lalaithion

Posts

Sorted by New

Wiki Contributions

Comments

Attempted Gears Analysis of AGI Intervention Discussion With Eliezer

Turing completeness is definitely the wrong metric for determining whether a method is a path to AGI. My learning algorithm of "generate a random Turing machine, test it on the data, and keep it if it does the best job of all the other Turing machines I've generated, repeat" is clearly Turing complete, and will eventually learn any computable process, but it's very inefficient, and we shouldn't expect AGI to be generated using that algorithm anytime in the near future.

Similarly, neural networks with one hidden layer are universal function approximators, and yet modern methods use very deep neural networks with lots of internal structure (convolutions, recurrences) because they learn faster, even though a single hidden layer is enough in theory to achieve the same tasks.

30-ish focusing tips

One warning to those who want to do

Be careful of this if your work requires you to switch between computers, or to work on computers you don't control, or to have others work at your computer. If you have a Dvorak keyboard, none of your coworkers can write a single sentence when you're sitting at your computer trying to figure out something hard. Making your own keyboard shortcuts is less of a problem, unless you rely on them and end up confused every time you go to help a coworker.

I'll also add another meta tip: If you can tell that you're not going to be productive for a while, consider actually taking time off (informally).  You slept poorly and won't be able to work well? Don't go to work and unproductively attempt to do things on no sleep. Sleep in and get working once you're feeling better. Obviously you can only do this if you're working fairly independently or flexibly; but if you're an office worker with a good boss, a student, or a freelancer, this is a good way to avoid burning out.

Dominic Cummings : Regime Change #2: A plea to Silicon Valley

The officer fired a single shot into a crowd of individuals who were breaking down a barricaded door. I think that even the strongest police officer would have a difficult time subduing 20 people peacefully.

The Best Software For Every Need

If you use cut (or awk or sed for cutting), try https://github.com/sstadick/hck

If you use less or cat for source files, try https://github.com/sharkdp/bat

If you can't ever remember the syntax for xargs (sorry, don't have a 2nd other program), try https://www.gnu.org/software/parallel/

If you're using standard command like tools for munging CSVs (like cut, grep, sed, etc.), try https://github.com/BurntSushi/xsv

If you use grep (or ag) try https://github.com/BurntSushi/ripgrep

Each of these programs only improves quality of life a little, but they make doing simple things without leaving the shell so much easier.

Boulder, CO – ACX Meetups Everywhere 2021

Looks to me like it’s in unincorporated Boulder County, just outside the city limits of Lafayette.

Chess and cheap ways to check day to day variance in cognition

I swing between 1100 and 900 in chess ratings really obviously. Some days I'll win and win and win until I'm around 1100; other days I'll fall and fall until I dip into the 800s. A 200 ELO point difference is equivalent to winning the game 76% of the time, so this means that "best day me" would beat "worst day me" in 76% of head-to-head chess matches. 

Is driving worth the risk?

Using numbers from this post: https://www.lesswrong.com/posts/ucjfY46L6qyXefvBT/quick-examination-of-miles-per-micromort-for-us-drivers-with

Assuming you're in a passenger vehicle, without correcting for drunk drivers/unsafe drivers, you have 132 miles per micromort. Using the $10 million value for your life:

I think that means that driving 13.2 miles is equivalent to $1.

If you're willing to categorize yourself in the least risky class in the linked post, that's 54.8 miles for $1.

The other thing to consider is how much you care about other risks to your life. For instance, I've been living with (a limit of) 200 microcovid a week for the past year. I'm young and healthy, so let's say that's 2 micromorts. Given that amount, I should be comfortable driving (a maximum of) 264 miles each week, as they provide the same risk.

How long would you wait to get Moderna/Pfizer vs J&J?

For the sake of argument, let's assume that the effectiveness of j&j and moderna/pfizer are 66% vs 95%, respectively, and this effectiveness comes in at the same time. Suppose you keep yourself to a constant risk of x per week, without factoring in vaccine protection, based on the case rates and variants in your area. If n is the number of weeks between getting the j&j vaccine vs the moderna/pfizer vaccine, and m is the number of weeks between getting the moderna/pfizer vaccine and the end of the pandemic, then you have the risk of (1-(x*.33)^(n+m)) if you get the j&j vaccine, and the risk of (1-x)^n * (1-(x*.04))^m if you get the moderna vaccine.

If I was offered the j&j vaccine tomorrow, vs the moderna vaccine I have scheduled in 3 weeks, and we assume the pandemic will be over in the first world by september 1st (20 weeks out), and my weekly (non-vaccinated risk budget) is 200 microcovids, then I would be looking at 0.132% chance of covid with j&j vs. 0.093% chance of covid with moderna/pfizer. So, it's worth it for me to wait it out. But, if you think the pandemic will be over in half the time, then it's 0.066% vs 0.085%, and so you should take the j&j vaccine.

Theory of Knowledge (rationality outreach)

Theory of Knowledge was perhaps the most useful class I ever took.

Oh, what did I do during it? I sat in the back and read the Sequences.

Load More