Quantum Computing : A preliminary research analysis report

[This is a linkpost to Quan­tum Com­put­ing : A pre­limi­nary re­search anal­y­sis re­port]

Quite re­cently, Google re­leased a pa­per claiming to have build a pro­grammable quan­tum com­puter ca­pa­ble of solv­ing in 20 mil­lisec­onds a sam­pling prob­lem that would take 2.5 days to solve in the fastest 2019 su­per­com­puter.

Among its po­ten­tial uses, Quan­tum Com­put­ing (QC) will al­low break­ing clas­si­cal cryp­to­graphic codes, simu­late large quan­tum sys­tems and faster search and op­ti­miza­tion.

This could have im­pli­ca­tions on some of the ar­eas of in­ter­est to a long ter­mist, in­clud­ing in Ar­tifi­cial In­tel­li­gence, Biose­cu­rity and Atom­i­cally Pre­cise Man­u­fac­tor­ing—as well as pre­sent­ing new tech­nolog­i­cal risks (such as un­der­min­ing the cur­rent in­fras­truc­ture for on­line trans­ac­tions).

In re­sponse to the re­cent de­vel­op­ments in the field and the above con­sid­er­a­tions, I re­solved to dis­cuss the rele­vance of Quan­tum Com­put­ing (QC) from the point of view of a philan­thropist, as to bet­ter un­der­stand the risks and benefits posed by this tech­nol­ogy.

A sum­mary of the main con­clu­sions I reached is re­pro­duced be­low. You can see the full re­port here. Sugges­tions and com­ments are wel­come!

Ex­ec­u­tive summary

  • QC could un­lock a sig­nifi­cant amount of com­put­ing power for quan­ti­z­able al­gorithms, and this could have im­pli­ca­tions on AI timelines and the kind of AI de­signs we will first see. This area of re­search seems well funded by big com­pa­nies (Google, IBM, Microsoft). More

  • Due to its po­ten­tial ap­pli­ca­tions to ad­vance trans­for­ma­tive AI, I would be medium ex­cited to see fur­ther re­search on the effects of QC com­pute over­hang on AI timelines and pos­si­ble policy im­pli­ca­tions for reg­u­la­tion of AI de­vel­op­ment. This line of re­search seems how­ever bot­tle­necked by our un­der­stand­ing of how ex­tra com­pute in gen­eral af­fects AI timelines, re­search on QC timelines and re­search in QC policy. More

  • I ex­pect QC de­vel­op­ments nei­ther hin­der nor help with bot­tle­necks in cur­rent re­search agen­das in AI al­ign­ment, and I see no need to cur­rently in­vest in ex­per­tise in this in­ter­sec­tion. More

  • QC com­pro­mises se­cure com­mu­ni­ca­tion and on­line trans­ac­tions, which could have profound eco­nomic effects. The scale of this prob­lem is un­clear. This is­sue seems to be re­ceiv­ing a com­men­su­rate amount of at­ten­tion from gov­ern­ment orgs (NIST) and in­dus­try (ETSI), and over­all tractable. More

  • QC could have an effect on com­puter hard­ware de­sign. I am un­sure of how promis­ing this re­search line is com­pared to other re­search lines in hard­ware de­sign. I’d be keen on see­ing a 10 hour re­search pro­ject re­flect­ing on this. More

  • Very ten­ta­tively, I’d guess that the de­vel­op­ment of QC would not sig­nifi­cantly in­crease bio risk nor it is a spe­cially promis­ing tool for miti­ga­tion. More

  • QC would likely not have a ma­jor effect on Atom­i­cally Pre­cise Man­u­fac­tur­ing. Cur­rent re­search in APM is able to ab­stract away from the quan­tum me­chan­ics of pre­cise pro­tein in­ter­ac­tions and thus able to use effi­cient clas­si­cal simu­la­tions for de­sign. More

  • QC has some promis­ing ap­pli­ca­tions with lit­tle ap­par­ent down­side in medicine, agri­cul­ture (fer­til­izer de­sign) and op­er­a­tions re­search. Big com­pa­nies like IBM, Google and Microsoft are aware of these ap­pli­ca­tions, so I would ex­pect lit­tle room-for-fund­ing. It is also un­clear whether there is lower hang­ing fruit for these ap­pli­ca­tions via other av­enues. More

  • Some re­searchers are pre­emp­tively ex­plor­ing QC simu­la­tions as a ba­sic re­search tool in physics and chem­istry, and re­search of QC limits as phys­i­cal limits. How­ever it is not clear to me that ba­sic re­search on physics is ur­gent, and I am also un­cer­tain about how use­ful QC tools would be in chem­istry re­search. An in­ter­view with an ex­pert in quan­tum chem­istry may clar­ify the is­sue. More

  • Some strate­gic and tractable re­search on gen­eral Quan­tum Com­put­ing that would be helpful to bet­ter un­der­stand its rele­vance in­cludes 1) fore­cast­ing QC timelines and 2) re­search­ing whether and how QC can be reg­u­lated. I think that a 100h re­search pro­ject on those ques­tions is worth fund­ing in­so­far it will al­low us to bet­ter un­der­stand the VOI of strate­gic re­search on con­crete ap­pli­ca­tions and reg­u­la­tion of QC. More

  • I would recom­mend to fund some more ex­plo­ra­tory re­search in QC grant­mak­ing as a high-risk high-re­ward re­search pro­ject. I would how­ever recom­mend against fund­ing ob­ject level re­search or pur­su­ing ca­reers in QC for philan­thropic rea­sons. More

In this sum­mary I have given guesses on what is worth fund­ing. Re­searchers may want to look into the sec­tions they are con­sid­er­ing in­ves­ti­gat­ing them­selves for open ques­tions.

This ar­ti­cle was writ­ten by Jaime Sevilla. This work was par­tially sup­ported by the Fu­ture of Hu­man­ity In­sti­tute sum­mer fel­low­ship pro­gram and partly by a grant made by the Effec­tive Altru­ism Foun­da­tion.

I want to thank Pablo Moreno for dis­cus­sion on Quan­tum Com­put­ing, Jassi Pannu and Gre­gory Lewis for dis­cus­sion on biolog­i­cal risks, Eric Drexler for dis­cus­sion on Atom­i­cally Pre­cise Man­u­fac­tur­ing, Max Daniel for men­tor­ship and Luisa Ro­driguez for gen­eral feed­back.

No comments.