Tuesday, July 29, 2014
Monday, July 28, 2014
I remember seeing a group of ants a few days ago, some of them lifting a piece of food, and others walking around those lifting the food. I thought - maybe one of these ants is thinking `I'm not going to be doing this for long. My life's gonna change. I'm gonna start my own company and leave this shitty job' but looking at him from above, it is so clear he is part of a pattern, maybe he will shift by 1/2% in some direction, but that's it.
I'm thinking lately, we need to get out of this inner conflict, thinking we have to be completely different.
Be a super-social family oriented person when we're not even close to that.
Be a super-athlete, doing flips in the air and being an expert in martial arts, when we could never do a cartwheel or decent somersault.
Just accepts our strengths and weaknesses, accept the path we're heading in, and just stir the 1/2% we have control over in the better direction.
I'm thinking lately, we need to get out of this inner conflict, thinking we have to be completely different.
Be a super-social family oriented person when we're not even close to that.
Be a super-athlete, doing flips in the air and being an expert in martial arts, when we could never do a cartwheel or decent somersault.
Just accepts our strengths and weaknesses, accept the path we're heading in, and just stir the 1/2% we have control over in the better direction.
Friday, July 25, 2014
to any fellow academics - on a scale of 1 to 10, how unethical would you rate the following: A fellow co-author retracts a paper that has already been accepted to a conference. He does this without the consent of other coauthors, and without even telling them he did it. They only find out, by accident, after the camera ready version was due.
I never thought academic papers could be like the stock market - you could wake up one morning with less
I never thought academic papers could be like the stock market - you could wake up one morning with less
Friday, July 18, 2014
Low influence functions
It is quite nice to see theory and practice connect.
I've recently become interested in digital crypto-currencies - a hot topic about a market currently worth at least 8 billion dollars, and I would bet, will easily go to a trillion in the next decades.
I started thinking, `ahh, these theory people spending so much time and effort and obscure mathematical questions. I'm never going to do that again..'
But suddenly, this research on crypto-currencies brought me back to old papers that seemed too theoretical to me even when I was interested 100% only in theory.
Let me describe their subject a little through the following story.
Suppose there is a tribe with a council of 99 men that needs to choose
a new leader.
Everybody agrees that in the tribe, there are two members- Xena and Hercules,
that are the most deserving.
They have supernatural powers, and are half-gods.
Both seem to equally deserve the job, so it seems to make sense to just flip a coin
and decide who will be the new leader.
The problem is, who can we trust to flip this coin?
If we assign a particular member of the council, he might use some trick to make, say, Xena come out as the leader - like a coin that seems normal but has one side slightly heavier and almost always falls on that side.
Xena would agree with him in advance that if she gets elected she will do special favors for him
as the tribe leader.
So we don't want to assign the task of choosing the leader to any particular council member.
It seems better to have all council members flip a coin. Then, combine these 99 coin flips in some way.
How should we combine the coin flips?
Here is a very bad idea:
We will count how many coin flips came out heads.
If it is an even number - Xena will be chosen.
If it is an odd number - Hercules will be chosen.
Why is this a bad idea? Because the last person flipping a coin has total control of who will be chosen.
He counts out of the 98 flips so far if there is an even or odd number of heads, and according to that
can say heads or tails to ensure the decision he likes.
Here is a better idea:
We count again how many heads we had.
If it is at least 50 , Xena will be chosen.
If it is less than 50, Hercules will be chosen.
Now notice that when it's the turn of the last person, it is very likely the result has already been determined.
Specifically, only if exactly 49 of the flips so far came out heads, he can control the result.
Are there ways to combine the coin flips such that each player will have an even smaller chance of controling the result?
There are, and this has to do with what is called `low influence functions'.
A nice thing for me was that this paper of ours was mentioned in the Ethereum blog:
https://blog.ethereum.org/2014/07/05/stake/
I've recently become interested in digital crypto-currencies - a hot topic about a market currently worth at least 8 billion dollars, and I would bet, will easily go to a trillion in the next decades.
I started thinking, `ahh, these theory people spending so much time and effort and obscure mathematical questions. I'm never going to do that again..'
But suddenly, this research on crypto-currencies brought me back to old papers that seemed too theoretical to me even when I was interested 100% only in theory.
Let me describe their subject a little through the following story.
Suppose there is a tribe with a council of 99 men that needs to choose
a new leader.
Everybody agrees that in the tribe, there are two members- Xena and Hercules,
that are the most deserving.
They have supernatural powers, and are half-gods.
Both seem to equally deserve the job, so it seems to make sense to just flip a coin
and decide who will be the new leader.
The problem is, who can we trust to flip this coin?
If we assign a particular member of the council, he might use some trick to make, say, Xena come out as the leader - like a coin that seems normal but has one side slightly heavier and almost always falls on that side.
Xena would agree with him in advance that if she gets elected she will do special favors for him
as the tribe leader.
So we don't want to assign the task of choosing the leader to any particular council member.
It seems better to have all council members flip a coin. Then, combine these 99 coin flips in some way.
How should we combine the coin flips?
Here is a very bad idea:
We will count how many coin flips came out heads.
If it is an even number - Xena will be chosen.
If it is an odd number - Hercules will be chosen.
Why is this a bad idea? Because the last person flipping a coin has total control of who will be chosen.
He counts out of the 98 flips so far if there is an even or odd number of heads, and according to that
can say heads or tails to ensure the decision he likes.
Here is a better idea:
We count again how many heads we had.
If it is at least 50 , Xena will be chosen.
If it is less than 50, Hercules will be chosen.
Now notice that when it's the turn of the last person, it is very likely the result has already been determined.
Specifically, only if exactly 49 of the flips so far came out heads, he can control the result.
Are there ways to combine the coin flips such that each player will have an even smaller chance of controling the result?
There are, and this has to do with what is called `low influence functions'.
A nice thing for me was that this paper of ours was mentioned in the Ethereum blog:
https://blog.ethereum.org/2014/07/05/stake/
Tuesday, July 01, 2014
God is telling you,
things are not going to get better..
basically from now, you are just going to get sick and die.
So you might as well,
BE HAPPY NOW!
BE PEACEFUL NOW!
you tell God -
thanks for the advice,
but I think I'm going to focus on getting rich and getting a hot girlfriend..
I'll work on this happiness and peacefulness thing later.
basically from now, you are just going to get sick and die.
So you might as well,
BE HAPPY NOW!
BE PEACEFUL NOW!
you tell God -
thanks for the advice,
but I think I'm going to focus on getting rich and getting a hot girlfriend..
I'll work on this happiness and peacefulness thing later.
Subscribe to:
Posts (Atom)