All of NoUsernameSelected's Comments + Replies

Seems like a lot of paragraphs got collapsed together in this version of the post (vs the Wordpress and Substack ones)?

2habryka
Indeed. I fixed it. Let's see whether it repeats itself (we got kind of malformed HTML from the RSS feed).

I don't get a progress bar on mobile (unless I'm missing it somehow), and the word count on hover feature seemingly broke on mobile as well a while ago (I remember it working before).

2habryka
Ah, I think showing something on mobile is actually a good idea. I forgot that the way we rearranged things that also went away. I will experiment with some ways of adding that information back in tomorrow.

Why remove "x min read"? Even if it's not gonna be super accurate between different people's reading speeds, I still found it very helpful to decide at a glance how long a post is (e.g. whether to read it on the spot or bookmark it for later).

Showing the word count would also suffice.

2habryka
Mostly because there is a prior against any UI element adding complexity.  In this case, with the new ToC progress bar which is now always visible, you can quickly glance the length of the post by checking the length of the progress bar relative to the viewport indicator. It's an indirect inference, but I've gotten used to it pretty quickly. You can also still see the word count on hover-over.

I compared the Manifold forecasts with the community prediction on Metaculus and calculated a time-averaged Brier Score to score forecasts over time.

2the gears to ascension
This is where I got the idea it was a median. Looks a little more complex than I remembered. Very interesting. https://www.metaculus.com/help/faq/#community-prediction

The so-called "nonsense" community prediction is still more accurate on average than Manifold for the same questions.

https://www.metaculus.com/notebooks/15359/predictive-performance-on-metaculus-vs-manifold-markets/

2the gears to ascension
wow, what! I really didn't see that coming. it ... is in fact the median of the community predictions, though, right?

+1

Never really got into VRChat, but I'd be happy to try a LW meetup there.

I agree about it having to fit on a single chip, but surely the neural net on-board would only have a relatively negligible impact on range compared to how much the electric motor consumes in motion?

8Daniel Kokotajlo
IIRC in one of Tesla's talks (I forget which one) they said that energy consumption of the chip was a constraint because they didn't want it to reduce the range of the car. A quick google seems to confirm this. 100W is the limit they say: FSD Chip - Tesla - WikiChip  IDK anything about engineering, but napkin math based on googling: FSD chip consumes 36 watts currently. Over the course of 10 hours that's 0.36 kWh. Tesla model 3 battery can fit 55kwh total, and takes about ten hours of driving to use all that up (assuming you average 30mph?) So that seems to mean that FSD chip currently uses about two-thirds of one percent of the total range of the vehicle. So if they 10x'd it, in addition to adding thousands of dollars of upfront cost due to the chips being bigger / using more chips, there would be a 6% range reduction. And then if they 10x'd it again the car would be crippled. This napkin math could be totally confused tbc.

I'd provide a counterexample analogy: speedruns.

Many high-level speedruns (and especially TAS runs) often look like some combination of completely stupid/insane/incomprehensible to casual players. Nevertheless, they work for the task they set out to do far more effectively than trying to beat the game quickly with "casual strats" would get you.

I think seeing a sufficiently smart AI doing stuff in the real world would converge to looking a lot like that from our POV.

Counterpoint while working within the metaphor: early speedruns usually look like exceptional runs of the game played casually, with a few impressive/technical/insane moves thrown in.

9Shankar Sivarajan
Counterpoint: such strategies typically requires lots of iteration with perfect emulation of the system targeted to develop (I'm thinking in particular of glitch exploitation). Robust strategies might appear more "elegant."

Any particular reason you've linked all those tweets, but blocked general access to them? I'd probably be interested in reading some of those threads just going by the titles.

