Open thread, Nov. 21 - Nov. 27 - 2016

More the former. In particular, there is one user who has been banned repeatedly, has many sockpuppets, and engages in mass-downvoting of people who have in the past expressed opinions contrary to his political position. (It is not clear that he now participates in any way other than by downvoting people he disapproves of, and it's possible that he would rather burn LW to the ground if he could.) Given better tools it might be easy to identify all his sockpuppets, terminate them with extreme prejudice, and stop him coming back; at present, those tools aren't in place. The LW moderators don't have the sort of direct access to the data that would make investigation easy without actually making code changes to the site, and are (I think) not programmer types anyway.

So what needs to happen is for people who have software expertise (but no mod power) to develop what they hope will be suitable tools, without any sort of access to the actual server or data (but there is a procedure, only a few years out of date, for making a sort of LW sandbox on your own computer); for them to provide these to the people who run the site; for the tools to get deployed; for the mods to try to use them to do their job; and for whatever iteration on this process is needed to happen.

(And of course this requires the people with expertise but no power to be sufficiently motivated to do their bit, without any actual way of knowing that the other people involved will actually be willing and able to turn their work into something usable.)

My guess:

The Chinese anti-hero's journey

Someone powerful lives in harmony with a society oppressed by some evil. After failing to defeat the evil s/he sets for a journey which grants him/her the humility / balance to live in harmony with society. S/he usually fails at that to.

I see 2 as a special case of 1. No computational model is inherently superior to another, it is just better for the data / computational power / necessities we have at hand.

I am not a ML Research Scientist -- but have studied and used it, and find it very interesting.

As far as I understand, deep learning is able to discover and fit any nonlinear dynamics in a set of data, which then of course has to be trained/regularized/cross-validated to prune away over-fitting. If we accept the view that reality is just a huge set of nonlinear equations and information, and NN/DL can discover these at any level of granularity, then it is a reasonable prediction that they are well posed to be the best.

Also I'm not confident in this next point, but would love some additional feedback. Read this part with skepticism: As I understand effective DL works so well because it combines filtering and tractability within the model structure, and with variation on layering/neurons/optimization techniques, it opens up a greater set of potential models than many other model classes. This makes the fact that it works so well with GPU not a lucky accident, but rather an intrinsic feature of the mathematical structure of the model. Perhaps that's why we evolved to use NN type structure in our brain -- due to its tractability and parallel information processing abilities?

To use an example from my own research, in Financial Econometric asset pricing we often use this tool called a Kalman Filter to filter out states of the world from sets of stochastic PDEs. Optimizing those models, when they have more than ~10 parameters, is such a hassle. It requires lots of optimization black-magic, which is a quasi-scientific method where over months you run different optimization algorithms on the whole model, then single parameters, then the whole model, and if some parameter looks 'weird' you manually change it to what you 'think' it should be.

Basically a neural network could learn the dynamics here (without revealing them to us), and provide a potentially equal forecast (haven't tested this). It could also do this much faster than our optimization method (I think). The forecast would be less useful to a human analyst, because without the model dynamics made explicit, it is much harder to run simulations and study specific parameters. But it's within the realm of reason to predict that a computer, were it self aware, would be able to understand the way the parameters work in the NN/DL model itself.

For this reason I think the natural structure of the model makes it somewhat true that it is #2, but that this makes it a special case of #1 (as MrMind pointed out before me).

Again, I'm not a ML Research Scientist -- so if I've totally messed something up I'd love to know what and why.

Archaeologists estimate that up to 500 persons were required to extract the heavy pillars from local quarries and move them

I wouldn't take this claim for granted, perhaps I can contact one of these archaeologists, to check how they reached their estimate?

There's not many pixels in this source but it explains a concept for low manpower block moving using simple tools probably available to pre-farming civilizations.

It seems like it should be absolutely devastating to any narrative of history that doesn't involve large, organized civilizations in, at the latest, 9,000 BC.

