Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Comment author: Gurkenglas 24 September 2017 05:34:48AM 2 points [-]

No, really, why reset the password when you can compose the new hash function with the old one?

Comment author: Gurkenglas 24 September 2017 05:23:00AM *  2 points [-]

A nonuniform distribution has two arguments with different probabilities. Changing both probabilities to their average increases the entropy.

Comment author: ImmortalRationalist 17 September 2017 03:36:35AM 0 points [-]

What is the probability that induction works?

Comment author: Gurkenglas 18 September 2017 01:14:41PM *  0 points [-]

By Solomonoff induction, the hypothesis that governs the universe under the assumption that induction works has less complexity penalty than one that counts to a number on the order of 10^80 to 10^18000 steps while the universe is running and then starts working differently by a factor of about 10^17 (since that's how many turing machines with 6 states there are, which is the number of states you need to count to that sort of number of steps), so the probability that induction works can be given an upper bound of about 1-10^-17.

Comment author: Gurkenglas 17 September 2017 12:25:46AM 1 point [-]

If you changed the hashing function so it's harder to invert, you could just hash the old hashes with your new hashing function so people don't need to make new passwords.

Comment author: Thomas 29 August 2017 06:56:59AM 0 points [-]

Sure. But not only to 7/16 but to the infinite number of other values, too. You just have to play with it longer.

The question now is, can the coin make it better, too? If not, why it can only make it worse?

Comment author: Gurkenglas 29 August 2017 07:12:01PM *  0 points [-]

If you say two numbers with nonzero probability, you can improve your chances by shifting all the probability mass to one of them.

In response to A Decision Problem
Comment author: Gurkenglas 25 August 2017 11:58:31PM *  0 points [-]

I agree with Dagon's first paragraph.

Could Omega's decision of which game to play depend on the algorithm I submit as an answer? One convenient ruling might be that if Omega tried to predict whether I would accept, that would count as the simulation accepting/rejecting and it would have to pay out >=100$.

One approach is worst-case analysis as employed in computer science - assume that Omega wants to minimize our reward, then choose the strategy that maximizes it. Here, that means always accepting because that never yields less than 100$.

If I had a random number oracle that Omega couldn't predict, I could accept 10000/10900 of the time, because that always yields an expected reward of 100000/109$, but since Omega can simulate the world this is unlikely.

Some interesting situations may have it be able to predict random number generators within its simulations, but not in the real world...

Comment author: Hafurelus 24 July 2017 10:22:30AM *  4 points [-]

TL;DR - A man is searching for a rival to grow together.

Hello everyone!
This comment was inspired by Anti-Lone Wolf and Stronger Together posts.

What I got from these posts is:
1. Anti-Lone Wolf: Trying to go alone, you may miss an important part of motivation.
2. Stronger Together: There is some kind of evidence that group improvement works.

What I got from my research of doing more is:
1. After a myriad of methods and tips tried during 4 years, I remained on the same level - doing things mostly under Panic Monster pressure. I desperately want to change this.

Having a group of people who have their views almost aligned is good, but what if one could add some sort of competition to this?
My current theory is that if two persons have the same goal and similar strategies of pursuing this goal, their competition can benefit each of them.
For example, I heard that while sprinters train on track, they mostly run at least in pairs while going for a record. (I haven't found written proofs of this). This implies, I suppose, that it's easier to get better results when you compete with someone.
The other example - It is recommended to organize startups not alone, but with a co-founder. I assume that in this case, among other benefits, there is a benefit from support while one of founders is lacking motivation.

Based on this, here I am, searching for a person who has the same goal and is on the same stage of reaching this goal, with similar strategy. My characteristics:
Goal: Survive.
Subgoal: Help in solving AI Safety problems.
Strategy: MS in CS/(perhaps ML) in Germany(winter 2018) -> PhD in ML/(perhaps CS) in USA -> work in AIS industry.
Current state: Finished BS in Applied Math, preparing for an application.
Additional info: 21 year, Russian, male, don't smoke/drink.

I apologize if this is the wrong place for such posts. In this case, it would be nice if one gave information about where it's appropriate to post this.
I also apologize if something from this post is naive or too vague. I'm open to any questions and suggestions.

EDIT (4:30, Jul 25 UTC) - It seems that I was wrong about focusing on competition. In games it's easy to tell who is stronger because usually it's the winner. However, there is no strict rules and definitions about winners and losers in real life. Some minor thing later can play a huge role and nobody knows whether it will happen or not.

So instead of "rivals with a bit of partnership" the focus shifts to "partners with a bit of competition". It seems that this idea is very similar to Stronger Together post, just they have a broader goal.

Comment author: Gurkenglas 25 July 2017 07:26:14PM *  0 points [-]

Sounds like a good idea. I, as well, would like my actions to be ones that follow from that goal and subgoal, though I've lost the last 1-2 years to akrasia. I'm trying to finish a BS in CS in Germany and planning to continue with an MS in CS in Germany. I'm 22 years old, Half-Russian, male, don't smoke/drink. I'm Gurkenglas on Freenode.

Comment author: Thomas 24 July 2017 07:32:02AM 1 point [-]

An easy problem this time?

Comment author: Gurkenglas 24 July 2017 07:56:28AM 0 points [-]

6 queens on A1, C3, G3, C6, G6 and H8 are an upper bound.

Comment author: Thomas 17 July 2017 11:08:11AM 2 points [-]

Try this

Comment author: Gurkenglas 17 July 2017 07:43:26PM *  0 points [-]

The same can be said of unit masses at every whole negative number.

The arrow that points to the right is at the same place that the additional guest in Hilbert's Hotel goes. Such unintuitiveness is life when infinities/singularities such as the diverging forces acting on your points are involved.

Comment author: Thomas 12 July 2017 05:49:09AM 0 points [-]

50 moves rule is totally inappropriate in 4D. Let us dismiss that rule here, yes.

Comment author: Gurkenglas 12 July 2017 10:11:45PM *  0 points [-]

An upper bound is 17 queens: 16 threaten all 6^4 inner squares, then the 17th moves to the inner square closest to the king.

Edit: Nevermind, this amounts to the 17th queen checkmating the king on a 3D board with warp sides.

View more: Next