<html><head></head><body><div><span data-mailaddress="avant@sollegro.com" data-contactname="Stuart LaForge" class="clickable"><span title="avant@sollegro.com">Stuart LaForge</span><span class="detail"> <avant@sollegro.com></span></span> , 23/2/2015 8:49 PM:<br><blockquote class="mori" style="margin:0 0 0 .8ex;border-left:2px blue solid;padding-left:1ex;">
<br>Perhaps the time period between a civilization developing recursively
<br>self-improving general AI and its subsequent development of
<br>computronium is relatively short compared to geologic time scales.
<br>Computronium, being the maximally optimized medium for computation,
<br>quickly saturates the Beckenstein bound of their region of space-time
<br>by being so information dense. This causes their space-time to warp to
<br>the point of pinching itself off, forming an event horizon around
<br>them</blockquote></div><div><br></div><div>The problem with standard spacetimes with localized event horizons is that they have curvature singularities on the inside, and all timelike paths crossing the horizon will intersect the singularity in finite proper time. In short, the computronium will not last long in its own reference frame. I think there are some rather strong theorems (Penrose?) showing this. </div><div><br></div><div>Now, rotating black holes have fairly complicated interiors and can in theory contain closed timelike curves, which blows most standard computronium out of the water since they allow future results to adjust past inputs: you get a class of hyperturing computation (check out Scott Aaronson's paper on CTC computing http://arxiv.org/abs/0808.2669 ). The practical problem is that imploding matter likely doesn't produce the full topology needed, and hence the black hole is useless.</div><div><br></div>As a Fermi explanation black holes are essentially identical to quiet M-brains. It requires strong cultural convergence, plus that the value of communicating with an exterior spacetime always becomes less than the value of computation, plus that one can do a non-neglible amount of computation on the inside.<div><br></div><div>Seth Lloyd's ultimate laptop paper analysed computing right at the Bekenstein edge; that would likely be pretty visible since it runs "hot". <div>http://arxiv.org/abs/quant-ph/9908043</div><div><br></div><div>Now, if you can make wormholes or more complex topologies things get real fun... but I suspect the result is a blob of smart Planck density stuff indistinguishable from Planckscale noise. <br><div><blockquote class="mori" style="margin:0 0 0 .8ex;border-left:2px blue solid;padding-left:1ex;"></blockquote></div><br><br>Anders Sandberg, Future of Humanity Institute Philosophy Faculty of Oxford University</div></div></body></html>