All of Petter's Comments + Replies

Thanks, that makes sense given your assumptions and results.

You think it is likely/possible that a nuclear war kills almost everyone in the US and Europe. Since you estimate the chance of this happening to 1 in 6, are you moving to South America or Australia for a year or two? 

With the right prompt, I get the following results for a few examples I tried (first attempts, no cherry-picking).

Input: ( ) ( ( ) )

Output: Balanced

Input: ( ) ( ( )

Output: Unbalanced

Input: ) (

Output: Unbalanced

Input: ( ) ( ) ( )

Output: Balanced

So it is definitely able to learn balancing a small number of parentheses.

Looks like a solid improvement over what’s being used in the paper. Does it introduce any new optimization difficulties?

The number 0.0000000000000000000000000001 does not tell me much. Numbers in high dimensions are tricky. For example, the volume of a unit sphere decays exponentially with dimension. A unit sphere in dimension 24 has very little volume.

Agreed. The diagram strongly suggests that they do sum to one, so this geometrical method is more confusing than helpful.

Are these misconceptions really common? I thought Kahneman was pretty clear on this in Thinking, Fast and Slow.

It is not a good proxy. Deepmind is a small team and there are many more teams within Alphabet doing machine learning. Remember that the market cap of Goog is $500 billion. (Although if one wants to invest in AI in general I think it is a cheap stock)

I'm convinced that humans must spike their blood sugar and/or pump their body full of stimulants such as caffeine in order to get past the natural tendency to find it unbearably dull to memorize words and syntax by rote and lifeless connection with the structures in their native language.

Just a comment: This is certainly not true for every human. Some people really enjoy that.

I actually think it would work pretty well. The banned user sees all of their contributions and any IP used by the banned user also sees their contributions. All other users and IPs do not see it.

I think it is better if banning decisions are not made public, even (especially) to the banned user.

The banned user should not notice anything, but their posts, messages, and votes do not appear to anyone else.

This exact topic comes up in the discussion you linked to – towards the end under “The difference between Eliezer and Nassim Taleb.” (not a descriptive caption)

Perhaps I will attend future meetups in the Stockholm area, but not today.

Stress and having trouble sleeping causes cancer?

Google will ask you for references before hiring, but will ignore and never read your cover letter.

Thank you for this article.

Many worlds seem not so much an interpretation anymore. They are really there as different non-interacting blobs!

I meant in general. I did not look up the numbers for Wikipedia.

Mobile is a larger platform than desktop 2015. That fact and the knowledge graph seem like very plausible explanations.

Then the entire result of the modification is tautologically true, right?

So, first you have the utility functions that pay both agents 10 if they cooperate and 1 if they don’t.

Then you change the utility functions to pay the agents 0 if they cooperate and 1 if they don’t. Naturally they will then stop cooperating.

I don’t get it. If you are the one specifying the utility functions, then obviously you can make them cooperate or defect, right?

This post had more statements of the type “p < 0.01” than I would expect at LW. I recently read “Frequentist Statistics are Frequently Subjective” here.

80,000 Hours (your employer?) has the following as its web page title:

“How to make a difference with your career”

and writes on the front page

“If you want to make the world a better place…”

To me, those are synonyms to “changing the world,” for the purpose of career description.

Done!

No, everyone who applies to Google is not ulta-smart but most who are hired are probably pretty smart.

Given that everyone who are hired are smart, gwerns point is valid.

Perhaps it’s the other way around? The study only suggests that the time of day affects the decision, not that worse decisions are made when hungry.

Sometimes the environment really is adversarial, though.

Regular expressions as implemented in many programming languages work fine for almost all inputs, but with terrible upper bounds. This is why libraries like re2 are needed if you want to let anyone on the Internet use regular expressions in your search engine.

Another example that comes to mind is that quicksort runs in n·log n expected time if randomly sorted beforehand.