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

Comment author: korin43 17 July 2017 05:19:07PM 3 points [-]

As a matter of short term practicality currently we don't have the hardware for GPU acceleration. This limits the things we can do, but at this stage of learning most of the time spent is on understanding and implementing the basic concepts anyway.

For what you're doing, GPU stuff probably doesn't make that big of a difference. Convolutional networks will train and run faster, but a digit recognition network should be tiny and fast anyway.

Comment author: vackosar 11 July 2017 07:21:45PM 2 points [-]

I am new joiner. This is the best post I have seen here so far. Its up there with Moloch post at Slate Star. What is "majority ethnicity" - majority ethnicity of LW? BTW, anybody uses eliptical trainer instead of threadmill?

Comment author: korin43 11 July 2017 11:16:41PM 1 point [-]

In 2016, the "Less Wrong Diaspora" was 86% non-hispanic "white": http://lesswrong.com/lw/nmk/2016_lesswrong_diaspora_survey_analysis_part_one/

Comment author: korin43 28 June 2017 03:13:45AM 0 points [-]

You might like the book "The End of Time" by Julian Barbour. It's about an alternative view of physics where you rearrange all of the equations to not include time. The book describes the result sort of similarly to what you're suggesting, where the system is defined as the relationship between things and the evolution of those relationships and not precise locations and times.

In response to Priors Are Useless
Comment author: korin43 22 June 2017 04:59:46PM 5 points [-]

I think you lost me at the point where you assume it's trivial to gather an infinite amount of evidence for every hypothesis.

Comment author: korin43 04 June 2017 03:18:57PM 2 points [-]

This seems like a good place to ask: How do people read long web based books like this without losing their place? I usually look for ebooks just because my ebook reader will remember what page I was on. I used to use bookmarks for this, but I use 4 different computers on a regular basis (two laptops, a tablet, and a phone). Instapaper / pocket work ok, but then if I add a bunch of links I'll forget about the older ones. Solutions?

Comment author: CronoDAS 23 May 2017 05:06:05PM 0 points [-]

Ah, pilot wave theory. It gets around the "no local realism" theorem by using non-local hidden variables...

Comment author: korin43 23 May 2017 08:07:10PM 0 points [-]

Does it use anything non-local? The experiments in the article use macroscopic fluids, which presumably don't have non-local effects.

Comment author: korin43 23 May 2017 04:42:51PM 1 point [-]

"The experiments involve an oil droplet that bounces along the surface of a liquid. The droplet gently sloshes the liquid with every bounce. At the same time, ripples from past bounces affect its course. The droplet’s interaction with its own ripples, which form what’s known as a pilot wave, causes it to exhibit behaviors previously thought to be peculiar to elementary particles — including behaviors seen as evidence that these particles are spread through space like waves, without any specific location, until they are measured.

Particles at the quantum scale seem to do things that human-scale objects do not do. They can tunnel through barriers, spontaneously arise or annihilate, and occupy discrete energy levels. This new body of research reveals that oil droplets, when guided by pilot waves, also exhibit these quantum-like features."

Comment author: korin43 23 May 2017 04:44:46PM 0 points [-]

Note that the theory seems to have been around since the 1930's, but these experiments are new (2016).

Comment author: korin43 23 May 2017 04:42:51PM 1 point [-]

"The experiments involve an oil droplet that bounces along the surface of a liquid. The droplet gently sloshes the liquid with every bounce. At the same time, ripples from past bounces affect its course. The droplet’s interaction with its own ripples, which form what’s known as a pilot wave, causes it to exhibit behaviors previously thought to be peculiar to elementary particles — including behaviors seen as evidence that these particles are spread through space like waves, without any specific location, until they are measured.

Particles at the quantum scale seem to do things that human-scale objects do not do. They can tunnel through barriers, spontaneously arise or annihilate, and occupy discrete energy levels. This new body of research reveals that oil droplets, when guided by pilot waves, also exhibit these quantum-like features."

[Link] Have We Been Interpreting Quantum Mechanics Wrong This Whole Time?

2 korin43 23 May 2017 04:38PM
Comment author: korin43 13 May 2017 02:43:23PM *  3 points [-]

From the perspective of the God of Evolution, we are the unfriendly AI:

  • We were supposed to be compelled to reproduce, but we figure out that we can get the reward by disabling our reproductive functions and continuing to go through the motions.
  • We were supposed to seek out nutritious food and eat it, but we figured out that we could concentrate the parts that trigger our reward centers and just eat that.

And of course, we're unfriendly to everything else too:

  • Humans fight each other over farmland (= land that can be turned into food which can be turned into humans) all the time
  • We're trying to tile the universe with human colonies and probes. It's true that we're not strictly trying to tile the universe with our DNA, but we are trying to turn it all into human things, and it's not uncommon for people to be sad about the parts of the universe we can never reach and turn into humantronium.
  • We do not love or hate the cow/chicken/pig, but they are made of meat which can be turned into reward center triggers.

As to why we're not exactly like a paperclip maximizer, I suspect one big piece is:

  • We're not able to make direct copies of ourselves or extend our personal power to the extent that we expect AI to be able to, so "being nice" is adaptive because there are a lot of things we can't do alone. We expect that an AI could just make itself bigger or make exact copies that won't have divergent goals, so it won't need this.

View more: Next