I recently completed an entry on Leibniz's Theodicy for the Stanford Encyclopedia of Philosophy. Leibniz predates AI, but I could not help making the connections while I wrote.
Leibniz published in 1710 to answer a hard problem: God created intelligent, agential beings in His own image, and these creatures caused enormous harm. Who is responsible, and could God have done better?
Now, we're building systems modeled on our own intelligence, granting them agency, and then asking: who is responsible when they cause harm? Their creators and designers? The systems themselves? The conditions that made failure inevitable?
Three ideas from Leibniz map directly to AI governance:
1. The author-of-sin problem → the accountability gap. Leibniz asked: if God creates "all the opportunities for sin" and even "provokes the will of the agent," how is God not responsible? We ask the same of companies deploying AI. If you design the system, train it, and release it, how much distance can you claim from its failures?
2. The best-possible-world defense → the tradeoff argument. Leibniz argued that some evils are inseparable from the goods they enable. "A dissonance in the right place gives relief to harmony." And we have recently heard, from AI companies, that some harms may be the price of a system that, on balance, does enormous good.
3. Deficient causation → emergent failure. Leibniz argued, with subtlety, that sin arises not from what God puts into creatures, but from their inherent finitude. Our connatural limitations become "deficient causes" of harm. And for AI: there can never be enough data, compute, and training to guarantee virtue.
Leibniz himself refused easy answers. The creator's good intentions are not enough. Human freedom does not excuse the author of our nature and circumstances. And evil itself is real enough to demand an explanation: you can't take credit for two bowings of the violin without taking responsibility for the dissonance. Most importantly, Leibniz insisted that philosophical answers provide real comfort to the afflicted.
Three centuries later and the engineers have theological problems now.
I'm not sure it's worth the upkeep to have comments. Seems that mostly spammers comment, and rarely real people. I just wanted a low-maintenance commenting system and Commento seemed to work decently at the time. I'm now noticing it's showing some CORS error, so I guess comments have been broken on my site for some time, doh...
Had the chance to see him speak at Google ten years back. I took a bunch of photos and added them to Wikimedia Commons for anyone interested in publishing articles: https://commons.wikimedia.org/wiki/Category:John_Searle
(this isn't a promotion, these are free to use photos)
Reducing methane emissions is the single fastest way to slow the rate of global warming today. In order to reduce emissions we first have to know where they're being emitted and who is responsible. The most effective way to do this on a global scale is from space, which is why we're launching a satellite to measure and attribute methane emissions. We are building tools to put in the hands of folks who will best be able to make the data actionable.
MethaneSAT is part of EDF, the Environmental Defense Fund.
I recently completed an entry on Leibniz's Theodicy for the Stanford Encyclopedia of Philosophy. Leibniz predates AI, but I could not help making the connections while I wrote.
Leibniz published in 1710 to answer a hard problem: God created intelligent, agential beings in His own image, and these creatures caused enormous harm. Who is responsible, and could God have done better?
Now, we're building systems modeled on our own intelligence, granting them agency, and then asking: who is responsible when they cause harm? Their creators and designers? The systems themselves? The conditions that made failure inevitable?
Three ideas from Leibniz map directly to AI governance:
1. The author-of-sin problem → the accountability gap. Leibniz asked: if God creates "all the opportunities for sin" and even "provokes the will of the agent," how is God not responsible? We ask the same of companies deploying AI. If you design the system, train it, and release it, how much distance can you claim from its failures?
2. The best-possible-world defense → the tradeoff argument. Leibniz argued that some evils are inseparable from the goods they enable. "A dissonance in the right place gives relief to harmony." And we have recently heard, from AI companies, that some harms may be the price of a system that, on balance, does enormous good.
3. Deficient causation → emergent failure. Leibniz argued, with subtlety, that sin arises not from what God puts into creatures, but from their inherent finitude. Our connatural limitations become "deficient causes" of harm. And for AI: there can never be enough data, compute, and training to guarantee virtue.
Leibniz himself refused easy answers. The creator's good intentions are not enough. Human freedom does not excuse the author of our nature and circumstances. And evil itself is real enough to demand an explanation: you can't take credit for two bowings of the violin without taking responsibility for the dissonance. Most importantly, Leibniz insisted that philosophical answers provide real comfort to the afflicted.
Three centuries later and the engineers have theological problems now.
Link to the Stanford Encyclopedia of Philosophy entry: https://plato.stanford.edu/entries/leibniz-evil/