Before I give my predictions for this year, a record of how I did on last year’s predictions.
General AI will not be made in 2011. Confidence: 90%
Was correct.
The removal of DADT by the US military will result in fewer than 300 soldiers leaving the military in protest. (Note that this may be hard to measure.) Confidence: 95%.
I haven’t been able to judge this. It looks hard to tell but seems to be correct. However, to a large extent this being correct extended from something I didn’t anticipate- it took much longer to actually implement the repeal than I expected, so the repeal took effect fairly late in the year.
The Riemann Hypothesis will not be proven.
I initially gave this 75% but further discussion suggested I was underconfident and so I bounced this up to 95% and was correct.
Ryan Williams recent bound on ACC circuits of NEXP (See here for a discussion of Williams work) will be tightened in at least one of three ways: The result will be shown to apply for some smaller set of problems than NEXP, the result will be improved for some broader type of circuit than ACC, or the bound on the circuit size ruled out will be improved. Confidence: 60%
Was incorrect.
At least one head pastor of a Protestant megachurch in the US will be found to be engaging in homosexual activity. For purposes of this prediction “megachurch” means a church with regular attendance of 3000 people at Sunday services. Confidence: 70%.
A few such scandals occurred but none of them were in churches nearly large enough. So this was incorrect.
Clashes between North Korea and South Korea will result in fatalities: Confidence 80%.
Was incorrect.
So with this out of the way new predictions. I’m not including here any predictions that have an end date of 2013 which I’ve already put in PredictionBook.
The first four predictions are predictions which are updated versions of predictions from last year:
Clashes between North Korea and South Korea will result in fatalities, or the North Korean government will collapse. Confidence: 75%
The Riemann Hypothesis will not be proven in 2012. Confidence 95%
P != NP will not be resolved in 2013. Confidence 95%
General AI will not be built in 2013. Confidence: 95%
The next set of predictions is about computational complexity:
The relationship between P and BQP will not be resolved in 2012. Confidence: 85%
There will be improvement in efficient matrix multiplication in 2012 that will be discussed in at least one of the following blogs: Combinatorics and more, Shetl-Optimized, Godel’s Lost Letter. Confidence: 52%
In 2012 there will be new results for either the group isomorphism problem or the graph isomorphism problem discussed at at least one of the following blogs: Combinatorics and more, Shetl-Optimized, Godel’s Lost Letter. Confidence: 52%.
No improvement in factoring integers in a classical setting that has better time-asymptotics than the best current ones will be made in 2012. Confidence: 85%
The next two predictions have to do with the integer complexity problem. (Background on that can be found here.)
No one one will resolve in 2012 whether integer complexity function is asymptotic to 3log_3 n. 58%
I will coauthor at least one paper on integer complexity by 2013. Confidence: 61%
The next few predictions concern space travel and exploration.
Humans will continue to have at least one functioning probe on Martian surface or a satellite around Mars for all of 2012. Confidence: 98%
Russia will lose at least one rocket launch in 2012 . Confidence: 55%
No contact with intelligent aliens be made in 2012: Confidence 96%
The Hubble Telescope will continue functioning though 2012. Confidence: 85%
Miscellaneous predictions:
HPMOR will update at least twice in 2012. Confidence: 75%
In 2012, I will go to at least two open viewing nights for telescopes in the greater Boston area. Confidence: 58%
I will not become a vegetarian in 2012. Confidence: 80%
My mother will not read any books by Steven Pinker this year. Confidence: 85%
I will have a total LessWrong karma of at least 10,000 by the end of the year. Confidence: 75%
Meta-predictions:
At least one of the above predictions will turn out to be correct for reasons that are surprising to me. Confidence: 80%
There’s a fair bit of reason to think that neither of BQP and NP contains the other. But the primary cause for my reduced confidence is that I don’t have a really good understanding of the quantum complexity classes whereas I do have more intuition for the classical classes like P and NP. So I’ve reduced the confidence accordingly.
If you feel you are relatively ignorant of quantum complexity and want to reduce your reliance on it, you should not simply reduce the number. That anchors on an arbitrary sign of the question. Why reduce 95% rather than the complementary 5%? Your prediction is, roughly, that P vs BQP will be resolved in 6 years. Phrased that way, isn’t it overconfident?
