Thursday, November 17, 2005
MU Book Club Update: Faster, Faster!
by Tom Bozzo
Towards the end of the first section of Accelerando, whose action takes place around 2020 (pre-Singularity, and thus still largely recognizable), Stross poses a good question for those of us who now need to be pried away from our portable computers and wireless highish-speed data networks: When will there be as much of one's identity in the "metacortex" as in the wetware "real" thing?
Right now, I will feel dislocation pretty quickly in the absence of halfway decent internet access and the information gathering tools it enables, and I'm far from the frontiers thereof. (The big advancement for me this year was giving in to RSS to eliminate the need for a lot of time-consuming pointing-and-clicking to follow the hundred-odd blogs and conventional news sources I follow daily or nearly so. Without it, work, family, and Legos would have a hard time fitting into the remainder of the day.) In a way, the what separates this world from Stross's is the absence, so far, of software that makes much good use of idle processor cycles beyond donations to the occasional massively parallelized computing project.
That transition could happen subtly, as indeed it does in Accelerando, where at first the state of the art is something like what's implied by the IBM "Park Bench" commercial of the bubble days of the previous decade. That's the one where the annoying guy scares the pigeons in the Piazza San Marco trading stocks with his wearable computer. (The Accelerando website has a page listing some of the predictions/assumptions for its future.) That is, computing is more ubiquitous but mostly does the sorts of things people presently make money and/or waste time doing.
Mass-market computing advancements seem to have been biased more strongly towards faster-but-still-dumb computing — as in computing jobs that took an afternoon when I started my job that now require no more than a minute or two, but where the computer's agency in initiating the job is comparable to the KitchenAid's in assembling the dough for a batch of cookies — than I'd have guessed back in the early nineties when I got some exposure to research in machine learning. Though it does look like there's quite a bit of effort is going into computational economics over at the Santa Fe Institute (and seemingly less in the way of other machine learning research than seemed to be going on, though that's based solely on the last couple years' online working papers). I should probably read this from Brian Arthur for a flyby of the latest.
Reading: Accelerando by Charles Stross.
Towards the end of the first section of Accelerando, whose action takes place around 2020 (pre-Singularity, and thus still largely recognizable), Stross poses a good question for those of us who now need to be pried away from our portable computers and wireless highish-speed data networks: When will there be as much of one's identity in the "metacortex" as in the wetware "real" thing?
Right now, I will feel dislocation pretty quickly in the absence of halfway decent internet access and the information gathering tools it enables, and I'm far from the frontiers thereof. (The big advancement for me this year was giving in to RSS to eliminate the need for a lot of time-consuming pointing-and-clicking to follow the hundred-odd blogs and conventional news sources I follow daily or nearly so. Without it, work, family, and Legos would have a hard time fitting into the remainder of the day.) In a way, the what separates this world from Stross's is the absence, so far, of software that makes much good use of idle processor cycles beyond donations to the occasional massively parallelized computing project.
That transition could happen subtly, as indeed it does in Accelerando, where at first the state of the art is something like what's implied by the IBM "Park Bench" commercial of the bubble days of the previous decade. That's the one where the annoying guy scares the pigeons in the Piazza San Marco trading stocks with his wearable computer. (The Accelerando website has a page listing some of the predictions/assumptions for its future.) That is, computing is more ubiquitous but mostly does the sorts of things people presently make money and/or waste time doing.
Mass-market computing advancements seem to have been biased more strongly towards faster-but-still-dumb computing — as in computing jobs that took an afternoon when I started my job that now require no more than a minute or two, but where the computer's agency in initiating the job is comparable to the KitchenAid's in assembling the dough for a batch of cookies — than I'd have guessed back in the early nineties when I got some exposure to research in machine learning. Though it does look like there's quite a bit of effort is going into computational economics over at the Santa Fe Institute (and seemingly less in the way of other machine learning research than seemed to be going on, though that's based solely on the last couple years' online working papers). I should probably read this from Brian Arthur for a flyby of the latest.
