It is time to stop quivering in our boots in pointless fear of the future and just roll up our sleeves and build it.
- Ray Pierrehumbert

Wednesday, November 11, 2015

The law of BS and the 8 Error Tweet

As proof of the proposition that
the amount of energy needed to refute bullshit is AT LEAST an order of magnitude greater than the effort required to produce it
which is the fundamental flaw with the "my opinion is every bit as good as yours" school of democracy, I offer this article on Medium, wherein I take 2500 words to refute a single tweet, a tweet which manages to hide at least eight identifiable implicit but substantive errors.

Indeed, although @AZComendador is somewhat obscure and probably not very influential, the misinformation density in that single tweet was so high that I decided to award him the Golden Horseshoe for (correction) 2015.

The author replies on Twitter that he "mostly agrees with me". Weird. Anyway he wants to discuss but doesn't want to take it up on Medium. (I agree that Medium is better for publishing than for discussing.)

But I don't want to take it up on Twitter. Arguing on Twitter is such a useless time sink. It's very hard to craft 140 characters that are immune to willful misinterpretation. Twitter is useful for sharing news and exchanging ideas with people with whom one has a shared understanding. As a medium for debate, it is a horrible time sink.

So if AZComendador wants to discuss matters, he is welcome here.

(Moderation is turned off again, until such time as he-who-shall-not-be-named decides to monopolize the conversation again.)


Mark said...

> Arguing on Twitter is such a useless time sink.

So I hear. One day I might log in to Twitter (or whatever it is you do with Twitter) to verify your observation, but probably not.

Seriously, complaining about Twitter is like complaining about the voting on Celebrity Dancing with the Stars Series 9. Just walk away.

Michael Tobis said...

I really enjoy and appreciate Twitter, until arguments start.

In addition to endless amusement, it is also my replacement both for newspapers and for Google Reader.

William Connolley said...

> the Golden Horseshoe for 2105

Forward looking?

William Connolley said...

> Gavin Schmidt recalls a snippet describing a bar

He may have; but I'm pretty sure he got it from

Which is well worth reading if you haven't already.

William Connolley said...

> he got it from

As he says himself, so I'm absolutely sure. But perhaps you deliberately preferred the known G over the unknown.

Mark said...

> I really enjoy and appreciate Twitter, until arguments start.

OK, well I might give it a go one day. But it sounds like a terrible space in which to conduct substantive discussions, and your post confirms my prejudices. So don't.

Michael Tobis said...

Boojums noted in a new footnote to the Medium article. Thanks.

Kevin O'Neill said...

"@theresphysics @nevaudit Lindzen & Houghton's tour de force QBO paper will apparently be overthrown by @WHUT's physics-free curve-fitting"

I understand the problem with 'cycles' and the fact deniers/skeptics are constantly using them to 'explain' global warming, but this tweet of yours can only be understood as ill-informed. Pukite's QBO letter is not devoid of physics. It uses the lunar gravitational force to explain (model) the 'sloshing' of the oceans. Obviously these same forces explain tides - are our tidal charts also physics free?

From a *physics* POV it should be obvious that rotational and gravitational forces will affect this sloshing and that the moon would have the largest effect. What Pukite has done is take the physics formula that *already exists* to describe sloshing and added the lunar terms. The result is an amazingly accurate predictor of the QBO.

Now, the fact that a machine learning tool identified the major components should not be dismissed, but it was Pukite that figured out what these numbers meant *physically* - i.e., the lunar gravitational forces.

An accurate predictive model of a physical event does not necessarily imply that the physics behind the model is correct - or that there are any physics behind it at all, but in this case the physics behind the model are exactly those physical forces one would believe to have the most effect.

Just as important, any predictive model that is accurate must have either accidentally hit on the correct physics or should provide very strong clues as to what the correct physics are. I understand WHUT's frustration when reputable scientists dismiss out-of-hand his rather strong finding. It's akin to dismissing a physical explanation of Earth's seasons because it relies upon the Earth's orbital 'cycles' to explain them or Milankovitch forcing and paleo climate - just more 'cycles'. Some cycles are significant and not mere curve-fitting.

Michael Tobis said...

This is off topic for this thread and refers to this stuff:

I will say that until Pukite publishes what he did in a way that is replicable I am not going to take it seriously even as number crunching.

Finding a model out of the set of all possible models that replicates a signal doesn't prove anything. That includes finding a model which continues to track a signal for a few cycles even if it wasn't explicitly trained on them, if one gets to look at anything and everything. (it took me a while to understand that.)

Pukite has to explain where his model came from other than (unless he is actually consciously cheating) a search over the space of all possible models. Backforming some handwaving explanation won't work either.

"Just as important, any predictive model that is accurate must have either accidentally hit on the correct physics or should provide very strong clues as to what the correct physics are" Okay. So let him make a notrivial prediction, and we can wait to see what comes out of it. There's nothing urgent about proving this result.

Reducing fluid dynamics to a one-dimensional model is just hard to believe. Whether he's consciously cheating or just overenthusiastic is hard to tell, but the likelihood that this is right (and wave action plays no role) is so small that it is rational for scientists to neglect it. There's no shortage of half-baked theories to waste time on.

If he really has found a key relationship by accident, he can win his case with an actual nontrivial prediction into the future and a bit of patience. What he has isn't science. Pending a sound prediction he hasn't even got a good challenge to science.

Further conversation on this subject should be taken somewhere that people are interested.