The view of academic experts in an area often differs considerably from the impression given in the popular media, such as in newspapers and magazines. But each academic knows only a few areas as an academic; most of the rest he knows mainly as a consumer of popular media. Unfortunately, academics typically accept media impressions of academic consensus in areas outside their own.
Reviewing the items, I think I mostly agree with Robin. I have different views on uploads, but wouldn't but the chance of anyone becoming an upload after centuries as low as Cowen's 1%. Rather, my view is that uploads are likely to come last - and have correspondingly low economic impact. p(success) = one in ten thousand also seems over-confident for Cowen's cryonics prediction to me. Rah, the belief-value dichotomy!
I have a practical question: Just $200 per year for Alcor? Even with term life insurance, an $80K (charge for head only) policy is over $500/year for someone with Robin's published stats; and that doesn't even protect against senescence like a more expensive whole life policy. Add Alcor's yearly dues of $600 or $1200, and that $200 figure starts to look an order of magnitude too low.
Matthew, thanks for the wiki link. But, if convenient, could you please provide some direct links (or summary of results) to these negative polls? The wikipedia article reports the following:
"A 2005 minor poll on the Interpretation of Quantum Mechanics workshop at the Institute for Quantum Computing University of Waterloo produced contrary results, with the MWI as the least favored."
But this poll consisted of a sample of 4 (four) researchers.
I'm also curious whether there is anybody here who subscribe to MWI "without the slighest degree of doubt about its correctness"?
And how can you be sure that the vast majority of participants on this blog subscribe to the MWI? My own probabily would be closer to the mid-way mark than to either 0 or 1, and I have no idea what most of the other people here think about the MWI.
Interesting. The high-water polls for MWI have been mentioned, but not any of the contrary polls (all listed in the Wikipedia article on MWI). I would suggest that is not very surprising on a blog where the vast majority of the partipants subscribe to the MWI because it fits best with their beliefs about the nature of reality, some of them without the slightest degree of doubt in its correctness.
However when the title and raison d'être of the blog is "overcoming bias", I think we ought to raise the expectations up a notch. Go ahead and mention the contrary polling data -- it will enhance your truth orientation (and credibility to boot!)
I think that physics today is in about the condition that physics was a hundred years ago. There is a well establshed model that cannot explain certain aspects of reality. It is unclear what to do about this. The only thing that is clear is that something new will probably be needed.
To assign high likelyhood (to "believe") in any of several deeper explanations is hardly prudent. In fact the most important thing may be to be contrary :-)
I think the most interesting question is on how important 'overcoming bias' is. I would rate it as fairly important, but certainly not more important than overcoming laziness or some such. And I suspect that a well governed bias can be a useful time-saving technique that has more positive effects than negative. So I value people like Robin who [mistakenly] think that overcoming bias is the preeminent value because their endeavours may reveal things which when integrated into my own biases are very useful.
But the effort to reward ratio in overcoming bias is like many things--subject to marginal returns.
Am I odd for thinking that blogs are an odd place to work out this meta-disagreement when physical separation is not an issue? If there are going to be rounds of back-and-forth about what exactly x means, lunch seems like a good time to pin that down; text report-outs can follow. Perhaps there is some value in showing the back-and-forth to the public or in fixing words in text, but I imagine you two will see each other some time in the next week or two.
On the MW, one poll is reported here: http://www.anthropic-princi.... It is probably a couple of decades old, but my impression is that MW has continued to have the wind in its sails, so I don't think support has declined since then:
"Political scientist" L David Raub reports a poll of 72 of the "leadingcosmologists and other quantum field theorists" about the "Many-WorldsInterpretation" and gives the following response breakdown [T].
1) "Yes, I think MWI is true" 58%2) "No, I don't accept MWI" 18%3) "Maybe it's true but I'm not yet convinced" 13%4) "I have no opinion one way or the other" 11%
If Robin, based on his own investigations, assigns a probability just slightly higher than the physicist consensus, I don't think this would count as a significant disagreement.
On "Probability of uploads within centuries < 1%" and "Probability of 'If your head is cryogenically frozen today, you will be alive in 2100.' < 0.01%", I have to say that these seem astonishingly low. Even if we just use arguments from authority, and even if we discount the most-informed authorities as being self-selected for optimism about these prospects, it is hard to see how we could get estimates that are that low. Would Tyler claim that if we assembled historical data on cases in which a better-informed scientific minority held a view that conflicted with a less-informed scientific majority about what technological feats might be possible a century later, we would find that the minority had been right only in 1/100 to 1/10,000 of the cases?
