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

Comment author: DanielFilan 08 August 2015 05:52:17AM 0 points [-]

For those who haven't been to my place before, it might be a bit hard to find. My phone number is 0402 447 287, feel free to ring me for help.

Meetup : Canberra: Guilt

1 DanielFilan 27 July 2015 09:39AM

Discussion article for the meetup : Canberra: Guilt

WHEN: 08 August 2015 06:00:00PM (+1000)

WHERE: 70/10 Thynne St, Bruce, ACT

What's guilt good for? How can we use it, or get rid of it if we don't find it useful? Nate Soares, blogging at Minding Our Way, is writing a series of blog posts on this issue. In this meetup, we will discuss the articles starting at Working yourself ragged is not a virtue, until the end of the series, or the last post in the series to be published before the meetup, whichever happens first. We will be discussing these posts over vegan snacks. General meetup info: * If you use Facebook, please join our group. * Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late.

Discussion article for the meetup : Canberra: Guilt

Meetup : Canberra: The Efficient Market Hypothesis

1 DanielFilan 13 July 2015 04:01AM

Discussion article for the meetup : Canberra: The Efficient Market Hypothesis

WHEN: 24 July 2015 06:00:00PM (-0700)

WHERE: 57 McKinlay Place, Narrabundah

Note the new location! Also, date and time for this might be off since I am posting from San Francisco: it should be at 6:00 pm on Friday the 24th of July.

What is the Efficient Market Hypothesis, is it supported by market data, and what are its implications for ordinary people looking to invest their spare cash? Rémy Hamilton-smith will tell us as we eat vegan snacks at his place.

For more detail on the talk: "Since the 80s the growing consensus among academics (though still hotly debated among practitioners) is that the major stock and security markets in developed countries are relatively efficient. This is because huge multinational investment banks can detect and take advantage of the slightest mispricings extremely quickly, making them disappear. They achieve this speed using automated trading programs and have been known to physically move their office next door to the stock exchange in order to get the lowest possible ping.

Thus from the point of view of everyone else this leads to the Efficient Market Hypothesis, which states that it is not possible to predictably ‘beat the market’. In this talk I'll discuss the EMH more formally as well as where it might suggest we should be investing our spare cash."

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late.

Discussion article for the meetup : Canberra: The Efficient Market Hypothesis

Meetup : Canberra: More Zendo!

1 DanielFilan 27 May 2015 01:13PM

Discussion article for the meetup : Canberra: More Zendo!

WHEN: 13 June 2015 06:00:00PM (+1000)

WHERE: 108 North Road, Acton, ACT

I really enjoyed Zendo last time we played it, so we're doing it again, but this time with things other than playing cards. The rules will be explained at the event, but in summary, one person is the 'Master', who has a secret rule in mind, and their 'Students' must guess the Master's rule in order to win. Further explanation is at the Wikipedia page (although note that we will be using simpler rules). As always, vegan snacks will be provided.

Note that after this, I will be in the USA until late July, and therefore unable to run meetups. However, other people are more than welcome to do so in my absence.

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late in the CSIT building, room N101.

Discussion article for the meetup : Canberra: More Zendo!

Meetup : Canberra: Deep Learning

1 DanielFilan 17 May 2015 09:34PM

Discussion article for the meetup : Canberra: Deep Learning

WHEN: 22 May 2015 06:00:00PM (+1000)

WHERE: 108 North Road, Acton, ACT

Deep learning has been making waves: many domains of machine learning have their best results from deep learning approaches. The cynical view is that deep learning is just a new term for neural networks, and the truth is probably somewhere in between, deep learning makes use of new tricks, better hardware and more data to better train neural networks. What is certain is that neural networks have made a comeback. During the talk, I [Mayank Daswani] hope to make you aware of some basic deep learning techniques, and then show you some very cool applications. My knowledge comes primarily from the course on deep learning taught at Stanford this semester (CS224d) that I'm following along (videos, notes, assignments available online). I highly recommend checking it out.

As always, vegan snacks are provided.

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late in the CSIT building, room N101.

