AI Alignment “Scaffolding” Project Ideas (Request for Advice)

I be­lieve that AI Align­ment is al­most cer­tainly the most press­ing is­sue for the fu­ture of hu­man­ity. It seems to me that the great­est thing that could hap­pen for AI al­ign­ment re­search is prob­a­bly re­ceiv­ing a whole lot more brains and money and poli­ti­cal spon­sor­ship. The pub­lic benefit is ex­traor­di­nary, and the po­ten­tial for pri­vate profit very small, and so this will need sub­stan­tial pri­vate or gov­ern­ment sub­sidy in or­der to re­ceive op­ti­mal re­source al­lo­ca­tion.

In or­der to start think­ing about strate­gies for achiev­ing this, I pic­ture sci­en­tific work as a sort of sig­nal­ling sys­tem be­tween re­search, the ed­u­ca­tional sys­tem, gov­ern­ment, and in­dus­try, as di­a­grammed be­low. I want to ap­ply the ne­glected/​tractable/​im­por­tant frame­work to this di­a­gram to ex­plore po­ten­tial pro­jects.

The Tech­ni­cal Side

a) Pro­fes­sional tech­ni­cal work on AI alignment

b) Ama­teur and stu­dent learn­ing con­tribut­ing or lead­ing to tech­ni­cal work

c) Meta-anal­y­sis of the state of the art, risks and re­wards, mile­stones, and big ques­tions of AI and AI al­ign­ment, via sur­veys, fore­casts, overviews of differ­ent per­spec­tives, etc.

d) Aware­ness-rais­ing dis­cus­sion and ex­pert ad­vice for policy-mak­ers, the pub­lic, and po­ten­tial ca­reer chang­ers/​donors

e) Laws, reg­u­la­tions, and pro­jects cre­ated by leg­is­la­tors, pub­lic policy mak­ers, and pri­vate institutions

f) Pres­sure by in­dus­try lob­by­ists and geopoli­ti­cal ten­sion to soften AI al­ign­ment con­cerns and go full steam ahead with AI de­vel­op­ment.

The Poli­ti­cal Side


1) Do any of the fol­low­ing ex­ist?

  • A com­pre­hen­sive AI al­ign­ment in­tro­duc­tory web hub that could serve as a “fun­nel” to turn the cu­ri­ous into the aware, the aware into am­a­teur learn­ers, am­a­teurs into for­mal ma­chine learn­ing PhD stu­dents, and PhDs into pro­fes­sional AI al­ign­ment re­searchers. I’m imag­in­ing one that does a great job of or­ga­niz­ing books, blogs, videos, cur­ricu­lum, fo­rums, in­sti­tu­tions, MOOCs, ca­reer ad­vis­ing, and so on work­ing on ma­chine learn­ing and AI al­ign­ment.

  • A for­mal cur­ricu­lum of any kind on AI alignment

  • A de­part­ment or even a sin­gle out­spo­kenly sym­pa­thetic offi­cial in any gov­ern­ment of any in­dus­tri­al­ized nation

  • Any gov­ern­ment spon­sor­ship of AI al­ign­ment re­search whatsoever

  • A list of con­crete and de­tailed policy pro­pos­als re­lated to AI alignment

2) I am or­ga­niz­ing an on­line com­mu­nity for older ca­reer chang­ers, and sev­eral of us are in­ter­ested in work­ing on this is­sue on ei­ther the policy or tech­ni­cal side. It seems to me that in the pro­cess of ed­u­cat­ing our­selves, we could prob­a­bly work to­ward cre­at­ing one or more of these pro­jects if any of them are in­deed ne­glected. Would this be valuable, and which re­source would it be most use­ful to cre­ate?