sendos

Are AI image and video generators trained with lots of porn images and video? They seem very capable of generating porn (even though they are filtered/moderated not to do so) with realistic physics and movement of nude body parts.

I assume this would not be something the models could have learned without porn being in the training data, is this correct?

I know that AI is improving, but for the sake of argument, if AI makes no progress in the next few years, are current capabilities enough to justify current valuations and investment?

That is, is the economic value that can be unlocked by today’s AI enough to justify the valuations, or are the valuations all predicated on AI becoming much much better than it is today? (AGI, ASI, etc)

I assume it’s a bit of both, just trying to get a sense of the balance between the two.

As I recall, as of a few years ago, that was the case, ie web apps could not do some of the things Flash was able to do.

But maybe web technologies have advanced recently so that that’s no longer the case.

Not sure if people have studied this in detail, but my anecdotal evidence is that when I read a tweet related to politics, the top response is a contrarian one (i.e. disagrees with the tweet)

This is even the case when the comment has fewer responses/likes/retweets than other comments lower down the list.

I can see that the algorithm might be trying to increase engagement, and so it might put the most liked/commented comments on top, but it doesn't seem to do just that.

It seems to also be looking at the 'semantics' of the comment (or maybe the social graph and history of the commenter) and chooses to put the contrarian comment on top.

Maybe a contrarian comment gets the most responses and that's why the algorithm favors it, but I've seen several cases when the top comment was not the most engaged-with.

It's possible this phenomenon is not due to the algorithm but due to bot networks gaming the system and liking the contrarian comment in patterns that make the algorithm put it on top.

Has anyone here noticed this pattern? And what could be the cause of it?

(It certainly makes the discussion on Twitter more polarizing, when the first comment is so often against the tweet even when the comment is not so popular/liked)

Assuming you're a programmer, researcher, or an entrepreneur (which I assume is a good chunk of HN readers), what would you be doing if you had enough money that you didn't have to work for a living?

Would you still be doing what you are currently doing (programming, research, entrepreneurship, other) or would you retreat to some other hobby, such as traveling, painting, etc?

One of the reasons I'm asking is I'm wondering if, in a world where everybody has a guaranteed income, would more technological or scientific progress be made (since people are now free to pursue this without worrying about their mortgage/rent/food) or would less progress be made (since there is no strong profit motive to motivate people to do all the hard work)?

I have set up a website for presidential election predictions (http://prespredict.com), and it is now up and running for the 2016 election. It uses daily state poll data and calculates the probability that each candidate will win. It also estimates the most likely number of electoral votes each candidate will get.

I originally did this back in 2008, and then applied this to the 2012 election. The methodology was also retroactively applied to the 2004 election using available state polls. In all three past elections, the calculation correctly predicted the winner. It also had a pretty good estimate of the number of electoral votes that each candidate ended up getting.

The approach and the website are not as fancy as 538.com but the results so far are pretty good. The methodology is explained here: http://prespredict.com/methodology.html.

It's a bit early and there are several caveats, including the accuracy of polls in the current political environment, the independence assumption in the model, the choice of the winning probability curve, etc. Since the predictions for the last three elections have been pretty accurate, I decided to leave the calculation as is, despite the presence of the caveats.

The website is automatically updated daily at 8am PST, so you can check back periodically.

The new feature that was recently added to HN, which is to be able to toggle the collapse of the comment threads, has the toggle switch at the end of the comment header.

The script below puts it at the beginning of the comment header, just after the up/down vote arrows.

This has been tested in Firefox.

  // ==UserScript==
  // @name        HN: Change location of comment-collapse toggle switch
  // @namespace   HN_sendos
  // @include     https://news.ycombinator.com/item?id=*
  // @version     1
  // @run-at      document-idle
  // @grant       none
  // ==/UserScript==
 
  (function() 
   {
       var allPosts    = document.getElementsByClassName('comhead');
     
       for (var i = 0; i < allPosts.length; i++)
       {
           allPosts[i].innerHTML = allPosts[i].innerHTML.replace(RegExp("^(.+)(<a class=\"togg\" .+</a>)(.*)$","mg"), "$2 $1 $3");
       }
   })();

I see they have a list of papers at https://deepmind.com/publications.html

Which ones are the most acclaimed/well-sited?