Profile avatar
lowd.bsky.social
CS Prof at the University of Oregon, studying adversarial machine learning, data poisoning, interpretable AI, probabilistic and relational models, and more. Avid unicyclist and occasional singer-songwriter. He/him
386 posts 3,992 followers 486 following
Regular Contributor
Active Commenter
Posts 22 Comments 28

New research from Costello, Pennycook, and Rand. Can conversations with AI help reduce belief in conspiracy theories? Quite possibly. What’s the mechanism? Evidence production. Let me explain my little-t theory about how this might work.

submitted 8 days ago • 11 comments

New blog post: Stop talking about AGI, it's lazy and misleading I argue that we all could do better than using the slippery, undefinable term Artificial General Intelligence. We would have a much better discussion by focusing on concrete tasks and capabilities togelius.blogspot.com/2025/01/stop...

submitted 35 days ago • 1 comment

Here are my notes on OpenAI's new ChatGPT Operator browser "agent", including initial thoughts on their approach to mitigating prompt injection risks simonwillison.net/2025/Jan/23/...

submitted 35 days ago • 9 comments

Here's the talk. I hope you enjoy it www.youtube.com/watch?v=rwfY...

submitted 36 days ago • 0 comments

If the current AI gold rush is anything like the dot com boom, some of these ventures will end up like pets dot com and some will end up like Amazon dot com.

submitted 38 days ago • 2 comments

After all these years, The Hampsterdance Album is still a masterpiece.

submitted 41 days ago • 0 comments

AI is like plastic — a lot of people hate it because it often comes across as fake and tacky, but it’s flexible and it’s cheap and there’s so many things that would be impossible or impractical without it. And yes, a lot of AI will be junk. Just like everything else. (cf Sturgeon’s Law.)

submitted 42 days ago • 1 comment

I think social media would be better if there were separate servers for PVP and PVE, just like in WoW. Let everyone who wants to troll, dunk, flame, and pile-on do so in their own space. And then let the rest of us gain xp, form parties, collect achievements, or whatever, without being ganked.

submitted 45 days ago • 0 comments

Garbage in, garbage out — and it may not take much garbage to do real harm! This is why data curation and model testing is so important.

submitted 46 days ago • 1 comment

My TV debut! On New Year's Eve, I spoke with Global News about AI — past and future, pros and cons. I think it went well! www.youtube.com/watch?v=LP0x...

submitted 47 days ago • 0 comments

(sung to my very muddy dog as I towel her off) "Shouldn't a dog who's so good inside Have a matching exterior? And since folks here to an absurd degree Seem fixated on your fur debris, Would it be alright by you If I de-muddified you?"

submitted 52 days ago • 0 comments

My first new book of the year is out — DISCOVERY OF THE WANDERLUST. If you love B’Elanna Torres from Star Trek: Voyager, you might like this book. If you like reading about animal characters on spaceships, you’re very likely to like this book. marylowd.com/books/#wande...

submitted 54 days ago • 0 comments

Shameless plug: If you like Star Trek and animals, you'll probably enjoy this series, written by my spouse. Book 2 was just published on January 2nd! www.amazon.com/dp/B0DQ4YXWVJ

submitted 56 days ago • 0 comments

Me: Hey Claude, could you help me build a simple app? Claude: Sure! Here's 100 different files and a setup script... Me: Wait, what's "npm"? [...2 hours later...] Me: Hey ChatGPT, could you help me build a REALLY simple app? [...2 hours later...] Me: So Claude, it's me again...

submitted 60 days ago • 2 comments

I’m a computer scientist, when I give a movie 7/10 it truncates to 0.

submitted 64 days ago • 0 comments

Merry Christmas, friends and colleagues! I've written a short letter reflecting on our progress. Thank you for everything. www.cs.toronto.edu/~cmaddis/let...

submitted 64 days ago • 3 comments

I stumbled on @eugenevinitsky.bsky.social 's blog and his "Personal Rules of Productive Research" is very good. I now do a lot of things in the post, & wish I had done them when I was younger. I share my "mini-paper" w ppl I hope will be co-authors. www.eugenevinitsky.com/posts/person...

submitted 74 days ago • 6 comments

I am a broken record on this, but everything strengthens the case for open access. There is no plausible path leading to a world where "no one can train on my words"; what we can do is ensure that everyone can train on our words.

submitted 72 days ago • 5 comments

I think many of the (quite gross) reactions to this are not grappling yet with how many their students already have what they think is this product in the form of chatgpt.

submitted 75 days ago • 3 comments

LLM output today may seem glib, generic, repetitive, and unnecessarily long-winded, but it can’t compare to the shallow banality of BlueSky discourse.

submitted 73 days ago • 0 comments

We wrapped up CS 8803 "Large Language Model" class at Georgia Tech for Fall 2024. Here is the reading list: • learning from human preferences (PPO, DPO, SimPO, CPO, RRHF, ORPO, CTO) • real-world LLM (Llama-3, Aya, Arena's) • efficient LLM (MoMa, LoRA, QLoRA, LESS)

submitted 86 days ago • 2 comments

Openly transgender or nonbinary US adults, among those age 18-29: 5.1% 30-49: 1.6% 50+: 0.3% 44% of US adults personally know someone who is trans. 20% know someone who is nonbinary. www.pewresearch.org/short-reads/...

submitted 90 days ago • 8 comments