Sunday, July 30, 2006
Saturday, July 22, 2006
Sue Savage-Rumbagh & Roger Lewin'sKanzi is an anecdotal presentation of the capabilities of modern, communicating apes, focusing on Kanzi, who has learned to use an ideograph-based keyboard. Savage-Rumbaugh relates many experiences that don't fit easily into the controlled scientific process in order to show us how much more capable Kanzi and his cousins are than the studies can demonstrate. At the end of it I'm fairly well convinced that Kanzi is communicating, has intentionality, has a mental model that includes the fact that others know different things than he does (sometimes more, sometimes less), can make plans, can and does lie, has a mental map of his (apparently large) normal range and can plan routes in it, and uses grammatic and semantic categories.
I think it is highly likely that these abilities are not shared with chimpanzees and bonobos in the wild, but that many of them would develop such skills if raised in appropriate environments. This conclusion isn't based directly on evidence that Savage-Rumbagh presents, but is my own deduction from the fact that more than a half dozen random individual chimps and bonobos have displayed similar talents when exposed to a variety of language tools more suited to their use than oral speech.
I'm not inclined to think that this changes the moral category of those species in toto, though it probably should have implications for how we treat the individuals that we have raised to such awareness. In order to justify drawing a line to include some chimps and bonobos as aware, and leave most as simple animals, I need to be clear about my thoughts about why we don't draw a similar line distinguishing normal humans from some subset that appears to be less aware than Kanzi. I think the argument that declaring some humans to be outside the protected boundary is a "slippery slope" is a good enough justification for treating all humans as if they were sapient. The difference in intellect between the chimps and bonobos who have acquired language and their kin who haven't is large enough that there's little hazard in drawing a line between them.
Savage-Rumbagh did most of her work in a lab that mainly focused on developmentally challenged clients. Some of the work with Kanzi suggested techniques for teaching them communication skills. These techniques turned out to help some of them significantly in mixing with their mainstream peers.
Tuesday, July 18, 2006
As I promised to try to do, I read Estep et. al's submission to TR's challenge, the one the judging panel declared to be the "most eloquent", along with de Grey's response and the author's rebuttal. Estep et. al. misunderstand their role in the challenge from beginning to end. de Grey has been promoting SENS, a proposal for a new approach to combat aging, for a few years. From the beginning he admitted it was a controversial approach combining leading edge biology and an engineering mindset and asked experienced scientists to tell him if he was making any fundamental mistakes in the science. He has led and participated in symposia with experts and published papers in respected journals, after earning a Ph.D. in the relevant discipline. Technology Review solicited written submissions that would demonstrate that de Grey's program was "so wrong that it was unworthy of learned debate."
There are good reasons for practicing scientists to recognize the signs of pseudo-science, and it's appropriate for their first reaction to be outright dismissal when they recognize those signs. When a proposal garners enough attention that outright dismissal is no longer effective, it's still appropriate for most scientists to ignore it until there's more evidence in its favor. But at the point at which someone decides that a head-on confrontation is called for, the participants in the process should no longer expect to get away with derision; their role at that point is to review the science, show that the conclusions are at odds with established results, or demonstrate that they can't be disproved in principal, and thus are outside the bounds of science. Their goal should be to analyze the proposals, and summarize their evaluation in enough detail that other scientist can read their summary rather than doing the primary analysis themselves.
Estep et. al. spend an inordinate amount of time referring to studies of pseudo-science and its practitioners, and explaining in what ways de Grey and his proposal resembles them. But that's of no value at this stage in the process. Pontin (the editor of TR, and instigator of the challenge) and de Grey's other detractors have already pointed out that SENS isn't normal science; the challenge is to come up with an evaluation of the science, and Estep skimps on that.
Estep accuses de Grey of tangling his new proposals in with other scientists' results, making it hard to distinguish what is well-founded from what is not. But that's part of the normal process of science. Scientists build on others' work, expand it in some areas, and correct mistakes in others. In effect, Estep's charge is that scientists perpetrating fraud should clearly separate their new fraudulent work from others' well-founded work in order to make the investigator's job easier. This is nonsense. It is Estep's responsibility to examine the foundations of SENS and find parts that aren't justified or justifiable. If he can't, he has failed to undercut the science. That failure wouldn't by itself show that SENS is valid, but it would be a hurdle passed. Becoming accepted as established science is a process of continuing to get past the hurdles.
