6 ms·

"Even with nuclear weapons, if you rewind the tape you notice that it turned out that in order to make a nuclear weapon you had to have these very rare raw mate

by uvdiv 15y ago

"Even with nuclear weapons, if you rewind the tape you notice that it turned out that in order to make a nuclear weapon you had to have these very rare raw materials like highly enriched uranium or plutonium, which are very difficult to get [a]. But suppose it had turned out that there was some technological technique that allowed you to make a nuclear weapon by baking sand in a microwave oven or something like that. If it had turned out that way then where would we be now? Presumably once that discovery had been made civilization would have been doomed." It works both ways; technology could find simpler ways to build superweapons, but it could also lower the entry barrier to complex, resource-intensive projects. Look at the consequences of even mild singularity predictions for robotics and AI -- things that lower industrial "costs" by orders of magnitude, things like 3D-printers, self-replicating machines, cheap and ubiquitous fab robots. Anyone could build incredibly sophisticated machines in their own homes -- including sports cars, jet engines, and giant TV screens, but equally, compact laser enrichment cascades [b] and nuclear weapons. Perhaps it's lack of knowledge on my part, but I don't see what will stop atomic bombs from being as common as handguns in 30-100 years. Fissile material is ubiquitous [c]; there's no barrier beyond economics and engineering, of exactly the sort near-term AI could unpredictably disrupt. [a] (Tangential silliness: terrestrial uranium WAS weapons-grade material a few billion years ago -- U-235 decays with a 0.7 billion year half life, compared to 4.5 billion years for U-238, so in geologic history the fissile fraction used to be extremely high. Another one for the "anthropic principle" bin: earth's intelligence must have evolved now, and not earlier, because if it had it would have trivially nuked itself...) [b] Check out [NYT][APS] -- it's actually a mainstream position among arms control experts to consider shutting down US research in laser enrichment, to prevent the knowledge from being developed at all. To me this plan sounds about as airtight as have suggested to close the NSA, to keep mathematical secrets like RSA from being discovered. [c] Here's a blogger who goes hiking and brings back uranium ore by the bucket [Willis], for his own hackery experimenting. Uranium ore occurs everywhere [Cameco]; even common granite is a low-grade ore containing 5-50 ppm (parts per million) U [AZGS]; and research suggests it's even feasible to extract it by bulk from seawater [NBF]. [NYT] http://www.nytimes.com/2011/08/21/science/earth/21laser.html?pagewanted=all http://www.nytimes.com/2011/08/21/science/earth/21laser.html... [APS] http://aps.org/units/fps/newsletters/201007/slakey.cfm http://aps.org/units/fps/newsletters/201007/slakey.cfm [Willis] http://carlwillis.wordpress.com/2008/02/20/uranium-chemistry/ http://carlwillis.wordpress.com/2008/02/20/uranium-chemistry... [Cameco] http://www.cameco.com/uranium_101/uranium_science/uranium/ http://www.cameco.com/uranium_101/uranium_science/uranium/ [AZGS] http://repository.azgs.az.gov/uri_gin/azgs/dlio/414 http://repository.azgs.az.gov/uri_gin/azgs/dlio/414 [NBF] http://nextbigfuture.com/2009/09/uranium-from-seawater-on-large-scale.html http://nextbigfuture.com/2009/09/uranium-from-seawater-on-la...

khafra 15y ago
Yup. Singleton superintelligent friendly AI is pretty much our only chance; and we have no idea how to build one.
angersock 15y ago
"Perhaps it's lack of knowledge on my part, but I don't see what will stop atomic bombs from being as common as handguns in 30-100 years. Fissile material is ubiquitous; there's no barrier but economics and engineering, of exactly the sort near-term AI could unpredictably disrupt." You guess right--it's your lack of knowledge. Fissile material is not useful for making nuclear weapons (of any more than the dirty bomb variety). So far, we seem to have converged on uranium and plutonium, and while at least uranium is relatively common in the Earth the processing and dredging required to get a useful amount of it, and then the refining to get the useful isotopes from that, is nontrivial. Engineering and fast computers are great and all, and will get you arbitrarily close to the physical limits--but we're there right now, and physics says you aren't getting a centrifuge with meaningful output in your garage. More troubling is the idea that frankly we've had the technology to build sports cars in our homes for decades and decades now. It's called buying a lathe, an arc welder, a mill, a forge. Despite the availability--cheap!--of these things to the general populace, not only is it not widespread among those who can--and there are damned few of those even with this magical Internet thing telling you how to do all of it--it will get less likely as people focus on cat pictures and tweeting. We aren't worthy of a singularity as a culture. EDIT: Thank you for attaching some interesting facts. Allow me to do the same: The useful fissile uranium isotope (235) accounts for bout .72% of naturally occurring uranium (http://web.ead.anl.gov/uranium/guide/facts/ http://web.ead.anl.gov/uranium/guide/facts/). Your hiking source claims uranium metal, but points out that it is completely locked up in slag and that only at large-scales does the approach seem tractable. Even allowing for that, the amount produced is negligible. That's just for metal--we aren't even talking about a usable isotope yet. Take 7/1000 of that result from the blogger, and wave a wand to make it pure enough to use. Now collect the (at least) several pounds needed of that to make a functioning device. Now do the machining on the rest of the device to make it function correctly (have fun with the berylium dust, if you go that route). Now do the timing electronics, and the charge shaping (if you go that route). This. Is. Not. Garage. Technology.

by uvdiv 15y ago

khafra 15y ago

Yup. Singleton superintelligent friendly AI is pretty much our only chance; and we have no idea how to build one.

angersock 15y ago

"Perhaps it's lack of knowledge on my part, but I don't see what will stop atomic bombs from being as common as handguns in 30-100 years. Fissile material is ubiquitous; there's no barrier but economics and engineering, of exactly the sort near-term AI could unpredictably disrupt." You guess right--it's your lack of knowledge. Fissile material is not useful for making nuclear weapons (of any more than the dirty bomb variety). So far, we seem to have converged on uranium and plutonium, and while at least uranium is relatively common in the Earth the processing and dredging required to get a useful amount of it, and then the refining to get the useful isotopes from that, is nontrivial. Engineering and fast computers are great and all, and will get you arbitrarily close to the physical limits--but we're there right now, and physics says you aren't getting a centrifuge with meaningful output in your garage. More troubling is the idea that frankly we've had the technology to build sports cars in our homes for decades and decades now. It's called buying a lathe, an arc welder, a mill, a forge. Despite the availability--cheap!--of these things to the general populace, not only is it not widespread among those who can--and there are damned few of those even with this magical Internet thing telling you how to do all of it--it will get less likely as people focus on cat pictures and tweeting. We aren't worthy of a singularity as a culture. EDIT: Thank you for attaching some interesting facts. Allow me to do the same: The useful fissile uranium isotope (235) accounts for bout .72% of naturally occurring uranium (http://web.ead.anl.gov/uranium/guide/facts/ http://web.ead.anl.gov/uranium/guide/facts/). Your hiking source claims uranium metal, but points out that it is completely locked up in slag and that only at large-scales does the approach seem tractable. Even allowing for that, the amount produced is negligible. That's just for metal--we aren't even talking about a usable isotope yet. Take 7/1000 of that result from the blogger, and wave a wand to make it pure enough to use. Now collect the (at least) several pounds needed of that to make a functioning device. Now do the machining on the rest of the device to make it function correctly (have fun with the berylium dust, if you go that route). Now do the timing electronics, and the charge shaping (if you go that route). This. Is. Not. Garage. Technology.