I too recently had a baby and left a government job in AI (in the UK: AISI).
Thank you for taking seriously, and helping society orient to, the prospect of AI-powered agents and firms.
Rather than all the AI action taking place in one explosive development setting and bursting into the world as a sovereign singleton, a lot of the scene-setting for the ‘really critical’ turning points and decisions may not take place in quite the world of today, but perhaps rather be shaped over a period of months or years where advanced intelligent resources and some semi-autonomous AI systems are widely proliferated in an ‘economy of agents’. We want that world to be as legible, harmonious, and judiciously managed by and for human(e) society as possible.
This means avoiding gradual disempowerment, and it also means maintaining trust in institutions and justified optimism for the potential of cooperation, rather than chaos and civic decay. It could matter what patterns are set in motion for this new economy before it emerges. Like urban planning vs sprawl - once patterns set in, they might be very hard to change.
Two crucial challenges emerge: establishing liability to prevent accountability voids, and ensuring that activity remains coupled to the ultimate motivating cause of some human principal. Without these, rogue accumulation of power and wealth could lead to gradual disempowerment of humans, or delegation could insulate actors (human or AI) from the collectively-aligning mechanisms of liability and incentive internalisation which have been the foundation of human society. Technical solutions, together with well-initialised norms and legal frameworks, can be developed now to ensure that AI agents operate within a framework where good equilibria are possible.
Mazal tov and congratulations! While still not a father, I resonate with the desire to be there for family. It sounds like you are on to bigger and hopefully better things.
Wise. Thanks for this post and its human fullness. I teach your epigram, and its trailing lines, in my English class. I am also (maybe) in your gm.edu email stack from about 12+ months ago—we discussed the possibility of sharing with high-schoolers the challenges of regulating “a being darkly wise, and rudely great” (not the words we shared at the time). If that laboratory interests you still, let’s see what we might do & learn: paul_erb@woodberry.org.
Fantastic foresight on prepping for the new family addition! I'm a father in a dual income household, and prioritizing time with young kids is the right outlook/plan to have. Good luck with everything, and looking forward to your future writings.
Hard to imagine a more productive four month stint in government, looking forward to your work here and at FAI!
> Barges into OSTP
> Writes AI Action Plan
> Refuses to elaborate further
> Leaves
Okay, actually:
> Goes back to elaborating further as a full-time job
Your self-aware articulations of your personal strengths and aspirations is inspiring. Bodes well for your future as a writer on AI policy.
Hope all goes well with your first child.
Congratulations Dean!
Fatherhood is amazing..
Rock on!
I too recently had a baby and left a government job in AI (in the UK: AISI).
Thank you for taking seriously, and helping society orient to, the prospect of AI-powered agents and firms.
Rather than all the AI action taking place in one explosive development setting and bursting into the world as a sovereign singleton, a lot of the scene-setting for the ‘really critical’ turning points and decisions may not take place in quite the world of today, but perhaps rather be shaped over a period of months or years where advanced intelligent resources and some semi-autonomous AI systems are widely proliferated in an ‘economy of agents’. We want that world to be as legible, harmonious, and judiciously managed by and for human(e) society as possible.
This means avoiding gradual disempowerment, and it also means maintaining trust in institutions and justified optimism for the potential of cooperation, rather than chaos and civic decay. It could matter what patterns are set in motion for this new economy before it emerges. Like urban planning vs sprawl - once patterns set in, they might be very hard to change.
Two crucial challenges emerge: establishing liability to prevent accountability voids, and ensuring that activity remains coupled to the ultimate motivating cause of some human principal. Without these, rogue accumulation of power and wealth could lead to gradual disempowerment of humans, or delegation could insulate actors (human or AI) from the collectively-aligning mechanisms of liability and incentive internalisation which have been the foundation of human society. Technical solutions, together with well-initialised norms and legal frameworks, can be developed now to ensure that AI agents operate within a framework where good equilibria are possible.
Mazal tov and congratulations! While still not a father, I resonate with the desire to be there for family. It sounds like you are on to bigger and hopefully better things.
Wise. Thanks for this post and its human fullness. I teach your epigram, and its trailing lines, in my English class. I am also (maybe) in your gm.edu email stack from about 12+ months ago—we discussed the possibility of sharing with high-schoolers the challenges of regulating “a being darkly wise, and rudely great” (not the words we shared at the time). If that laboratory interests you still, let’s see what we might do & learn: paul_erb@woodberry.org.
Fantastic foresight on prepping for the new family addition! I'm a father in a dual income household, and prioritizing time with young kids is the right outlook/plan to have. Good luck with everything, and looking forward to your future writings.
Congratulations to you and your wife!
Congratulations—and best of luck with the new addition!
Thank you for your important contributions, Dean!
Mazel tov! Sincere best wishes for the blessed event.