I do think that if we are going to improve the way we deal with disagreements, one important first step is to try to be very careful about how we specify the contested claim. One of the most significant methodological advances in philosophy over the past century or so has been to appreciate the importance of precision of language and care in how we articulate our theses. This methodological insight could probably be more widely applied. Taking a vague disagreement, and, through analysis and distinctions, transforming it into a (set of) precise disagreement(s) can be a major research contribution if done well.
Robin's choice of experts on quantum computing as an estimator of p is a good one -- such people are much more likely to have thought about the foundations of qm than average physicists.
I posted the list below on the other thread, but it may be of use here. I have been told that Feynman and Gell-Mann would ofted *aggressively* proselytize about many worlds to visitors at Caltech.
Some eminent physicists who (as far as I can tell) believe(d) in MW: Feynman, Gell-Mann, Hawking, Steve Weinberg, Bryce DeWitt, David Deutsch, Sidney Coleman ... In fact, I was told that Feynman and Gell-Mann each claim(ed) to have independently invented MW, without any knowledge of Everett!http://infoproc.blogspot.co...
So the main evidence that I am a "logical atomist" is that I want to distinguish the possibility that our differing values lead to our differing priorities on overcoming bias, from the possibility that our differing priorities result from some as yet unstated factual belief related to overcoming bias? Is my interest in the particular distinction I am considering an especially potent evidence of "logical atomism", or is the inclination to draw any distinction whatsoever equally suggestive evidence of such an alarming tendency?
Robin is free to tell me that I don't disagree with him on the many worlds interpretation of quantum mechanics. I don't know much about what the scientific consensus is, but I do believe that such a consensus is my best estimator, given my limited understanding. Robin talks "as if" he has a higher p than the consensus, but of course he is free to announce that he doesn't.
Robin juxtaposed these two sentences:
"Nerd, I haven't accepted Tyler's label of "logical atomism", and I'm still not even very clear on what it means.
Quantum, it is not clear that Tyler and I disagree about any fact regarding overcoming bias; it could just be a preference difference."
Again, this is the heart of the matter. And the second sentence shows just how strong Robin's logical atomism is. It reflects Robin's reductionism to call it "just a preference difference." Nerd remains right on target.
Matt, I don't think I have made any claims about consciousness, just on reproducing I/O patterns. I agree with you that AI advocates have been too optimistic about rates of progress, though as I was an AI researcher for nine years, I'm pretty sure few AI researchers ever endorsed 5-7 year forecasts. And I consider myself to be a singularity skeptic in many ways.
My claim is not that "informed scientists expect us all to become uploads" but that those most informed specifically about both brain modeling and emulation expect, with at least "p=0.1", whole brain emulation to be technically feasible within a few centuries at most.
I would not concede that "brain modeling" scientists are experts on consciousness. In fact, I suspect that the people I would consider experts on consciousness and your list of such experts would have very little overlap.
In fact, a great many of my differences with you on your list of fourteen wild ideas (I agree with some of them) have to do with differences in whose opinion should be worth listening to on any particular subject.
You want to defer to an "expert" opinion, and I feel that your lists of "experts" are so subject to selectionist and other effects that their opinion is very biased and questionable.
Just as a single example, I would submit that the fifty years of hearing "AI in 5-7 years" should produce more skepticism about claims of strong AI in the future. I most certainly would not trust the predictions of people working in AI over the predictions of well-informed outsiders.
We could say the same thing about many other putative experts in very speculative fields, with vested financial, career and emotional interests in rapid progress or at least perceptions of rapid progress.
Frankly I feel bad making these kinds of comments on a board full of true believers in these kinds of Singularian ideas. I have no particular desire to rain on anyone's parade, and some of the Singularian activities are likely to produce modest or substantial fruit (even though not the everlasting life hoped for by their proponents). The only reason I bother to speak up here is the putative purpose of this forum to "overcome bias", the lack of many (any?) Singularity-skeptic voices here over an extended period of time, and a genuine wish to support Robin Hanson's heartfelt continuously stated desire to discover truth. I do believe he is very sincere in that quest, and I do feel it would be a worthwhile quest for many other people who contribute here to take up as their own, which would require less epistemic arrogance than some of the posts here demonstrate.