500 people is not a "large, organized" civilization. The town of Jericho is supposed to have had a population of 2-3,000 people around the same time. Calling it a "civilization" is aIso debatable. I really don't see anything earth-shaking about this.

There seems to be a decent case for this theory

That's evidence for the theory that the Younger Dryas were caused by an extraterrestrial impact, not for the theory of "true civilizations before 10,000 BC".

The idea here is that humanity had started forming true civilizations before 10,000 BC, and a comet impact or airburst over one of the ice sheets caused a huge fireball and flood that led to mass extinctions and the annihilation of civilization

There's no mystery about what caused the quaternary mass extinction--humans reached the Americas and wiped out the ecologically naive megafauna.

Have you heard of Gobekli Tepe?

If I remember correctly, it showed up in Sapiens. One suspicion is that a bunch of tribes used it as their periodic meeting ground, and that it led to the invention of agriculture in the region because of the frequent human visitation.

This seems plausible to me, especially if it's constructed bit by bit over many years. (How they would do the quarrying in the first place seems weirdest in my model.)

back when there were no social media, and even further away in time when there was no Internet at all, we all had one name: our complete name, given to us by our parents / family.

That is incorrect -- there were nicknames, various nom de guerre/plume/..., etc. Besides, most societies have a variety of ways to address an individual depending on the formality of the occasion and the relative status of the speaker: compare "Hey Lizzy" to "Your Majesty" :-D

That might be true, but the fact remains that one person is still completely dependent on and consequently bound to one neural substrate, and the support system for said neural substrate, which permits us (for now) to tie a person's personality to their distinct 'body'.

This remains true even in the spectrum of cases when multiple personalities or possibility even people (appear to) inhabit or share the same neural substrate.

If the substrate is disrupted, the indivdual(s) are destroyed.

This will not always be the case, once technology progresses to such an extent that synthetic and digital persons living in distributed neural substrates become first possible at all and then commonplace in society, the the above statement will be false and a society wide identity crisis will emerge for those not born into the new status quo.

We biological humans, with our antiquated instincts for people in single bodies will be old & regressive compared to the zeitgeist of society.

As Lumfier says naming is often more complicated than that. Scientists who published in Latin for example often didn't simply write their name as they get called in their native language on their manuscript but Latinized it.

People who move to Spain often adopt the Spanish naming convention where the last name is inherited from both mother and father.

I remember (I think from Code Complete) that the assumption that everybody has a name is also false.

Does Tetlock make good arguments? Doesn't he just say that they have a small empirical advantage? An empirical advantage means that he doesn't know what the virtues actually are. Though maybe a small mean advantage corresponds to a large difference at the tails, and all superpredictors are foxes.

Tetlock himself praises hedgehogs: he says that they are necessary for asking good questions. I don't he provides any evidence for this and I'm a little worried that it's just bullshit for the purpose of being inclusive.

There is a long history of decision science praising models and condemning humans who second-guess the models. Is this praise of hedgehogs? They certainly don't use the word "hedgehog" — that's Tetlock's brand. Or does a precise model not count, only verbal models?

When Tetlock talks about foxes and hedgehogs, he sounds to me like a hedgehog, insisting that every new result has vindicated his verbal model that I can't pin down.

Okay, I watched the video and it was impressive. If I start my own cult, I will definitely want to hire your video guy.

My impression is that there are a few good ideas in the core, but they are blown out of proportion ("motte and bailey"). The video is an advertisement, using the standard techniques to create a mountain out of a molehill. It's impressive and technically well-done, but when you are familiar with the pattern, it's just the same pattern again.

Generally, it feels like a LessWrong-inspired Scientology clone. They even talk about registering as a religion.

This is how it probably works in reality:

1) Create an effective spiral around a few keywords such as "logic", by providing a few clever insights, and then insisting that investing as much emotional energy as possible in these few insights will fix everything that is wrong with the world, or at least give you personal superpowers.

2) Subtly redefine "logic" to mean "what our leader says".

3) Profit!