Instead you should regress to an outside model. For example, it has been 30 years since Feynman’s suggestion, so my outside model is that it won’t be resolved in 30 years, so < 3% per year. Edit: this is a doomsday argument.
Also, if you think your inside model says that something is hard, but the number it yields is easier than the outside model, you probably aren’t combining your information correctly.
Before I give my predictions for this year, a record of how I did on last year’s predictions.
Was correct.
I haven’t been able to judge this. It looks hard to tell but seems to be correct. However, to a large extent this being correct extended from something I didn’t anticipate- it took much longer to actually implement the repeal than I expected, so the repeal took effect fairly late in the year.
I initially gave this 75% but further discussion suggested I was underconfident and so I bounced this up to 95% and was correct.
Was incorrect.
A few such scandals occurred but none of them were in churches nearly large enough. So this was incorrect.
Was incorrect.
So with this out of the way new predictions. I’m not including here any predictions that have an end date of 2013 which I’ve already put in PredictionBook.
The first four predictions are predictions which are updated versions of predictions from last year:
Clashes between North Korea and South Korea will result in fatalities, or the North Korean government will collapse. Confidence: 75%
The Riemann Hypothesis will not be proven in 2012. Confidence 95%
P != NP will not be resolved in 2013. Confidence 95%
General AI will not be built in 2013. Confidence: 95%
The next set of predictions is about computational complexity:
The relationship between P and BQP will not be resolved in 2012. Confidence: 85%
There will be improvement in efficient matrix multiplication in 2012 that will be discussed in at least one of the following blogs: Combinatorics and more, Shetl-Optimized, Godel’s Lost Letter. Confidence: 52%
In 2012 there will be new results for either the group isomorphism problem or the graph isomorphism problem discussed at at least one of the following blogs: Combinatorics and more, Shetl-Optimized, Godel’s Lost Letter. Confidence: 52%.
No improvement in factoring integers in a classical setting that has better time-asymptotics than the best current ones will be made in 2012. Confidence: 85%
The next two predictions have to do with the integer complexity problem. (Background on that can be found here.)
No one one will resolve in 2012 whether integer complexity function is asymptotic to 3log_3 n. 58%
I will coauthor at least one paper on integer complexity by 2013. Confidence: 61%
The next few predictions concern space travel and exploration.
Humans will continue to have at least one functioning probe on Martian surface or a satellite around Mars for all of 2012. Confidence: 98%
Russia will lose at least one rocket launch in 2012 . Confidence: 55%
No contact with intelligent aliens be made in 2012: Confidence 96%
The Hubble Telescope will continue functioning though 2012. Confidence: 85%
Miscellaneous predictions:
HPMOR will update at least twice in 2012. Confidence: 75%
In 2012, I will go to at least two open viewing nights for telescopes in the greater Boston area. Confidence: 58%
I will not become a vegetarian in 2012. Confidence: 80%
My mother will not read any books by Steven Pinker this year. Confidence: 85%
I will have a total LessWrong karma of at least 10,000 by the end of the year. Confidence: 75%
Meta-predictions:
At least one of the above predictions will turn out to be correct for reasons that are surprising to me. Confidence: 80%
I find this confusing; I would expect P vs. BQP to be harder to resolve than P vs. NP.
There’s a fair bit of reason to think that neither of BQP and NP contains the other. But the primary cause for my reduced confidence is that I don’t have a really good understanding of the quantum complexity classes whereas I do have more intuition for the classical classes like P and NP. So I’ve reduced the confidence accordingly.
If you feel you are relatively ignorant of quantum complexity and want to reduce your reliance on it, you should not simply reduce the number. That anchors on an arbitrary sign of the question. Why reduce 95% rather than the complementary 5%? Your prediction is, roughly, that P vs BQP will be resolved in 6 years. Phrased that way, isn’t it overconfident?
Instead you should regress to an outside model. For example, it has been 30 years since Feynman’s suggestion, so my outside model is that it won’t be resolved in 30 years, so < 3% per year. Edit: this is a doomsday argument.
Also, if you think your inside model says that something is hard, but the number it yields is easier than the outside model, you probably aren’t combining your information correctly.
Maybe because that pushes the probabilities towards the zero-knowledge position of 50%. (However, as you say, this isn’t a zero-knowledge situation.)