[Question] What are CAIS’ boldest near/​medium-term predictions?

Back­ground and questions

Since Eric Drexler pub­li­cly re­leased his “Com­pre­hen­sive AI ser­vices model” (CAIS) there has been a se­ries of analy­ses on LW, from ro­hin­mshah, ricraz, PeterMcCluskey, and oth­ers.

Much of this dis­cus­sion fo­cuses on the im­pli­ca­tions of this model for safety strat­egy and re­source al­lo­ca­tion. In this ques­tion I want to fo­cus on the em­piri­cal part of the model.

  • What are the bold­est pre­dic­tions the CAIS model makes about what the world will look in <=10 years?

“Bold­est” might be in­ter­preted as those pre­dic­tions which CAIS gives a de­cent chance, but which have the low­est prob­a­bil­ity un­der other “wor­ld­views” such as the Bostrom/​Yud­kowsky paradigm.

A pre­dic­tion which all these wor­ld­views agree on, but which is nonethe­less quite bold, is less in­ter­est­ing for pre­sent pur­poses (pos­si­bly some­thing like that we will see faster progress than places like main­stream academia ex­pect).

Some other re­lated ques­tions:

  • If you dis­agree with Drexler, but ex­pect there to be em­piri­cal ev­i­dence within the next 1-10 years that would change your mind, what is it?

  • If you ex­pect there to be events in that timeframe caus­ing you to go “I told you so, the world sure doesn’t look like CAIS”, what are they?

Clar­ifi­ca­tions and suggestions

I should clar­ify that an­swers can be about things that would change your mind about whether CAIS is safer than other ap­proaches (see e.g. the Wei_Dai com­ment linked be­low).

But I sug­gest avoid­ing dis­cus­sion of cruxes which are more the­o­ret­i­cal than em­piri­cal (e.g. how de­com­pos­able high-level tasks are) un­less you have a neat op­er­a­tional­i­sa­tion for mak­ing them em­piri­cal (e.g. whether there will be ev­i­dence of large economies-of-scope of the most prof­itable au­toma­tion ser­vices).

Also, it might be re­ally hard to get this down to a sin­gle pre­dic­tion, so it might be use­ful to pose a cluster of pre­dic­tions and differ­ent op­er­a­tional­i­sa­tions, and/​or us­ing con­di­tional pre­dic­tions.