Aeschylus

Someone asked me about electricity rates increasing in areas that are serving new AI datacenters. It's a fascinating story because it reveals multiple levels of coordination failure. 1) Yes, the rates went up. But you are purchasing kWh (energy) for $ (money) and their relative scarcity is part of the picture. 2) We have known now for many decades how to make arbitrary quantities of kWhs, but have largely chosen not to for various intangible reasons like utility regulation path dependency and the environmental impact of nuclear power, which is low enough that enlisted submariners access the reactor through the galley. 3) We have chosen to create many many more $ and spend them on a variety of more-or-less inflationary things, and almost no compounding, wealth creating deflationary things. As a result, to leading order, your electricity bill went up because your money became worth less, relative to the actually valuable and artificially scarce electricity you are purchasing with it. 4) Increasing electricity bills are primarily a problem for poorer people and people on fixed incomes, often the same classes of people who inflationary public spending is intended to help! This process disproportionately impacts older people and retirees, who can't simply re-enter the workforce to make more money now that their savings have been effectively taxed through profligate spending and inflation. 5) Something like 20% of the US are now people 60 and over, in the twilight years of economic productivity and thus especially vulnerable to debasement of currency, in addition to the indignities of physical and mental decline wrought by old age. 6) The existence of such a large number of dependent people is a testament to the success of public health, medicine, industrialized agriculture, and other infrastructure without which nearly everyone simply doesn't live long enough to get old and frail. 7) And yet human mortality increases 9% per year of life, a rate that far exceeds economic growth, now creating a widely shared burden of old age. In short, why do we let people get old and sick? 8) The NIH spends $50b a year on health research, publicly funded, upstream of starkly unproductive expenditure of public wealth, regardless of its social valence. Less than 0.1% of that studies the mechanisms of aging. The US spends $2.4t on Medicare and Social Security every year, attempting to put a band aid on an unavoidable and growing problem (human mortality) unless we attack it at its source. But we have chosen not to allocate any serious resources to the core remaining human health challenge - senescence. 9) Data centers will, by necessity, end up with captive internal power plants and thus become net sources rather than sinks of power to their communities. 10) AI has the potential to solve each of these issues: efficiency in government spending, expansion of power supply, human aging and health. It will certainly improve the efficiency of current human efforts in these directions. 11) It's not clear exactly who "we" is in each of these cases. This is a problem because legibility is key to responsibility. The problem isn't that datacenters are making great aunt Mabel's power expensive. The problem is that it's illegal to make more power, apparently impossible to spend less money, and aunt Mabel's age and lack of economic power is an addressable challenge that is mostly ignored.