Estep shows no evidence that he understands the distinction that de Grey makes in referring to SENS as an engineering approach. de Grey is very clear in drawing the distinction:
Concerning the difference between scientists and engineers in mindset and motivation — as opposed to laboratory expertise — that I have often mentioned, Estep et al. expertly make my point for me by noting that the only reason they engineer model organisms is to find things out. To quote them: "If we could easily predict the outcome, why bother going through all the trouble of actually doing the engineering?". I wonder if Estep et al. think the Wright brothers built their airplane in order to discover whether it would fly? I personally suspect that they built it because they were confident that it would fly and they wanted to build something that would fly. Estep et al.'s oversight of this motivation is quite breathtaking to anyone who understands that, since aging causes immense suffering and death, it is something to be explored not for the sake of curiosity alone but with the goal of actually doing something about it.
In their response, Estep et. al. point out that the Wright brothers used the scientific process to test and evaluate the components they were building, and that they were systematic in their efforts to evaluate all the effects that mattered in getting into and remaining in the air. This response completely misses the point. Estep et. al. think that progress is made by systematically expanding the frontier of what is known. de Grey proposes to take what is known and build an effective mechanism to solve a problem, learning as he goes. The relevant question is whether we know enough at this point to start the process. The Wright brothers didn't know the answers when they started. They weren't satisfied with asking all the interesting questions, either; they asked the questions that were blocking their path to building a successful flying machine. Useful questions for critiquing an engineering proposal include: Do we know enough to get started? Is the cost estimate reasonable? Are there reasons to believe that there is no solution to the problem (within the budget)?
Estep et. al. focus on showing that some of the particular approaches that de Grey suggests aren't completely justified in the literature. But the heart of the SENS proposal is a search for solutions to 7 identified pathologies; de Grey has identified multiple approaches to each of them, and tried to give plausible detail about at least one solution for each problem. The useful counter-arguments are that these 7 aren't exhaustive, or that no (affordable) solution is available to one or more of the seven. Estep et. al. say that de Grey's list of 7 pathologies "unscientifically exclud[es] others"; the only exclusion I can identify in their paper is "unrepaired DNA damage in post-mitotic tissues", though they also mention "largely uncharacterized and undiscovered damage and pathologies". de Grey's original claim about the seven pathologies was that these seven had all been identified and characterized by 1982, and no new ones have become apparent since then despite all the advances and attention in this area, so "uncharacterized and undiscovered" looks like a clear miss to me. de Grey's response to the post-mitotic damage is to point out that he had addressed that issue in a draft paper that Estep has been asked to review. Estep talks about other items covered in that paper, but drops the thread on post-mitotic damage. I count that as no challenge to de Grey's claim that there are only seven pathologies to worry about.
de Grey wrote in his rebuttal:
Particularly incongruous is their accusation that I use the media to skirt expert criticism, when the SENS Challenge itself is my most conspicuous effort to do just the reverse, exposing the public reticence of SENS's off-the-record detractors and thereby forcing them to make their supposed case in print.
(de Grey has made other efforts to get gerontologist to put their objections on the record.) Estep responds thusly:
TR Editor Jason Pontin claims this challenge was initiated by him and all available evidence supports this claim. Furthermore, we have evaluated only a few of SENS' weaknesses, as did Warner and colleagues (Warner et al. 2005), and Aubrey de Grey's responses demonstrate beyond a shadow of a doubt that he is completely insincere in his wish for scientists to critique SENS.
Notice that Estep wrote in the original paper:
ThisSENS Challengeis itself part of the pseudoscience archetype, and it is simply a culmination of ongoing challenges made by Aubrey de Grey to opponents of SENS to prove him wrong. This is a classic attempt to subvert the scientific process, it is known to be typical of pseudoscience, and it is described as such in Dr. Friedlander's book, which predates SENS by several years.
So, Estep started this particular exchange by saying that the SENS Challenge is stereotypical behavior of pseudoscientists; de Grey rebuts that the challenge is an example of his having solicited criticism, at which point Estep backtracks and says that de Grey wasn't responsible for the challenge anyway. de Grey's Methuselah Foundation had matched Technology Review's prize for which Estep competed. Seems like de Grey is on solid ground in describing his role as helping to encourage debate.