You certainly won't overcome bias and discover truth with a room full of true believers. But the temptation is ever-present for me to move on and let the millenarian parade-planners have their fun, and sooner or later I am likely to succumb to it.
Tyler, I don't yet accept this "logical atomist" label you have put on me, and even after reading a bit I'm not entirely clear on what it means. I certainly don't think the label applies to me merely because I objected to having my views called "strange" and "crazy" on the basis of some other vague labels you invoked, and therefore ask you to be as clear as possible about where you think we disagree.
My claim is not that "informed scientists expect us all to become uploads" but that those most informed specifically about both brain modeling and emulation expect, with at least "p=0.1", whole brain emulation to be technically feasible within a few centuries at most. Are you suggesting scientists with this specific expertize are biased because their topic is "unusual"?
As per the title of this post, the contrarian point, if any, (and not my contribution btw) is to combine this technical conclusion with social analysis to predict social consequences. My contribution, if any, is to apply standard formal economic analysis to this situation.
Selection bias -- which experts work on certain unusual topics -- may appear to support him but ... Robin here is just cherry-picking a scientific community. ... I'm not a scientific expert, I'm simply relying on common sense for my skepticism.
This is, in a nutshell, exactly the problem of the topic of this post. When people hear a claim that the relevant experts deviate from the perceived media consensus, they find it easier to attribute this to cherry-picking than to an actual deviation. Do you think I am cherry-picking on many worlds as well?
Tyler: This blog has so far lacked deep analysis as to how one should pick one's relevant community of experts, but surely common sense skepticism of IQ combined with the opinions of most psychologists who haven't actually studied it would tell you one thing while asking scientists who have studied IQ would tell you another. Is that, to you, cherry picking? I certainly think that for the only plausibly relevant definition of informed then informed scientists expect uploads to exist within several centuries with P >.1 Since most informed scientists in the relevant manner are not informed economists or are informed about other matters that counter-indicate for Robin's conclusion, I don't think that they expect us to all become uploads or anything even fairly close to it.I appreciate your agreement with me that it is misleading to label every type of progress as economic growth.
Reviewing the items, I think I mostly agree with Robin. I have different views on uploads, but wouldn't but the chance of anyone becoming an upload after centuries as low as Cowen's 1%. Rather, my view is that uploads are likely to come last - and have correspondingly low economic impact. p(success) = one in ten thousand also seems over-confident for Cowen's cryonics prediction to me. Rah, the belief-value dichotomy!
I have a practical question: Just $200 per year for Alcor? Even with term life insurance, an $80K (charge for head only) policy is over $500/year for someone with Robin's published stats; and that doesn't even protect against senescence like a more expensive whole life policy. Add Alcor's yearly dues of $600 or $1200, and that $200 figure starts to look an order of magnitude too low.
Matthew, thanks for the wiki link. But, if convenient, could you please provide some direct links (or summary of results) to these negative polls? The wikipedia article reports the following:
"A 2005 minor poll on the Interpretation of Quantum Mechanics workshop at the Institute for Quantum Computing University of Waterloo produced contrary results, with the MWI as the least favored."
But this poll consisted of a sample of 4 (four) researchers.
I'm also curious whether there is anybody here who subscribe to MWI "without the slighest degree of doubt about its correctness"?
And how can you be sure that the vast majority of participants on this blog subscribe to the MWI? My own probabily would be closer to the mid-way mark than to either 0 or 1, and I have no idea what most of the other people here think about the MWI.
Interesting. The high-water polls for MWI have been mentioned, but not any of the contrary polls (all listed in the Wikipedia article on MWI). I would suggest that is not very surprising on a blog where the vast majority of the partipants subscribe to the MWI because it fits best with their beliefs about the nature of reality, some of them without the slightest degree of doubt in its correctness.
However when the title and raison d'être of the blog is "overcoming bias", I think we ought to raise the expectations up a notch. Go ahead and mention the contrary polling data -- it will enhance your truth orientation (and credibility to boot!)
My two cents on "many worlds".
I think that physics today is in about the condition that physics was a hundred years ago. There is a well establshed model that cannot explain certain aspects of reality. It is unclear what to do about this. The only thing that is clear is that something new will probably be needed.
