Mmm. I note that they cite my top-level article introducing the IM, but didn't find the later LJ post which more closely parallels their topic of what problems you'd choose to solve with one. (Not that that's a surprise, really; my LJ is neither linked from my main website nor Google-indexable.)
They've got an interesting list there; some things exactly match my own thoughts on the matter (simplified computer system architecture since the CPU needs no optimised specialist hardware to do jobs like graphics, eliminate all the hassle of designing code for performance rather than clarity, all 3D graphics done by full-on raytracing in real time, the coolest fractal exploration programs ever), but some seem downright unambitious (solving the TSP optimally in cases where they say the best heuristics are already only 1-3% off optimum, and similarly while the box-packing problem would certainly be done better by IMs it hardly seems like the most pressing issue that would drive their notional development) and some thoroughly unrealistic (greater speed on its own does not an AI make, you still need to actually know how to write AIs that work; similarly I'm unconvinced by the claim that computational improvements will make weather forecasting work properly without needing insanely accurate observations to base the computation on). I think my commenters tended to think bigger, at least, though of course who knows about better :-)
One of these days I keep meaning to get round to rewriting my IM article to incorporate more thought on the implications of the thing existing in the real world (and fix a couple of other points while I'm at it, but anyway). The fact that I have no good answer to the cryptography problem is mentioned in the current article but its social consequences aren't explored: currently we're living in a golden age in which codemaking generally beats codebreaking, but that would suddenly be reversed and nobody's communications would be private or unforgeable or anything unless they'd exchanged an infinite amount of OTP data in person beforehand. Then there's the question of spam: what effect would IMs have on the power balance between spammers and spam filter writers, and what would happen when somebody had the idea of sending an infinite amount of spam so that any nonzero failure rate in the filter was instant game-over? And that's without even getting into more woolly social questions such as whether you could use an IM to accurately model people in quantity in game-breaking ways: IM-derived optimal strategies for politicians? For economists? Stock market speculators? Fraudsters? One colleague to whom I talked about all this suggested that actually, it was a damn good thing Infinity Machines can't exist, because they would represent the end of civilisation as we know it!
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
They've got an interesting list there; some things exactly match my own thoughts on the matter (simplified computer system architecture since the CPU needs no optimised specialist hardware to do jobs like graphics, eliminate all the hassle of designing code for performance rather than clarity, all 3D graphics done by full-on raytracing in real time, the coolest fractal exploration programs ever), but some seem downright unambitious (solving the TSP optimally in cases where they say the best heuristics are already only 1-3% off optimum, and similarly while the box-packing problem would certainly be done better by IMs it hardly seems like the most pressing issue that would drive their notional development) and some thoroughly unrealistic (greater speed on its own does not an AI make, you still need to actually know how to write AIs that work; similarly I'm unconvinced by the claim that computational improvements will make weather forecasting work properly without needing insanely accurate observations to base the computation on). I think my commenters tended to think bigger, at least, though of course who knows about better :-)
One of these days I keep meaning to get round to rewriting my IM article to incorporate more thought on the implications of the thing existing in the real world (and fix a couple of other points while I'm at it, but anyway). The fact that I have no good answer to the cryptography problem is mentioned in the current article but its social consequences aren't explored: currently we're living in a golden age in which codemaking generally beats codebreaking, but that would suddenly be reversed and nobody's communications would be private or unforgeable or anything unless they'd exchanged an infinite amount of OTP data in person beforehand. Then there's the question of spam: what effect would IMs have on the power balance between spammers and spam filter writers, and what would happen when somebody had the idea of sending an infinite amount of spam so that any nonzero failure rate in the filter was instant game-over? And that's without even getting into more woolly social questions such as whether you could use an IM to accurately model people in quantity in game-breaking ways: IM-derived optimal strategies for politicians? For economists? Stock market speculators? Fraudsters? One colleague to whom I talked about all this suggested that actually, it was a damn good thing Infinity Machines can't exist, because they would represent the end of civilisation as we know it!