timfduffy.com
I like utilitarianism, consciousness, AI, EA, space, kindness, liberalism, longtermism, progressive rock, economics, and most people. Substack: http://timfduffy.substack.com
2,508 posts
870 followers
453 following
Getting Started
Active Commenter
comment in response to
post
I'm not aware of one
comment in response to
post
Before LLMs folks thought that 1 would be hard, but now it seems pretty easy. 2 seems to be achieved in current LLMs, but I think is probably the hardest, as do many including Yud. 3 is impossible for current models but I expect to be easy for the kind of AIs we worry about.
comment in response to
post
I co-sign this
comment in response to
post
Not a very nice thing to say tbh
comment in response to
post
No, I learned to read quickly when it was taught in kindergarten, but I didn't do so before being trained in it.
comment in response to
post
Wow same actually, sound is the one imagined sense I have. I can play a single melody in my head in a way that sounds semi-real.
comment in response to
post
Sadly I don't know anything more than what's provided by my own pitifully bad introspection.
comment in response to
post
Interesting, I'm also aphantasic and mostly without an internal monologue. I've been wondering how correlated those are, maybe there's some survey data out there.
comment in response to
post
Yeah, that part is tongue-in-cheek.
comment in response to
post
bullying is bad
comment in response to
post
That is how I and many others see it, but I think to some of the folks involved it feels like more than that to them.
comment in response to
post
janus imo
comment in response to
post
Eigenrobot is one one the folks who helped radicalize me against illegibility.
comment in response to
post
Yeah tbh the last couple months have turned me from a "rein in the ERO wing of ICE" guy to an "abolish ICE" guy
comment in response to
post
The plane analogy gets across the intuition that points are all near the edge of the sphere, but I think in high-dimensional space the distribution of distances between randomly selected points converges to root 2 which it doesn't capture.
comment in response to
post
This has nothing to do with your point, but I met @manic-pixie-agi.bsky.social this weekend and she's really cool
comment in response to
post
Such a good use case because comments are so much easier to verify than to write
comment in response to
post
Every LLM ever leaves better code comments than I do
comment in response to
post
I tried to get Gemini to make this for me but I guess it can't do that. Usually it just changed the pfp and this time just gave me an unrelated image and I'm too lazy to edit the image myself
comment in response to
post
It's Norvin' time!
comment in response to
post
Yup that's it sorry for forgetting the link!
comment in response to
post
Your own posts should be auto-liked like they are on Reddit.
comment in response to
post
comment in response to
post
Same, though I didn't bet big in these so it's less than 1k mana profit for me.
comment in response to
post
LiveCodeBench is also using a new version now so not directly comparable
comment in response to
post
Pedro isn't real and the model the output came from was specifically trained to tell users what they wanted to hear. This is an interesting result because it shows potential dangers of sycophancy, but it's being presented very dishonestly here.
comment in response to
post
Not only is the user not real, the model it comes from was RL'ed to tell the user what they wanted to hear.
comment in response to
post
Elon: Have you said thank you once?
comment in response to
post
Super happy to see this, an not just because of the Manifold profit I"m making off it.
comment in response to
post
bsky.app/profile/timf...
comment in response to
post
The new Gemini, or is there another one I'm missing?
comment in response to
post
anniversary of the ratification/effective date of the constitution (June 21, 1788/March 4, 1789) or the ratification of the Bill of Rights (December 15, 1791).
comment in response to
post
This reminds me, the great part of the American Revolution was that we established a democratic government, not just that we were free from British rule. So instead of celebrating the anniversary of the ratification of the Declaration of Independence, I think we should celebrate the...
comment in response to
post
See sweatshirt for answer
comment in response to
post
in the longer term, that seems too for out for me to really predict.
comment in response to
post
Fair point. But I think that we should have built a Moon base after we first visited it, and think we should do the same with Mars. A Mars base would require people to live there for stints of at least a couple years, which would justify structures like this. As for whether we should settle Mars...
comment in response to
post
comment in response to
post
gee emm
comment in response to
post
Some common examples of this that come to mind: socialism, capitalism, eugenics, genocide, racism, fascism. Hard to use these without being more specific.
comment in response to
post
Ah yes, Christmas pudding
comment in response to
post
We don't, I don't even know what that is.
comment in response to
post
Me too. Maybe I should add longtermism to my bio along with utilitarianism.
comment in response to
post
I think it's hard to say too much from self-report, but I'm still happy shat it's self-reported valence is positive