Hacker News Reader

Comment by grantlmiller | Hacker News Reader

I'm a natural skeptic, so when everyone believes something I think that is the most important time to question it. I found Economist Bjorn Lomborg's take on climate change to be very impactful... http://www.ted.com/talks/bjorn_lomborg_sets_global_prioritie...

The basic idea is: that given our limited resources how can we do the most good? The answer is that focusing that effort on climate change is frivolous even by the most optimistic predictions. Now that doesn't mean we should ignore it entirely. I'm generally of the mindset that problems aren't solved by "steady progress" rather by radical innovation that changes the game. To that point... it can be seen as a waste of time to enforce rules today that really won't make a difference tomorrow. Instead today's effort should be spent to enable the rapid innovation that when applied will actually make a significant difference tomorrow.

View on HN

> The basic idea is: that given our limited resources how can we do the most good? The answer is that focusing that effort on climate change is frivolous even by the most optimistic predictions.

This is actually a fantastic point. Given the real, quantifiable health effects pollution has on humans (and other species), it seems more beneficial as a whole to deal with more immediate issues. Rather than quibbling over something that might, maybe, possibly reduce global temperatures by a few tenths of a degree. Maybe.

Great video, by the way. Thank you for sharing! Bjorn's basic premise could apply to so many things beyond simply world problems, which is what makes his talk fantastic.

Judging by the emotional investment some people have had in this thread, it'd be shameful if your post is missed. But, reasonable, balanced, rational measures seldom get people quite as excited as those that evoke emotion.

The problem with doing net present value calculations on climate change is the possibility of "infinite" cost events.

Sure it might make more sense from a NPV perspective to spend a given number of dollars on say improving the water treatment in poor countries rather than stopping GHG emissions, but only if we know the maximum cost of climate change. The problem becomes how do you do a NPV calculation on the possibility of a positive feedback warming event that causes the Earth to become uninhabitable. This is effectively an infinite cost event. Even if such a risk is small, no NPV calculation has any value.

...the possibility of "infinite" cost events.

Never having taken place, thinking about such events would seem to challenge to both frequentist and Bayesian schools of thought.

A charitable interpretation of daniel's comment would assume "infinite" meant "very, very, very large", for example: the extinction of humanity. There have been similar events to the extinction of humanity (past mass extinctions), so there is every reason to assign non-zero probability to human extinction from both frequentist and Bayesian perspectives. There are actually a multitude of events that happen in this universe with non-zero probability (from both probability philosophy perspectives) that could eradicate humanity.

Catastrophic climate change may very well be one of them. See http://en.wikipedia.org/wiki/Atmosphere_of_Venus#Evolution

Yes. I really don't know how to do NPV calculations on any event that involves the extinction of humanity, but with any NPV calculation where the cost is more than three orders of magnitude greater than benefit it doesn't really matter how much greater the cost is.

I assume for this reason you also have become religious to avoid a small probability of burning in hell for an infinite amount of time? (See pascals wager.)

These are very different. There is evidence for the possibility of future catastrophic climate change. There is no evidence for the existence of hell.

There are also arguments against Pascal's wager that do not apply to climate change. For example, you can postulate an anti-Christianity Hell, where only non-Christians are spared. This poses a new Pascal wager. There's no reason to think this wager has any different odds than the original: they have the same prior, hence cancel out in an expected value calculation.

>"There are also arguments against Pascal's wager that do not apply to climate change. For example, you can postulate an anti-Christianity Hell, where only non-Christians are spared. "

Could one not also postulate that we are approaching a new ice age, and that carbon emissions are keeping the world from freezing over?

Definitely, although the evidence seems to point to warming as a greater risk than cooling. In addition, there is little evidence for the possibility of a catastrophic cooling capable of making the planet uninhabitable (and not just less ideal), whereas there is evidence that a runaway greenhouse effect could actually make the planet uninhabitable for humans (Venus).

Sure there is some evidence - the existence of a guy claiming to be god's son, a bunch of people claiming miracles occurred, etc. These things are weak evidence in favor of christianity - similarly, there is weak evidence in favor of islam, mormonism, etc.

Given that there is no similar evidence for the Anti-Christianity you postulate, the odds don't cancel.

Another interesting result of this line of thinking: a rational Buddhist/Hindu should convert to Islam/Christianity. If Christianity is correct then you avoid burning in hell forever. If Buddhism is correct you reincarnate and get another shot at enlightenment.

> the existence of a guy claiming to be god's son

Why doesn't such a person show up in Roman records or any one else's records besides Christians then? This is not evidence.

http://rationalwiki.org/wiki/Evidence_for_the_historical_exi...

> These things are weak evidence in favor of christianity

I'm not convinced this is true. Certainly the story of Jesus is evidence, but it's not clear which direction the evidence should push us. Someone making a claim does not always mean our posterior estimate on the claim should increase.

Here's a simple example:

I hand you a small bottle filled with medical-looking pills and a nice label. I ask you, "What's the chance these pills cure malignant melanoma cancer?" You think for a moment, then answer, "Very small. There may be a cure eventually, but I've never seen evidence. I'm not on the cutting edge of research though, so perhaps this is a new breakthrough I haven't heard of. Plausible, but unlikely."

I then say, "Well, let me reassure you by showing you the website I ordered the pills from."

I bring up the website, which shows testimonial after testimonial of people claiming the pills cure malignant melanoma cancer. You then respond (correctly), "I've decreased my estimate. There is still a chance, of course, that the pills work, but I now know these pills are not from a legitimate medical breakthrough I hadn't heard of yet. In fact, this website pattern matches as a scam, so I'm decreasing my posterior to an estimate even lower than my completely uninformed prior."

