action over a