Thinking about AI. Trying to build a Rat, EA, TPOT meetup scene in Asheville, NC
Matt Brooks
Karma: 310
whoops, fixed!
Which metric would you use to compare welfare across species?
I don’t think we know enough about consciousness/qualia/etc. to say anything with conviction about what it’s like to be a nematode. And operationally, I don’t think you won’t be able to convince enough people/funders to take real action on soil animals because it’s just too epistemically unsound and doesn’t fit into people’s natural world views.
When I say net negative, I don’t mean if you try to help soil animals you somehow hurt more animals on the whole.
I mean that you will turn people away from the theory of animal suffering because advocating for soil animals will make them think the field/study of animal suffering as a whole is less epistemically sound or even common sense as they previously thought.
I’m going to write a post next week about this, but consider the backlash on twitter regarding Bentham’s Bulldog’s post about bees and honey. More people came out in force against him than for him. I think that post, for instance, reduced the appetite for animal suffering discussion/action
Thank you!
On the code sharing. Yes, I thought about it, but it would take us a bit of effort to pull it all together and publish it online, I didn’t want to spend that effort if no one was going to get value from it. So far, no one has found the courage and 3 seconds of effort to put a comment asking for the data/code (or more likely, people just don’t want to spend the time wading through the code/data)
On nematodes, I think 169x the total number of neurons compared to humans is a poor/confused way to attempt to measure total welfare. And I think the second order effects of trying to convince people they should care about nemotodes (unless they are already diehard EA) is likely net negative for the animal suffering cause at large.
From Rob (waiting for his comment to be approved):
Thanks for trying Winnow! My guess is that you were redirected to the homepage after logging in and created a fresh document (no reviewers included by default). Now that you’re logged in, try creating a document directly from this page and it should work: https://www.winnow.sh/templates/ea-rationalist-writing-assistant
On the Egregore / religion part
I agree! Egregore is occult so definitely religion-adjacent. But I also believe EA as a concept/community is religion-adjacent (not necessarily in a bad way).
It’s a community, ethical belief, there is suggested tithing, sense of purpose/meaning, etc.
Funny—I don’t think it feels written by a critic, but definitely a pointed outsider (somewhat neutral?) 3rd party analysis.
I do expect the Egregore report to trigger some people (in good and bad ways, see the comment below about feeling heard). The purpose is to make things known that are pushed into the shadows, the good and the bad. Usually things are pushed into the shadows because people don’t want to or can’t talk about them openly.
Wow, I’m so happy something in our analysis was able to make you feel seen and heard 💙
I agree that there are issues in EA and I also hope the community is able to build up the courage and effort to solve them
Thanks for sharing!
Thanks for the kind words!
We’d be very happy to jump on a call and discuss our methods in more detail, share the code, datasets, etc.
I’ll let @alejbo take this question—I think it’s a good one
Although at the high level I somewhat disagree with “I don’t think chatbots are very good at epistemology”, my guess would be they’re better than you think, but I agree not perfect or amazing
But as you admit, most humans aren’t either, so it’s already a low bar
Thank you! 🙏🏻
I’d ask you to consider, when have you ever taken action due to a 0.1–1 min video?
I think basically no one takes action from any video, even 30 min high quality YouTube videos.But what you get when you have someone watch your 1 min video is that their feed will steer in that direction and they will see more videos from your channel and other AI Safety-aligned channels.
I think this might be where a lot of the value is.
If you can get 20M people to watch a few 45 second videos, you are making the idea more salient in their minds for future videos/discussions and are bending their feed in a good way.
If someone watches a 30 min YouTube video it’s because they’re already bought in to the idea and just want to stay abreast.
I would rather The Inside View get 20M 25 second views (8.3×10^6 mins) than The Cognitive Revolution get 500k 30 minute views (1.5×10^7 mins) because I think a lot of Cognitive Revolution viewers are already taking action or just enjoy the entertainment, etc.
On TikTok you might be the only AI Safety content someone sees (huge boost from literally 0 awareness for millions of people) while the marginal Cognitive Revolution video might expand the concept to like a thousand new people
I mean more specifically, what is the additional risk of death per person across the next 10 years if you’re lonely vs not lonely? Is it even 1/1000 an affect compared to deaths due to cars?
How tractable are the interventions? It might take hundreds of hours over many months to solve your loneliness. That’s actually pretty hard / costly
I’m not saying it’s not a problem, it definitely is, but I’m just trying to understand if it makes sense to be in this particular guide. Short term serious (yet tractable) risks
Hmm… but wouldn’t the main impact of loneliness be suicide in the short term (the relevant part for this guide)? Which we’re already addressing?
I’m sure loneliness impacts your long-term health, but I don’t think it’s going to raise your likelihood of death in the next 10 years if you’re relatively young and healthy
If there other resources that are doing something similar, please link them in the comments so I can use the information to improve the guide (with a reference). Thanks!
I think this type of thinking and work is useful and important.
It’s very surprising (although in some ways not surprising) that this analysis hasn’t been done before elsewhere.
Have you searched for previous work on analyzing the AI safety space?
If OpenPhil (and others) are in the process of funding billions of dollars of AI safety work and field building, wouldn’t they themselves do some sort of comprehensive analysis or fund / lean on someone else to do it?
What’s the “whole manifund debacle”? People complaining about Curtis Yarvin or something?
Btw, two small suggestions for the chatbot:
smaller max width for container div, size 16px font, and 150% line height
before : https://i.imgur.com/AHjaJHD.png
after: https://i.imgur.com/jAO4ozG.png
Ask the LLM to use standard markdown in it’s output. This will automatically create headings and bolded elements that make it much easier to skim/read
I like this idea and it looks great!
I had a similar concept in mind that I wanted to build but with more of a questionnaire/survey design rather than solely text articles or an open-ended chatbot. More of a hand-holding guided experience through the concerns/debate points.
How’s it going so far? How many daily active users do you have?
Maybe they’re training “GPT-4.5”, maybe they’ve come up with a new name and they’re training “Assistant-1″
But he’s said elsewhere publicly that they’re not training GPT-5
Maybe they’re going to focus on plugins, fine-tuning, visual processing, etc.
Do you not trust Ilya when he says they have plenty more data?
oh good call out, I’ll ping Niki to make sure he sees this comment