Estep's rebuttal contains a number of explicit "predictions". That interests me because of the obvious application of prediction markets to monitor the argument. Most of the predictions are untestable or uninteresting: "de Grey's complaints that he isn't getting a fair hearing will continue", "Very soon it will be apparent to all [...] that SENS [...] constitutes overt scientific misconduct." Two exceptions are worth noting.
First, Estep et. al. say "[de Grey]'s only contribution to pharmacological inhibition and modification [...] of telomerase, therapeutic deletion of the entire telomerase gene [...] from the genome, will be recognized to be a crude biomedical fantasy. It will be abandoned by all sensible people--and even de Grey's co-authors will cease to [...] discuss it publicly." This is a very hard statement to write a clear prediction market claim for. We've tried before, on FX for claims that a consensus will arise, or that some research field will be dropped, but the claim details are hard to agree on and judge. I'll leave the prediction standing here as a reminder that de Grey's opponents can be judged to have made a valid prediction if, in your opinion, this subfield goes away later.
The second prediction is easier to adapt as a prediction market claim. The background is that de Grey is apparently accepted as an expert in the area of mitochondrial DNA (mDNA) mutation and repair. His proposal to combat progressive damage to mDNA is to ensure that Nuclear DNA (nDNA) can construct the few proteins (that's "allotropic expression") that the mitochondrion can't already get from the cell's nucleus. de Grey's response says
the shallowness of Estep et al.'s analysis is revealed especially starkly by their contradictory statements concerning allotropic expression in cell culture, which midway through their analysis they call an example of "assumptions and technologies that reside firmly in the realm of fantasy" but, in their summary, they call an example of "routine biology experiments."
Estep et. al. respond
successful allotropic expression of all 13 mitochondrial coding regions while maintaining mitochondrial and cellular function is a technology that resides in the realm of fantasy. Nevertheless, attempts to achieve these things are without a doubt routine biology experiments.
This makes it appear that Estep et. al. are willing to use the word "fantasy" to describe things that haven't been achieved yet. At least they didn't make any attempt to show that it couldn't be achieved.
Estep et. al. point out that de Grey bet someone publicly in 2000 that this would be known technology before the end of 2005, and it still hasn't been achieved. de Grey responds that he was counting on the imminent release of a paper showing the technique was possible (which was in review at that point) to spur further work. The paper only came out in 2005, so the follow-on work has only just started.
Estep's one explicit prediction is "We believe that this problem is [complicated] and it will remain unsolved for a very long time--if it is ever solved. Aubrey de Grey will continue to make excuses [...] to explain the lack of progress in this area." This, unfortunately, doesn't give a useful time scale, but de Grey's original projection can be extended to give a reasonable deadline.
Here is the background de Grey gives:
They lampoon my prediction from 2000 concerning AE, without mentioning that I made it assuming that Zullo et al.'s seminal breakthrough ( which I presented at the time I made the bet) would be published imminently in Science (where it was then in review), stimulating effort to perfect this approach; in fact, followup effort remained negligible until it was finally published in 2005. Thus, it is grossly misleading to suggest that my overoptimism arose from underestimating how hard AE is — and I fully explained this recently in a reply to Estep on a well-known mailing list (subscription required in order to review archives).
So in 2000, de Grey expected that imminent publication in Science would produce results by 2005. The paper was published in 2005, but not in a publication as prestigious as Science. Estep says it will be a very long time if ever. I'll propose a claim for the Foresight Exchange that it will be solved by 2012, which is longer than the 5 years de Grey originally expected, but gives some leeway for the reduced publicity the paper actually got. If it still seems to be 5 years away at that point, we'll have to conclude that de Grey's timeframes were way off.
Overall, I have to agree with TR's panel. Estep et. al. may have found some holes in de Grey's specific proposed therapies (it's hard for me to tell; unlike de Grey, Estep et. al. don't provide layman's versions of any of their technical arguments), but they didn't show that these holes make the entire effort unlikely to succeed, and they didn't show that de Grey's proposal doesn't address a useful goal.