To assign high likelyhood (to "believe") in any of several deeper explanations is hardly prudent. In fact the most important thing may be to be contrary :-)
I think the most interesting question is on how important 'overcoming bias' is. I would rate it as fairly important, but certainly not more important than overcoming laziness or some such. And I suspect that a well governed bias can be a useful time-saving technique that has more positive effects than negative. So I value people like Robin who [mistakenly] think that overcoming bias is the preeminent value because their endeavours may reveal things which when integrated into my own biases are very useful.
But the effort to reward ratio in overcoming bias is like many things--subject to marginal returns.
Zubon, a good question.
Am I odd for thinking that blogs are an odd place to work out this meta-disagreement when physical separation is not an issue? If there are going to be rounds of back-and-forth about what exactly x means, lunch seems like a good time to pin that down; text report-outs can follow. Perhaps there is some value in showing the back-and-forth to the public or in fixing words in text, but I imagine you two will see each other some time in the next week or two.
On the MW, one poll is reported here: http://www.anthropic-princi.... It is probably a couple of decades old, but my impression is that MW has continued to have the wind in its sails, so I don't think support has declined since then:
"Political scientist" L David Raub reports a poll of 72 of the "leadingcosmologists and other quantum field theorists" about the "Many-WorldsInterpretation" and gives the following response breakdown [T].
1) "Yes, I think MWI is true" 58%2) "No, I don't accept MWI" 18%3) "Maybe it's true but I'm not yet convinced" 13%4) "I have no opinion one way or the other" 11%
If Robin, based on his own investigations, assigns a probability just slightly higher than the physicist consensus, I don't think this would count as a significant disagreement.
On "Probability of uploads within centuries < 1%" and "Probability of 'If your head is cryogenically frozen today, you will be alive in 2100.' < 0.01%", I have to say that these seem astonishingly low. Even if we just use arguments from authority, and even if we discount the most-informed authorities as being self-selected for optimism about these prospects, it is hard to see how we could get estimates that are that low. Would Tyler claim that if we assembled historical data on cases in which a better-informed scientific minority held a view that conflicted with a less-informed scientific majority about what technological feats might be possible a century later, we would find that the minority had been right only in 1/100 to 1/10,000 of the cases?
I do think that if we are going to improve the way we deal with disagreements, one important first step is to try to be very careful about how we specify the contested claim. One of the most significant methodological advances in philosophy over the past century or so has been to appreciate the importance of precision of language and care in how we articulate our theses. This methodological insight could probably be more widely applied. Taking a vague disagreement, and, through analysis and distinctions, transforming it into a (set of) precise disagreement(s) can be a major research contribution if done well.
Robin's choice of experts on quantum computing as an estimator of p is a good one -- such people are much more likely to have thought about the foundations of qm than average physicists.
I posted the list below on the other thread, but it may be of use here. I have been told that Feynman and Gell-Mann would ofted *aggressively* proselytize about many worlds to visitors at Caltech.
Some eminent physicists who (as far as I can tell) believe(d) in MW: Feynman, Gell-Mann, Hawking, Steve Weinberg, Bryce DeWitt, David Deutsch, Sidney Coleman ... In fact, I was told that Feynman and Gell-Mann each claim(ed) to have independently invented MW, without any knowledge of Everett!http://infoproc.blogspot.co...
So the main evidence that I am a "logical atomist" is that I want to distinguish the possibility that our differing values lead to our differing priorities on overcoming bias, from the possibility that our differing priorities result from some as yet unstated factual belief related to overcoming bias? Is my interest in the particular distinction I am considering an especially potent evidence of "logical atomism", or is the inclination to draw any distinction whatsoever equally suggestive evidence of such an alarming tendency?
Robin is free to tell me that I don't disagree with him on the many worlds interpretation of quantum mechanics. I don't know much about what the scientific consensus is, but I do believe that such a consensus is my best estimator, given my limited understanding. Robin talks "as if" he has a higher p than the consensus, but of course he is free to announce that he doesn't.
Robin juxtaposed these two sentences:
"Nerd, I haven't accepted Tyler's label of "logical atomism", and I'm still not even very clear on what it means.
Quantum, it is not clear that Tyler and I disagree about any fact regarding overcoming bias; it could just be a preference difference."
