Trying to understand the world
(my relationship to twitter: https://t.co/7UrZIBBeKS…)
5 subscribed
Apr 12 • 5 tweets • 1 min read
Has their ever been a human culture that trained it's citizens in virtue, sufficiently, such that transparent flattery doesn't work on them?
This thought prompted by the ads in the airport, many of which have barely relevant flattering slogans.
dath ilani would shocked at that. The slogans are trying to hack your decision process and they're not even logically valid!
But has any real human culture done better?
Mar 21 • 17 tweets • 3 min read
I'm monogamous by nature. I've never wanted to be in a romantic relationship with more than one person.
But I've exclusively dated poly people, since the people I wanted to date were poly. And ~all my friends are poly.
So I've been pretty close to some poly relationships.
I'm exploring strategic "mainstream" dating right now. Seeing about meeting people on bumble, and so on.
I've been surprised by how strong the negative reaction to poly is.
Mar 20 • 21 tweets • 3 min read
I just read @nabeelqu's post on how to use twitter.
I don't get it. This is not my experience with Twitter at all.
nabeelqu.co/twitter
I basically already follow his advice, with regards to tweeting.
Mar 19 • 4 tweets • 1 min read
So....it was probably bad that Trump lost in 2020?
Because if he had won, he wouldn't be running again, and neither would Biden (probably).
And it seems likely that the intelligence explosion is going happen in the next presidential term.
If Trump had won, we would at least be seeing different candidates this time around.
And that's not a guarantee that they would be better, than these two, but we at least get a different draw.
Mar 18 • 9 tweets • 2 min read
It's dawning on me how much the incentives of software, and the world, are warped by VC.
The possibility of unicorn returns distorts the whole industry, most of which would probably have more wholesome impacts on the world if they focused on making more revenue than costs.
The whole world is using software, but most (?) of the producers of the software are beholden to something other than their customers: they need to grow, so that they're attractive to the next round of investors, so that they can grow more, so that the founders can exit.
Mar 13 • 31 tweets • 4 min read
Rereading Eliezer's post on wasted motions.
"[Fix] in your mind the scenario where your goal has been achieved, and ask whether a thought will predictably not have contributed to getting there in retrospect."
facebook.com/yudkowsky/post…
This feels like a new insight: this is a two step process. You don't just try and notice wasted motions. You first visualize the world where you succeed, and then, looking back from there, ask if the motions that you're doing are actually helpful.
Mar 11 • 8 tweets • 1 min read
I bet the underlying thing that IQ measures will turn out to be log normally distributed.
It just makes way more sense for it to be a product of (mostly) independent factors than the sum of (mostly) independent factors.
It only looks like a bell curve because we normalize the results of our cognitive ability tests.
Mar 9 • 24 tweets • 4 min read
What's the cutoff point at which I should stop seeking a long term romantic partner, because the singularity is going down and I'm better off just focusing on making it go as well as I can, on my own, instead of investing in a relationship of mutual support?
I think it probably takes 6 months to a year for a relationship to mature to the point that I can really rely on a person in the way that I want and need?
Mar 8 • 7 tweets • 1 min read
So AR and VR has slowly been improving for the past...20 year? Long enough that investors have been hyped about it and other investors are like "no, not this decade."
And so far the later investors have been correct.
But it seems to me that that's about to change? I'm an amateur here, but it looks like we're now like 2 to 5 years away from AR and/or VR being good enough that I use it every day?
Mar 8 • 31 tweets • 5 min read
I can be a lot more concrete here, since I did basically the same thing.
Between 2017 and 2020, I helped run AIRCS workshops, which were a recruiting program for MIRI.
We would invite a bunch of professional programmers to the CFAR venue in Bodega Bay. There were talks on x-risk, some CFAR content, break-out sessions on mathy stuff, and a lot more circling than you would expect.
Mar 7 • 14 tweets • 2 min read
For clarity: There's a much larger group of people who influenced my intellectual trajectory, or who I intellectually respect a lot, but who I'm not affiliated with as strongly.
The first three people on my list above are all people that I've either worked with, or whose projects I've supported, in large part on the basis of my assessment of their judgement.
Mar 7 • 11 tweets • 2 min read
I was reflecting on how badly my society let me down with regards to giving me guidance about dating and finding a partner.
Lots of people are happy to give drive-by advice, without bothering to understand my situation. That's not just unhelpful, it's sometimes actively harmful.
In the same way, my college counselors gave me a bunch of advice, most of which I ignored.
In retrospect, I was right to ignore a lot of it, because my situation and life path really actually was atypical.
Mar 7 • 15 tweets • 2 min read
I don't know if this is a bullshit measurement or not.
But I gotta say, if we were coming up with some bright lines for when to pause frontier AI development, "we have AI as smart as the average human" is an extremely strong contender.
We want to strike a balance between getting as much of the material and human benefits of AI tech as we can, while still leaving a sufficient wide safety margin between our tech and existentially dangerous tech.
Mar 6 • 11 tweets • 2 min read
I'd like to say this publicly when there's nothing in particular happening:
@willmacaskill does not represent me, personally.
He did not contribute to my intellectual trajectory. I've never read any of his books or, to my knowledge, been influenced by any ideas he originated.
If we were voting for leaders or figureheads for EA, and I got a vote, I wouldn't vote for him.
I have nothing in particular against the guy, but he's not someone I defer to or someone whose choices and actions I would like to reflect on me.
Mar 5 • 5 tweets • 1 min read
Here's a proposal: any time you are tempted to point out or complain about a problem on social media, also suggest a probably-shitty, first draft solution.
We absolutely want to be raising problems, and not deluding ourselves about what would work. @ESYudkowsky is always going on about how people dodge the real problems. That's bad.
But we (or at least I) also want energy in the direction of solving problems.
Of course, this is only true to the extent that people actually transact with bitcoin. I don't actually know that the network is so robust that the US government couldn't shut it down, or put it outside the scope of things that "normal" "well-adjusted" people do.
Mar 4 • 7 tweets • 1 min read
I'm trying some supplements for cognitive performance and general brain health My plan is to try them, one at a time, for a week each.
I want to know if anyone has had a strong positive reaction to any of these, and I should start with that one.
For each supplement, I'll add it into my existing stack for one week, and then if it doesn't have negative side effects, I'll take a break for a week and then add it in more regularly.
Mar 4 • 4 tweets • 1 min read
Suppose I had the aim of doubling the number of housing units in the united states over the next 4 years.
Which single person, what organizational role, has the MOST power to make that happen?
If I could be dropped into the position of anyone in the world, who should I choose?
The president of the united states? Secretary of Housing and Urban Development? The senate majority leader? The head of the Senate Housing committee? The governor of California in particular? Jeff Bezos?
Is it some obscure bureaucrat that I've never heard of?
Mar 4 • 4 tweets • 1 min read
We should figure out what language norms would help us in to think skillfully and fluidly in probability.
Stuff we have already:
* Obviously, giving calibrated credences. And betting odds.
• And the words "prior" and "posterior" and "odds ratio".
Feb 26 • 9 tweets • 1 min read
Is there an LLM that is capabilities -competitive with GPT-4 yet?
If there isn’t, does that mean that OpenAI, currently has a technical lead that is lower-bounded at 348 days?
(That’s the time between the GPT-4 release and today.)
How does Gemini beat GPT-4 on benchmarks?
Some quick googling suggest yes, at least for Gemini ultra.
5 subscribed