As Marc Stiegler commented when I told him who was on the review committee, they seem predisposed to finding merit in de Grey's plan. Ventner is best known for his public commitment to develop technology to sequence the human genome before it was scientifically justifiable--and delivered. Rod Brooks' mantra at the Mobile Robot Lab was "fast, cheap, and out of control"; they were building stuff that worked rather than pursuing pure knowledge. Nathan Myhrvold is more focused on inventions than basic research, even though his academic credentials are impeccable.
Wednesday, July 12, 2006
A year ago, I wrote about Jason Pontin (the editor of Techology Review) and his attack on Aubrey de Grey's SENS (Strategies for Engineered Negligible Senescence) proposal. Pontin was so upset at de Grey's efforts to defeat aging, that he offered $10,000 as a prize to anyone who could come up with a convincing argument that the proposal was "so wrong that it was unworthy of learned debate." (de Grey's foundation matched the money, making the prize $20,000.) They appointed a distinguished committee (Rodney Brooks; Nathan Myhrvold; Craig Venter; Anita Goel, a physicist and nanotech entrepreneur; and Vikram Kumar, an innovator and pathologist) which has issued its report. They received only three submissions that merited detailed evaluation, and decided that none of them constituted the convincing rebuttal that Pontin had asked for. (de Grey's response)
The Judges' statement emphasized that de Grey's proposals are closer to engineering than science, and none of the submissions had evaluated them on that ground. Brooks is quoted as having said "I have no confidence that they understand engineering, and some of their criticisms are poor criticisms of a legitimate engineering process." Since de Grey is talking about plausible approaches, fallback solutions, and costs versus benefits, it's not reasonable to attack him for not having proven experimentally all that he proposes, as you might a scientific paper or career. The committee even went so far as to describe some of the attacks as "name calling". That was also the reaction I had to Pontin's original article on SENS.
Pontin seems to have calmed down since he wrote the original diatribe. He salved his pride to some extent by giving MIT's half of the prize money to the entry the judges found to be the "most eloquent", even though that wasn't called for in the terms of the contest. I'll try to read the three submissions and the back and forth between de Grey and their authors. I'd like to understand the best arguments for and against; I'm hopeful that I'll learn something more about the biology and the chances for SENS' success. I somehow suspect that as scientists criticizing an engineering proposal, they'll spend their time showing that his suggestions aren't proven. What matters is whether the proposals are close enough to right that they can be corrected as we learn more. That's much harder to justify in a new area, and also much harder to attack convincingly.
I'm not sure where the burden of proof should be in that part of the argument. The parts of de Grey's proposal that I found most compelling the first time I heard him were that there are only seven interesting mechanisms underlying progressive metabolic deterioration, and that each of them can be defeated if you take an engineering approach. I hope there's at least some discussion of this point.
Tuesday, July 04, 2006
C. J. Cherryh's Chanur Series (Pride of Chanur, Chanur's Venture, The Kif Strike Back, and Chanur's Homecoming) tells a wonderfully involved story from an alien point of view. This is the quintessential example of a story in which the aliens seem to have different motivations and points of view that aren't quite human. In my review of MacLeod's Learning the World, I mentioned that his aliens seemed like people in bat suits. Cherryh's aliens are understandable to us, but different.
The story is told from the point of view of the crew of The Pride, a Hani ship. The Han are a feline species that are relatively new to space. The Pride becomes the unwilling caretakers of Tully, a human, who was a captive of the Kif, but escaped. Tully is the only human we ever meet, though the recent arrival of humans is affecting interspecies relations throughout Compact space. Through the course of the series, Tully is returned to humankind, and comes back to the Pride for the final installments.
The characters we meet (at least the oxygen breathers) all have plausible motivations that differ among members of the same species, but the way of thinking seems somewhat consistent within each species. This isn't the Dune style of interstellar civilization, where each planet has a single kind of habitat, and most of the members of any species are indistinguishable. We meet a few different Stsho, and they're uniformly fearful in the face of (even hinted at) violence, but they have different goals, and approach negotiations differently. The Kif are clearly the kind who look for and follow an alpha-leader, but we see some of the beta kif scheming to supplant a current leader, while others scheme to gain status by supporting the current leader. The different races of methane breathers are inscrutable in widely varying ways.
