space_fountain

Research on using LLMs and diffusion models to directly predict weights have been going on for a bit now for example https://arxiv.org/pdf/2402.18153

Spending tens of millions is a lot to see if you could get there too? I realize the internal price is measured in opportunity cost rather than dollars, but still a bit surprising to see

The issue is that when we discovered Pluto we thought it was much bigger than it turned out to be. If it was as big as they thought then sure it would be a planet. As it is it isn’t even the most massive dwarf planet

Even if we just have AI that can do every job that can be remote that would be hugely disruptive. I think you're right that plumbing is safe for a bit. Human hands are insanely dexterous

I'm really sad about how little creative control these tools have. They seem great for creating slop, but pretty useless for creating content that someone would love. I love the potential but text isn't really a great medium for describing artistic vision.

All these demos are focused on how easy it makes everything. Easy is great, but if everyone is able to make instant cute cat videos or whatever it just devalues it. I want to see turning a photo into a rigged 3d model, letting the artist animate and then generate the video. This technology could be used to increase creative expression, but instead it's being used to squeeze out creative expression

We'll see if the admin also restricts access to OpenAI's new models, but if they don't it seems like a policy that is based around perceived fealty to the current admin won't do much to prevent misaligned/or dual function AI from causing problems

I see this really confidently stated, but when I looked into it the exact line is very blurry. It is still possible that writing a prompt is enough creative input for copyright to be granted. What has been decided is you can't have your artwork copyrighted by the AI that produced it, but that isn't quite the same as AI writing not being copyrighted

https://sites.usc.edu/iptls/2025/02/04/ai-copyright-and-the-...

This isn't DRM right? This is metadata attached to the image that makes it clear it was synthetically generated. The public has a huge incentive to know when images are AI generated and the harm to legitimate users seems pretty small: aka someone might complain online that you use AI

I mean we haven't had major power conflict in 50 years and they haven't seemed very useful in Ukraine

Yes, I am interested in understanding why this is hard

Why not? It's not exactly hard? The exact capabilities aren't public, but there are companies that provide daily synthetic radar captures for the whole globe and similarly companies that do this for imagery. It would take ages but I think you and I could both right an algorithm to classify if an image or high map has an aircraft carrier in it? Even if you can't get data for the whole globe taking a photo of a 10 mile of ocean every 10 minutes and shifting the center based on where a boat isn't hard. There aren't that many aircraft carriers. Aircraft carriers and tanks are both capable machines built and designed for a different environment than we have now. They can still be useful. Most countries don't have the ability to image the whole globe or lob weapons across an ocean, but I'm pretty skeptical they'd survive long in a war with say China

Some quick Googling implies China has satellites capable of tracking shipping via radar from geostationary orbit. I'm not really convinced that aircraft carriers can hide these days?

I just mean is the claim that the open source models where the closed models were 12 to 6 months ago true? They do seem to be for some specific tasks which is cool, but they seem even more uneven in skills than the frontier model. They're definitely useful tools, but I'm not sure if they're a match for frontier models from a year ago?

I've seen this claimed, but I'm not sure it's been true for my use cases? I should try a more involved analysis but so far open models seem much less even in their skills. I think this makes sense if a lot of them are built based on distillations of larger models. It seems likely that with task specific fine tuning this is true?

I don't think this person belongs in prison, but the internet also isn't the place it was in 2004? You do bear responsibility for what you do online, and this was irresponsible. We should encourage kids and others to experiment and make mistakes, but the kid shouldn't have put up this website and should have taken it down as a responsible member of the community

I've tried to use langchain. It seemed to force code into their way of doing things and was deeply opinionated about things that didn't matter like prompt templating. Maybe it's improved since then, but I've sort of used people who think langchain is good as a proxy for people who haven't used much ai?

My view is that testosterone is a reasonable thing to discriminate on because:

1. It is causally connected to primary and secondary sex characteristics

2. It has a large impact on performance in many sports

3. It's easy to explain to most people and somewhat matches people's intuitions around fairness

But, yes, it is true that there are cis women with high T levels and it is somewhat unfair and arbitrary to include them when not excluding other random advantages that people have. I'm just not sure if I have a better solution

I said "Sports should only be segregated by this <gender identity> category, except that to remove perverse incentives it’s reasonable to require hrt"

That was trying to elaborate on citruscomputing's argument where they said "Otherwise you have trans women having to choose between being more competitive and receiving necessary medical care."

I'm rephrasing those two points. Apologies if I initially described that badly, but I'm just restating the perverse incentive they were talking about

Nope, I’m consistently saying the same thing. When have I said something else?

Again I don’t take it be saying that. It’s saying that encouraging women to be forced to be in emotional distress to succeed at sport is problematic so we should require hrt so that elite sport doesn’t require trans women to skip hrt

No, but that’s not what the statement is saying. It’s arguing that we should add the minimum restrictions we can to the women’s sports category and that hormones might be a reasonable one

I’m not defending this definition, but I will point out that gender has never been about the chromosomes you were born with. It has been about how people around you perceived you and people often have overly simplistic ideas about exactly what that meant.

Plus it’s totally normal for words to have more technical detail than they first appeared. The idea of a sex binary doesn’t fully exist so we’d need something to deal with that anyway.

I personally support segregation based on hormones as the fairest option available. Otherwise if you use purely a genetic test there are plenty of women with high t levels without an sry gene and no one disputes that high t levels confer a biological advantage in many sports

I think not many people are arguing that we shouldn’t exclude people based on testosterone in elite events, but none of these were trans women, these were all women who lived their entire lives as women from the moment they were born

It’s interesting how the evidence based analysis switched as soon as the republicans came into power. Maybe this is less about evidence and more about opinion actually?

When I’ve researched this it’s turned out that among elite athletes it tended to be a bit higher since some of these intersex conditions can confer benefits

I read the statement as follows:

There is a category called woman, it’s defined by something that’s identify related.

Sports should only be segregated by this category, except that to remove perverse incentives it’s reasonable to require hrt

I’m unclear on what you find absurd about this?

Because then trans men will dominate the "women"'s category. What's frustrating about this entire subject is that many of these things were tried. After finding that too many cis athletes were being disqualified they switched to the current rules that in most cases split things based on testerone levels. You can choose to do it some other way, but all of them come with some problems that people won't like

But trans women aren't in anyway enhanced in their ability to compete?

I'm not asking for a double blind study. I'm asking for examples of someone who took puberty blockers, regretted it and stopped, and then went on to not be able to live the life they wanted to live. I'm not aware of any such stories and I'm pretty familiarly with the population of people who regret taking hormones. When I double checked with Claude it also failed to find anything accept the issue around bone density I mentioned.

There are plenty of studies that point to strong evidence that this protocol results in better mental health outcomes because for whatever potential consequence there is for delaying natural puberty, there are plenty of known irreversible impacts of allowing it to progress.

If you have other evidence, even just observational studies it would be good to share that.

And again the recommendation is to continue until 15 or 16, not until 18

You're making scientific claims, but with the only evidence that I'm aware of contradicting the claim. The usual approach with puberty blockers is prescribing them around the onset of natural puberty and one way or another stopping them around the age of 16. While there are sadly some cases of people who started hormone therapies and later regretted it, I'm aware of no cases of long term health impacts that are attributed to delaying puberty until 16. If you do know of some reports please let me know.

I asked Claude to see if it could find anything and the only reports it could find was some long term bone density issues, but only in trans women and it seemed potentially related to estrogen dosing