Changbai
@changbai.bsky.social
Opinions are not my own, but merely a dance of signals traversing the neural network. MS AI @ Oregon State University | Creative Technologist | AI Safety | memes that aren't funny
Reposted by Changbai
I will always love this blog post. You, yes you, have a responsibility to the world not to respond to bad incentives
every claim that "the incentives" support or deter certain kinds of behavior is also a statement about what kinds of external signals the claimant views as rewards or penalties #linklog
No, it’s not The Incentives—it’s you
There’s a narrative I find kind of troubling, but that unfortunately seems to be growing more common in science. The core idea is that the mere existence of perverse incentives is a valid and…
talyarkoni.org
December 25, 2025 at 1:02 AM
I will always love this blog post. You, yes you, have a responsibility to the world not to respond to bad incentives
Reposted by Changbai
I think all the time about how Ray Kurzweil's dad died young and then Kurzweil spent his entire life imagining a world where a magic moment would arrive where technology suddenly advanced so quickly he could somehow get his dad back.
It's very sad and very sweet, but that does not make it real.
It's very sad and very sweet, but that does not make it real.
IN DEPTH | If you could speak to your dead grandmother forever, would you?
If you could speak to your dead grandmother forever, would you?
www.independent.co.uk
December 10, 2025 at 11:05 PM
I think all the time about how Ray Kurzweil's dad died young and then Kurzweil spent his entire life imagining a world where a magic moment would arrive where technology suddenly advanced so quickly he could somehow get his dad back.
It's very sad and very sweet, but that does not make it real.
It's very sad and very sweet, but that does not make it real.
Reposted by Changbai
Today, we're opening up Seams Alpha (seams.so), a social web annotation built in the Atmosphere. We're hoping it helps with collective sense-making.
This is just the start, there's still lots to do!
We'd love if you tried it with us.
This is just the start, there's still lots to do!
We'd love if you tried it with us.
Lab Notes #001 - Sealight Lab Notes
Announcing Seams
sealight.leaflet.pub
December 8, 2025 at 7:37 PM
Today, we're opening up Seams Alpha (seams.so), a social web annotation built in the Atmosphere. We're hoping it helps with collective sense-making.
This is just the start, there's still lots to do!
We'd love if you tried it with us.
This is just the start, there's still lots to do!
We'd love if you tried it with us.
Happy Noam Chomsky Day!
www.youtube.com/watch?v=dXE6...
www.youtube.com/watch?v=dXE6...
Happy Noam Chomsky Day - Captain Fantastic
YouTube video by spoon2001
www.youtube.com
December 8, 2025 at 1:43 AM
Happy Noam Chomsky Day!
www.youtube.com/watch?v=dXE6...
www.youtube.com/watch?v=dXE6...
Reposted by Changbai
Let me tell you a story. Perhaps you can guess where this is going... though it does have a bit of a twist.
I was poking around Google Scholar for publications about the relationship between chatbots and wellness. Oh how useful: a systematic literature review! Let's dig into the findings. 🧵
I was poking around Google Scholar for publications about the relationship between chatbots and wellness. Oh how useful: a systematic literature review! Let's dig into the findings. 🧵
December 5, 2025 at 10:35 PM
Let me tell you a story. Perhaps you can guess where this is going... though it does have a bit of a twist.
I was poking around Google Scholar for publications about the relationship between chatbots and wellness. Oh how useful: a systematic literature review! Let's dig into the findings. 🧵
I was poking around Google Scholar for publications about the relationship between chatbots and wellness. Oh how useful: a systematic literature review! Let's dig into the findings. 🧵
Reading math forum discussions where mathematicians attempt to crack open long-standing problems make me wish all online discourses are like that. Civil, constructive, curious. Humans working together to advance knowledge. Perhaps the atmosphere comes from the feeling of awe?
Erdős Problem #124 - Discussion thread
www.erdosproblems.com
November 30, 2025 at 11:23 PM
Reading math forum discussions where mathematicians attempt to crack open long-standing problems make me wish all online discourses are like that. Civil, constructive, curious. Humans working together to advance knowledge. Perhaps the atmosphere comes from the feeling of awe?
How it feels to receive unprompted support texts from family and friends from the other side of the continent, and other side of the earth, as I struggle to finish my thesis:
a group of people are standing in a field with their arms up
ALT: a group of people are standing in a field with their arms up
media.tenor.com
November 27, 2025 at 9:52 AM
How it feels to receive unprompted support texts from family and friends from the other side of the continent, and other side of the earth, as I struggle to finish my thesis:
Reposted by Changbai
We are not "banning" reviews; we are just requiring peer review first. Good review articles are important for the field!
You can’t really blame arXiv for the decision to stop publishing computer science stuff (given the flood of slop) but this is also a textbook example of a global public good being gratuitously degraded www.nature.com/articles/d41...
Preprint site arXiv is banning computer-science reviews: here’s why
The repository is taking steps to tackle a surge in low quality, AI-generated content.
www.nature.com
November 8, 2025 at 6:57 PM
We are not "banning" reviews; we are just requiring peer review first. Good review articles are important for the field!
Reposted by Changbai
“…a major advance that we’ve been trying to achieve for a very long time, and a great aid to scholarship. It allows human beings to focus their time on the important, profound work of understanding another human being, rather than staring at a curlicue to grasp if it’s an L or an I.”
New issue of my newsletter: "The Writing Is on the Wall for Handwriting Recognition" — One of the hardest problems in digital humanities has finally been solved, and it's a good use of AI newsletter.dancohen.org/archive/the-...