P(Pills work | testimonials) < P(Pills are a scam | testimonials)

Of course, the testimonials would be positive evidence if the pills had passed clinical trials.

Now, which is bigger:

P(Jesus is the son of God | Jesus said he is the son of God)

P(Jesus is was just a guy | Jesus said he is the son of God) ?

Honestly, this is not clear to me at all. But I think once we start taking in all the other evidence related to (but not necessarily in favor of) Christianity things become more clear. Just like the pill's website pattern matched for a scammy website, it seems to me that the evidence related to Christianity pattern matches to the evidence related to other man made religions. So I conclude:

P(Christianity is man-made | evidence) > P(Christianity is correct | evidence)

Hence P(Christianity is correct | evidence) may very well be less than the uniform prior that a particular god exists.

Edit: The above symbols aren't right, as yummy points out below, but the argument and analogy is sound. What we care about is the change in probability of each possibility. If

P(Christianity is man-made | evidence) > P(Christianity is man-made)

then we can conclude that Christianity is less likely due to the evidence.

The right question: is P(prophet|says prophet) > P(prophet|says ordinary guy)?

My argument is only that any semi-mainstream religion is more likely to be true than some random arbitrary belief system. I.e., Christianity and Islam are both more likely than Anti-Christianity.

I just don't buy your argument that the evidence for all possible religions (including ones you made up for this argument only) cancel out. That's a massive coincidence. And if the evidence only cancels out 99.9999999% rather than 100%, you find yourself back in a Pascal's Wager situation.

Your argument's invalid, because P(prophet) has nothing to do with whether $diety is telling the truth through said prophet or lying. The only way to predict whether $diety is an earnest hypocrite or a master of deception is by pulling a prior out of your posterior.

Computing a posterior, whether about god or controlling climate via human actions, requires a prior. All decisionmaking requires a prior. What's your point?

My point is that your question about who's a prophet is irrelevant. You claim that stories of Jesus is evidence of Christianity -- or that Christians go to heaven -- but it's really evidence of anti-Christianity if you think God would set up honeypot ideologies. You're not doing anything but looping back around to the original question, that if Christianity is real, is it actually a honeypot? The religion's paraphernalia doesn't matter, and the "evidence" you've suggested doesn't affect the chance of it being a honeypot. That it's a question of picking priors is not news to the anti-Christianity anti-Pascal's wager people, that's what they're saying. For example, your act of saying that Jesus is evidence for Christianity and not anti-Christianity is just a roundabout way of saying your priors are anti-honeypot.

Also, the phrase about pulling a prior out of your posterior was meant as a double entendre between your circular reasoning and the well known anti-Bayesian epithet. But it is easily pointed out that controlling climate via human actions requires a much easier prior than computing a posterior about god: its prior is essentially that reality is real and so is Doppler radar. A choice of priors about God, on the other hand, doesn't require throwing away belief in the empirical reality.

Controlling climate via human actions requires far more than simply postulating that Doppler radar works. A huge number of mathematically unproven simplifications are performed when going from Navier-Stokes to GCM models, in addition to a number of known to be false assumptions (e.g., anisotropic viscosity of water). A large number of assumptions are thrown into GCM models for parameters that are unknown.

Admitting the possibility that god (or god-like beings) exists does not require throwing away belief in empirical reality. See, for example, simulation theories.

In any case, my reasoning is merely a continuity argument. TheEzEzz asserted that the very small probability of Christianity and Anti-Christianity, and Islam+Anti-Islam cancel exactly. I'm asserting that if the distributions are continuous then this has probability 0 - it takes a ridiculously tiny amount of evidence or an infinitesimally different prior to put you back into Pascal's wager territory.

> Controlling climate via human actions requires far more than simply postulating that Doppler radar works.

Obviously I'm not saying that it doesn't (well it was obvious to me...). It's a stand-in for everything you'd use. The point is, it's beliefs that stuff you see working will continue working, be it radar or temperature gauges, the laws of physics, or mathematical simplifications or approximations. There is presumably some reason why some seemingly totally made up constants aren't actually 10 times the chosen values, or why it doesn't matter, or how likely that's the case. The difference between this and beliefs about God is that if you don't have reasons why you think the CO2 reading are wrong or why physics will change or why certain mathematical simplifications will stop working, you have to give reasons, you're, epistemologically speaking, discarding empiricism from your toolbox. You can't simply go "Laa dee daa I'm picking different priors than you." There has to be some reason (though it might be that they're going laa dee daa).

> Admitting the possibility that god (or god-like beings) exists does not require throwing away belief in empirical reality.

That's what I said. The difference is that unlike science and measurements of reality, there's no endless stream of evidence to beat your initial priors out of you. Unless you have 100% faith that past events can't be used to predict future outcomes, you'll get dragged over to the empiricist side. But you could have 90% faith in Christianity and maybe that'll only get beaten down to 40%, because there's just no evidence. (Not to mention the more practical way to get people to change their beliefs, no visible consequences for being wrong, either.)

> it takes a ridiculously tiny amount of evidence or an infinitesimally different prior to put you back into Pascal's wager territory.

Sure. But the same goes for putting you into anti-Christianity territory. The expected displacement a tiny amount of evidence would push you is 0 (if we regard the continued absence of new evidence to be evidence itself (e: and assuming potential "big" evidence is neglible, of course)). The whole argument against Pascal's wager is that it's about what your priors are.

And yeah, I don't know what TheEzEzz's 50/50 symmetry is all about.

Reply on news.ycombinator.com