3DragonGod
Oh, it's not clear but this thread was originally written over 6(?) months ago (my Twitter was public back then). I just updated it today to add the top line. I have this thread linked in my Twitter bio (and I guess that's where most visitors to it come from), so that's the main use case. I don't really blog outside LW, so don't have a separate home page.   Sorry that you can't see the linked tweets but: 1. I'm not actively maintaining this thread/haven't updated it for several months before a sudden desire to make it more prominent than I'm focusing on AI safety 2. This thread's modal reader is someone who saw me on Twitter not someone finding this blog post by browsing LW blog posts   I do not plan to make my Twitter account public (I accept follow requests by default, but there are many bad incentives of public Twitter that I enjoy being shielded from).

Hey, I'm an interested Lithuanian :)

Only saw your comment just now 2 months later, but I've sent you a DM.

1Mnephisto
Great - sent a response :) Good to hear I'm not the only one and sounds like a good start to the community

Is there any GDP-like measure that does do a better job of capturing growth from major tech breakthroughs?

4Lost Futures
Total Factor Productivity would fit the bill

Advice for Smart People with Autism

Even if you never write this one (or any of the other ideas with "autis" in the title), I'd love a 1-3 sentence summary of your suggestions.

How long would you expect to be able to enjoy your newfound fortune in Google stock before death? I could maybe see an AGI starting to disassemble the planet before the stock even has a chance to rise all that much...

4Razied
In the modal future, no time at all, we probably all die at the same time before any economic effects are felt. That was partly a joke, and partly said to maximize value in those futures where we do get some time of economic growth before catastrophe.

In the limit, the pessimists will eventually be correct. There's only so many technologies and improvements that can be made under the laws of physics (e.g. Landauer's limit), and once we hit those real limits, that'll be as good as it gets.

Of course, the catch that a lot of people might not realize is that a) we're at least centuries away from hitting said limits, even with aligned superintelligence, and b) those limits are far, far better than what we have and live with today.

Answer by NoUsernameSelected60

Maybe not a normal blog post per se, but AI Impacts did a survey back in 2016 about when AIs would reach certain capabilities. Some narrow tasks have already passed their 10 or 50% median estimate dates and are arguably doable with 2022 technology.

They are: aesthetics, mental health, social capital, wealth, physical attractiveness, and niceness.

So if I'm below average on probably all of these (and extremely below average on at least mental health and social capital), is the right thing to do just to give up on dating for the foreseeable future?

5Xodarap
Here is a table of who women message on a dating app, organized by attractiveness quintile: From Kreager, Derek A.; Cavanagh, Shannon E.; Yen, John; Yu, Mo (2014). “Where Have All the Good Men Gone?” Gendered Interactions in Online Dating. Journal of Marriage and Family, 76(2), 387–410. doi:10.1111/jomf.12072  Men in the bottom quintile clearly get substantially fewer messages, but they don't get zero messages. 3% of messages from top quintile women go to bottom quintile men!   I don't feel qualified to give advice about your personal life though. Whether or how you should be dating depends on a bunch of factors like what your goals are, how costly it is for you to date, etc.
6omegastick
While it's important to bear in mind the possibility that you're not as below average as you think, I don't know your case so I will assume you're correct in your assessment. Perhaps give up on online dating. "Offline" dating is significantly more forgiving than online.

I wonder how plausible it would be to develop some kind of tool that offers alterations to any message you leave online into something that has essentially the same meaning and content, but no longer possesses your "digital fingerprint".

Change the wording into something you're less likely to use, add or remove subtle details like, say, a double space after a period, randomized posting times, etc.

7orthonormal
Obfuscation might be feasible, yeah. Though unless you can take down / modify the Wayback Machine and all other mirrors, you're still accountable retroactively.
Answer by NoUsernameSelected50

Metaculus predicts with ~75% probability that UBI will be implemented in at least one EU country.

Are there any good examples of useful or interesting sub-problems in AI Alignment that can actually be considered "solved"?

2paulfchristiano
I don't think so. Of course the same is true for machine learning, though it's less surprising there. I think subproblems getting solved is only something you'd really expect on a perspective like mine where you are looking for some kind of cleaner / more discrete notion of "solution." On that perspective maybe you'd count the special case "AIs are weak relative to humans, individual decisions are low-stakes" as being solved? (Though even then not quite.)