Pyanfar Chanur, the Captain of "The Pride", an interstellar merchant ship, must navigate an ever-deepening network of entangling alliances and enmities. Her friends and enemies are Hani and other; sometimes she understands their motivations, other times she has to guess wildly, or infer motivations from capabilities and actions.
I've been making my way through this series for several years, but it's apparently always been on the back burner: one of the books I take with me when I expect to have free time in a waiting room or between acts at a play. The series never showed up in the "Currently reading:" portion of my .sig line. Even so, I remember all the characters vividly, and can keep the various races straight. I think I half-consciously measure out my progress slowly in a book or series when I discover a kind of poetry in it. I want it to last. This was that kind of story. Cherryh has written several stories that have had that effect on me. The Chanur series was published in the early 1980's, but it doesn't seem dated. I enjoyed it immensely.
Oh, and there's a follow-on separate story: Chanur's Legacy. I look forward to it.
Sunday, July 02, 2006
Charles Mann's 1491 is a very interesting read. The main point is that the natives to the Americas were a lot more numerous and a lot more advanced before the arrival of the European adventurers than anyone realized until recently. Whatever you learned in school vastly underestimates the accomplishments and the scale of their civilizations. There are a lot of details still to be worked out, but that much seems clear.
Mann is very good at taking the discoveries and recent analysis and making the possibilities clear: there is disagreement in many cases about how many people were here, how much contact there was between groups, when and how they died off, and just how sophisticated they were, but at the end there is so much evidence that there was large-scale engineering in so many places (the North American Midwest, the Amazon basin, the Andes and along South America's west coast, and Central America) that the conclusion stands above the minor disagreements.
More than anything, it's an enjoyable read. Even though Mann is heaping up evidence, he tells good stories of how the people must have lived and struggled. He seems to make it clear when he is speculating, when the experts are in agreement on the basics, and when he is reporting on aspects of the history that are still in dispute. The North American natives didn't write anything down, so he can't tell many stories about individuals, but the South Americans had a variety of writing systems, so he's able to report on political struggles that demonstrate the extent of the civilizations and gives some hints about how the internal wars may have contributed to their demise in the face of the conquistadors and their diseases.
In all these areas, there are constructions that were large enough, and that have endured well enough that they are still visible if you know what to look for. Until recently, anthropologists and archaeologists didn't realize they were there, and so hadn't studied them. In the Andes and the western plains, there are dikes and large scale water works that have only been identified as such in the last twenty years. In the Midwest, there are humongous mounds that are clearly artificial. New pyramids have been discovered in the Amazon in the last twenty years using LIDAR that can penetrate the tree cover.
The mounds are a fascinating chapter all to themselves. My favorite is the Cahokia mound. Cahokia, near present day Saint Louis, was the largest (maybe the only?) city north of the Rio Grande for 300 years starting about 1000 years ago. They built a series of mounds, of which Monks Mound is the most interesting.
Its core is a slab of clay about 900 feet long, 650 feet wide, and more than 20 fee tall. From an engineering standpoint, clay should never be selected as the bearing material for a big earthen monument. Clay readily absorbs water, expanding as it does. The American Bottom clay can increase in volume by a factor of eight. Drying, it shrinks back to its original dimensions. Over time the heaving will destroy whatever is built on top of it.
To minimize instability, the Cahokians kept the slab at a constant moisture level: wet but not too wet. Moistening the clay was easy—capillary action will draw up water from the floodplain, which has a high water table. The trick is to stop evaporation from drying out the top. In an impressive display of engineering savvy, the Cahokians encapsulated the slab, sealing it off from the air by wrapping it in thin, alternating layers of sand and clay. [...] The final result covered almost fifteen acres and was the largest earthen structure in the Western Hemisphere: though built out of unsuitable material in a floodplain, it has stood for a thousand years.
The time of arrival of the pre-columbian inhabitants is more contentious than I realized. I've been using the figure of 13000 years before the present as the time that the first wave came over the Bering Strait, and wiped out all the mega-fauna. It appears that there were probably two waves of immigration before that, and possibly more, though the timing is completely unclear. There are indications that early humans were here as early as 30,000 years ago, and that there was another wave around 20000 years ago based on genealogical evidence and tantalizing archaeology. Those people didn't leave many artifacts, so there is still plenty of disagreement.