Discussion article for the meetup : Canberra: Deep Learning

Meetup : Canberra: Putting Induction Into Practice

1 DanielFilan 28 April 2015 02:40PM

Discussion article for the meetup : Canberra: Putting Induction Into Practice

WHEN: 09 May 2015 06:00:00PM (+1000)

WHERE: 108 North Road, Acton, ACT

At our last meetup, we talked about a mathematical theory of how to do induction well. At this meetup, we will get some practice at actually doing it by playing (a facsimile of) Zendo! In small groups, one player will be the 'Master', and the rest will be their 'Students'. Students will have playing cards, and the Master will have some criterion to pick out 'good' arrangements of playing cards. The 'Students' have to try to guess the Master's criterion by putting out arrangements of playing cards, and asking the Master if they are 'good'. If a Student guesses the Master's criterion, they win! As always, vegan snacks will be provided.

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late in the CSIT building, room N101.

Discussion article for the meetup : Canberra: Putting Induction Into Practice

Meetup : Canberra: Intro to Solomonoff induction

1 DanielFilan 19 April 2015 10:58AM

Discussion article for the meetup : Canberra: Intro to Solomonoff induction

WHEN: 24 April 2015 06:00:00PM (+1000)

WHERE: 108 North Road, Acton

Assume we are walking through the world and see a bunch of objects. Some of these objects are ravens, and all of the ravens turn out to be black. So we start entertaining the hypothesis that 'all ravens are black'. But how can we believe in this hypothesis? It talks about an infinite number of ravens, almost all of which we haven't seen!

What we need is a method of induction, generalizing a finite number of examples into a universal rule. It has been claimed that Solomonoff induction is the best method out there. Is that true? Does that mean all scientists should use Solomonoff induction? How does it work? And what can it do for me?

Jan will explain these and related questions giving a brief tour from probability theory to the universally intelligent agent AIXI. No prior knowledge about math is required. As always, vegan snacks will be provided.

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late in the CSIT building, room N101.

Discussion article for the meetup : Canberra: Intro to Solomonoff induction

Meetup : Canberra: A Sequence Post You Disagreed With + Discussion

1 DanielFilan 06 April 2015 10:38AM

Discussion article for the meetup : Canberra: A Sequence Post You Disagreed With + Discussion

WHEN: 11 April 2015 06:00:00PM (+1000)

WHERE: 108 North Road, Acton, ACT

In February we had a meet up based around us discussing our favorite sequence post. As a counterbalance to that, this time we will be discussing a sequence post that we disagreed with. Each person will summarize the article that they chose and why they disagree with it and that will lead into group discussion.

Please post the article you are bringing on this FB event page so everyone has a chance to read all of them before hand. The disagreement can be major issues or minor nit picks, buy if you can't think of any that you disagree with that might be a bad sign :P

As usual, vegan snacks will be provided.

General meetup info:

  • If you use Facebook, please join our group.
  • Structured meetups are (usually) held on the second Saturday and fourth Friday of each month from 6 pm until late in the CSIT building, room N101.

Discussion article for the meetup : Canberra: A Sequence Post You Disagreed With + Discussion

In response to Optimal Exercise
Comment author: DanielFilan 21 March 2015 02:25:48AM 0 points [-]

Of the forms of exercise I cover, weight training has the most rigorous evidence separating what works and what doesn’t. This study (pdf warning) examines what sort of resistance training results in the most rapid improvements.

Link is now broken, but this one works (assuming that this is the article you were thinking of).

Comment author: G0W51 10 March 2015 01:47:23AM 1 point [-]

Perhaps it would be beneficial to make a game used for probability calibration in which players are asked questions and give answers along with their probability estimate of it being correct. The number of points gained or lost would be a function of the player’s probability estimate such that players would maximize their score by using an unbiased confidence estimate (i.e. they are wrong p proportion of the time when they say they think they are correct with probability p. I don’t know of such a function off hand, but they are used in machine learning, so they should be able to be found easily enough. This might already exist, but if not, it could be something CFAR could use.

Comment author: DanielFilan 10 March 2015 05:19:53AM 3 points [-]

One function that works for this is log scoring: the number of points you get is the log of the probability you place in the correct answer. The general thing to google to find other functions that work for this is "log scoring rules".

At the Australian mega-meetup, we played the standard 2-truths-1-lie icebreaker game, except participants had to give their probability for each statement being the lie, and were given log scores. I can't answer for everybody, but I thought it was quite fun.

View more: Next