The Writing Is on the Wall for Handwriting Recognition
One of the hardest problems in digital humanities has finally been solved
newsletter.dancohen.org
November 26, 2025 at 1:02 AM
“…a major advance that we’ve been trying to achieve for a very long time, and a great aid to scholarship. It allows human beings to focus their time on the important, profound work of understanding another human being, rather than staring at a curlicue to grasp if it’s an L or an I.”
Reposted by Changbai
Sora is here and will only get better. Video provenance will become weird. LLMs for coding are imperfect but are serious productivity boosts. These are just the first things to really matter
November 9, 2025 at 5:31 AM
Sora is here and will only get better. Video provenance will become weird. LLMs for coding are imperfect but are serious productivity boosts. These are just the first things to really matter
Reposted by Changbai
New Google DeepMind paper: "Consistency Training Helps Stop Sycophancy and Jailbreaks" by @alexirpan.bsky.social, me, Mark Kurzeja, David Elson, and Rohin Shah. (thread)
November 4, 2025 at 12:18 AM
New Google DeepMind paper: "Consistency Training Helps Stop Sycophancy and Jailbreaks" by @alexirpan.bsky.social, me, Mark Kurzeja, David Elson, and Rohin Shah. (thread)
Reposted by Changbai
The challenge in learning using AI is very similar to the same learning issue discovered about internet search
When we are given answers we think we learn, but we don’t. Learning is work. However, things like the “learning modes” from the AI providers help, as does using AI for tutoring not answers
When we are given answers we think we learn, but we don’t. Learning is work. However, things like the “learning modes” from the AI providers help, as does using AI for tutoring not answers
October 31, 2025 at 1:55 PM
The challenge in learning using AI is very similar to the same learning issue discovered about internet search
When we are given answers we think we learn, but we don’t. Learning is work. However, things like the “learning modes” from the AI providers help, as does using AI for tutoring not answers
When we are given answers we think we learn, but we don’t. Learning is work. However, things like the “learning modes” from the AI providers help, as does using AI for tutoring not answers
While the outcome of science is a tower we build collectively, the process of science is more akin to the exploration of a maze. So do not be discouraged when your finding is null or method fails. Taking steps back is just as important as steps forward when in a maze.
October 31, 2025 at 4:40 AM
While the outcome of science is a tower we build collectively, the process of science is more akin to the exploration of a maze. So do not be discouraged when your finding is null or method fails. Taking steps back is just as important as steps forward when in a maze.
Reposted by Changbai
This is a very good paper on how methodologies from comparative cognition can help improve AI evaluation:
openreview.net/pdf?id=gCPJF...
Other papers along the same lines:
⬇️
openreview.net/pdf?id=gCPJF...
Other papers along the same lines:
⬇️
openreview.net
June 20, 2025 at 7:58 PM
This is a very good paper on how methodologies from comparative cognition can help improve AI evaluation:
openreview.net/pdf?id=gCPJF...
Other papers along the same lines:
⬇️
openreview.net/pdf?id=gCPJF...
Other papers along the same lines:
⬇️
Reposted by Changbai
Hands off working COBOL code - via @krypt3ia.bsky.social infosec.exchange/@krypt3ia/11...
Krypt3ia (@krypt3ia@infosec.exchange)
Attached: 1 image
infosec.exchange
April 6, 2025 at 12:50 AM
Hands off working COBOL code - via @krypt3ia.bsky.social infosec.exchange/@krypt3ia/11...
Reposted by Changbai
United States researchers surveyed 135 long COVID patients and interviewed 29 more.
They found that long haulers face major information barriers and have become their own health advocates, sharing stories, doing research, and building support networks.
www.sciencedirect.com/science/arti...
They found that long haulers face major information barriers and have become their own health advocates, sharing stories, doing research, and building support networks.
www.sciencedirect.com/science/arti...
Theorizing forgotten crisis publics: COVID long haulers’ information marginalization
An estimated 400 million people live with long COVID. Long COVID is a chronic condition with more than 200 symptoms, many of them debilitating. Alarmi…
www.sciencedirect.com
March 26, 2025 at 11:52 AM
United States researchers surveyed 135 long COVID patients and interviewed 29 more.
They found that long haulers face major information barriers and have become their own health advocates, sharing stories, doing research, and building support networks.
www.sciencedirect.com/science/arti...
They found that long haulers face major information barriers and have become their own health advocates, sharing stories, doing research, and building support networks.
www.sciencedirect.com/science/arti...
An anecdotal observation: it seems easier for physicists, mathematicians, biologists, philosophers, and security researchers to take catastrophic AI risk seriously, and harder for machine learning researchers to do so. If this is true, I wonder why.
March 23, 2025 at 12:47 AM
An anecdotal observation: it seems easier for physicists, mathematicians, biologists, philosophers, and security researchers to take catastrophic AI risk seriously, and harder for machine learning researchers to do so. If this is true, I wonder why.
I wonder why polymaths are a lot more common before modernity. It probably mostly has to do with inequality restricting opportunity to education to only the leisure class.
March 7, 2025 at 7:39 AM
I wonder why polymaths are a lot more common before modernity. It probably mostly has to do with inequality restricting opportunity to education to only the leisure class.
Reposted by Changbai
Third suicide associated with generative AI, and maybe first clear case of using generative AI to plan a crime.
January 7, 2025 at 10:39 PM
Third suicide associated with generative AI, and maybe first clear case of using generative AI to plan a crime.
