AI Impacts: Historic trends in technological progress

AI Im­pacts just re­leased a col­lec­tion of posts show­ing his­tor­i­cal progress in a va­ri­ety of tech­nolo­gies. Rather than cross-post­ing all of these ar­ti­cles in­di­vi­d­u­ally, I’ve col­lected them in this post.

The speed of de­liv­er­ing a short mes­sage across the At­lantic Ocean saw at least three dis­con­ti­nu­ities of more than ten years be­fore 1929, all of which also were more than one thou­sand years: a 1465-year dis­con­ti­nu­ity from Colum­bus’ sec­ond voy­age in 1493, a 2085-year dis­con­ti­nu­ity from the first tele­graph ca­ble in 1858, and then a 1335-year dis­con­ti­nu­ity from the sec­ond tele­graph ca­ble in 1866.

The speed at which a mil­i­tary pay­load could cross the At­lantic ocean con­tained six greater than 10-year dis­con­ti­nu­ities in 1493 and be­tween 1841 and 1957.

We mea­sure eight dis­con­ti­nu­ities of over ten years in the his­tory of longest bridge spans, four of them of over one hun­dred years, five of them ro­busts to slight changes in trend ex­trap­o­la­tion.

The an­nual av­er­age in­crease in bridge span length in­creased by over a fac­tor of one hun­dred be­tween the pe­riod be­fore 1826 and the pe­riod af­ter (0.25 feet/​year to 35 feet/​year), though there was not a clear turn­ing point in it.

Max­i­mum light in­ten­sity of ar­tifi­cial light sources has dis­con­tin­u­ously in­creased once that we know of: ar­gon flashes rep­re­sented roughly 1000 years of progress at past rates.

An­nual growth in light in­ten­sity in­creased from an av­er­age of roughly 0.4% per year be­tween 424BC and 1943 to an av­er­age of roughly 190% per year be­tween 1943 and the end of our data in 2008.

The num­ber of books pro­duced in the pre­vi­ous hun­dred years, sam­pled ev­ery hun­dred or fifty years be­tween 600AD to 1800AD con­tains five greater than 10-year dis­con­ti­nu­ities, four of them greater than 100 years. The last two fol­low the in­ven­tion of the print­ing press in 1492.

The real price of books dropped pre­cip­i­tously fol­low­ing the in­ven­tion of the print­ing press, but the longer term trend is suffi­ciently am­bigu­ous that this may not rep­re­sent a sub­stan­tial dis­con­ti­nu­ity.

The rate of progress of book pro­duc­tion changed shortly af­ter the in­ven­tion of the print­ing press, from a dou­bling time of 104 years to 43 years.

There do not ap­pear to have been any greater than 10-year dis­con­ti­nu­ities in telecom­mu­ni­ca­tions perfor­mance, mea­sured as:

  • band­width-dis­tance product for all tech­nolo­gies 1840-2015

  • band­width-dis­tance product for op­ti­cal fiber 1975-2000

  • to­tal band­width across the At­lantic 1956-2018

Ra­dio does not seem likely to have rep­re­sented a dis­con­ti­nu­ity in mes­sage speed.

Group in­dex of light ap­pears to have seen dis­con­ti­nu­ities of 22 years in 1995 from Co­her­ent Pop­u­la­tion Trap­ping (CPT) and 37 years in 1999 from EIT (con­den­sate). Pulse de­lay of light over a short dis­tance may have had a large dis­con­ti­nu­ity in 1994 but our data is not good enough to judge. After 1994, pulse de­lay does not ap­pear to have seen dis­con­ti­nu­ities of more than ten years.

Peni­cillin did not pre­cip­i­tate a dis­con­ti­nu­ity of more than ten years in deaths from syphilis in the US. Nor were there other dis­con­ti­nu­ities in that trend be­tween 1916 and 2015.

The num­ber of syphilis cases in the US also saw steep de­cline but no sub­stan­tial dis­con­ti­nu­ity be­tween 1941 and 2008.

On brief in­ves­ti­ga­tion, the effec­tive­ness of syphilis treat­ment and in­clu­sive costs of syphilis treat­ment do not ap­pear to have seen large dis­con­ti­nu­ities with peni­cillin, but we have not in­ves­ti­gated ei­ther thor­oughly enough to be con­fi­dent.

The max­i­mum su­per­con­duct­ing tem­per­a­ture of any ma­te­rial up to 1993 con­tained four greater than 10-year dis­con­ti­nu­ities: A 14-year dis­con­ti­nu­ity with NbN in 1941, a 26-year dis­con­ti­nu­ity with LaBaCuO4 in 1986, a 140-year dis­con­ti­nu­ity with YBa2Cu3O7 in 1987, and a 10-year dis­con­ti­nu­ity with BiCaSrCu2O9 in 1987.

YBa2Cu3O7 su­per­con­duc­tors seem to cor­re­spond to a marked change in the rate of progress of max­i­mum su­per­con­duct­ing tem­per­a­ture, from a rate of progress of .41 Kelvin per year to a rate of 5.7 Kelvin per year.

The Elo rat­ing of the best chess pro­gram mea­sured by the Swedish Chess Com­puter As­so­ci­a­tion did not con­tain any greater than 10-year dis­con­ti­nu­ities be­tween 1984 and 2018. A four year dis­con­ti­nu­ity in 2008 was no­table in the con­text of oth­er­wise reg­u­lar progress.

We es­ti­mate that Eli Whit­ney’s cot­ton gin rep­re­sented a 10 to 25 year dis­con­ti­nu­ity in pounds of cot­ton ginned per per­son per day, in 1793. Two in­no­va­tions in 1747 and 1788 look like dis­con­ti­nu­ities of over a thou­sand years each on this met­ric, but these could eas­ily stem from our ig­no­rance of such early de­vel­op­ments. We ten­ta­tively doubt that Whit­ney’s gin rep­re­sented a large dis­con­ti­nu­ity in the cost per value of cot­ton ginned, though it may have rep­re­sented a mod­er­ate one.

Flight air­speed records be­tween 1903 and 1976 con­tained one greater than 10-year dis­con­ti­nu­ity: a 19-year dis­con­ti­nu­ity cor­re­spond­ing to the Fairey Delta 2 flight in 1956.

The av­er­age an­nual growth in flight air­speed markedly in­creased with the Fairey Delta 2, from 16mph/​year to 129mph/​year.

We do not know if breech load­ing rifles rep­re­sented a dis­con­ti­nu­ity in mil­i­tary strength. They prob­a­bly did not rep­re­sent a dis­con­ti­nu­ity in fire rate.