Again, this is the heart of the matter. And the second sentence shows just how strong Robin's logical atomism is. It reflects Robin's reductionism to call it "just a preference difference." Nerd remains right on target.
Matt, I don't think I have made any claims about consciousness, just on reproducing I/O patterns. I agree with you that AI advocates have been too optimistic about rates of progress, though as I was an AI researcher for nine years, I'm pretty sure few AI researchers ever endorsed 5-7 year forecasts. And I consider myself to be a singularity skeptic in many ways.
My claim is not that "informed scientists expect us all to become uploads" but that those most informed specifically about both brain modeling and emulation expect, with at least "p=0.1", whole brain emulation to be technically feasible within a few centuries at most.
I would not concede that "brain modeling" scientists are experts on consciousness. In fact, I suspect that the people I would consider experts on consciousness and your list of such experts would have very little overlap.
In fact, a great many of my differences with you on your list of fourteen wild ideas (I agree with some of them) have to do with differences in whose opinion should be worth listening to on any particular subject.
You want to defer to an "expert" opinion, and I feel that your lists of "experts" are so subject to selectionist and other effects that their opinion is very biased and questionable.
Just as a single example, I would submit that the fifty years of hearing "AI in 5-7 years" should produce more skepticism about claims of strong AI in the future. I most certainly would not trust the predictions of people working in AI over the predictions of well-informed outsiders.
We could say the same thing about many other putative experts in very speculative fields, with vested financial, career and emotional interests in rapid progress or at least perceptions of rapid progress.
Frankly I feel bad making these kinds of comments on a board full of true believers in these kinds of Singularian ideas. I have no particular desire to rain on anyone's parade, and some of the Singularian activities are likely to produce modest or substantial fruit (even though not the everlasting life hoped for by their proponents). The only reason I bother to speak up here is the putative purpose of this forum to "overcome bias", the lack of many (any?) Singularity-skeptic voices here over an extended period of time, and a genuine wish to support Robin Hanson's heartfelt continuously stated desire to discover truth. I do believe he is very sincere in that quest, and I do feel it would be a worthwhile quest for many other people who contribute here to take up as their own, which would require less epistemic arrogance than some of the posts here demonstrate.
You certainly won't overcome bias and discover truth with a room full of true believers. But the temptation is ever-present for me to move on and let the millenarian parade-planners have their fun, and sooner or later I am likely to succumb to it.
Tyler, I don't yet accept this "logical atomist" label you have put on me, and even after reading a bit I'm not entirely clear on what it means. I certainly don't think the label applies to me merely because I objected to having my views called "strange" and "crazy" on the basis of some other vague labels you invoked, and therefore ask you to be as clear as possible about where you think we disagree.
My claim is not that "informed scientists expect us all to become uploads" but that those most informed specifically about both brain modeling and emulation expect, with at least "p=0.1", whole brain emulation to be technically feasible within a few centuries at most. Are you suggesting scientists with this specific expertize are biased because their topic is "unusual"?
As per the title of this post, the contrarian point, if any, (and not my contribution btw) is to combine this technical conclusion with social analysis to predict social consequences. My contribution, if any, is to apply standard formal economic analysis to this situation.
Selection bias -- which experts work on certain unusual topics -- may appear to support him but ... Robin here is just cherry-picking a scientific community. ... I'm not a scientific expert, I'm simply relying on common sense for my skepticism.
This is, in a nutshell, exactly the problem of the topic of this post. When people hear a claim that the relevant experts deviate from the perceived media consensus, they find it easier to attribute this to cherry-picking than to an actual deviation. Do you think I am cherry-picking on many worlds as well?
Tyler: This blog has so far lacked deep analysis as to how one should pick one's relevant community of experts, but surely common sense skepticism of IQ combined with the opinions of most psychologists who haven't actually studied it would tell you one thing while asking scientists who have studied IQ would tell you another. Is that, to you, cherry picking? I certainly think that for the only plausibly relevant definition of informed then informed scientists expect uploads to exist within several centuries with P >.1 Since most informed scientists in the relevant manner are not informed economists or are informed about other matters that counter-indicate for Robin's conclusion, I don't think that they expect us to all become uploads or anything even fairly close to it.I appreciate your agreement with me that it is misleading to label every type of progress as economic growth.