The ma­chines may not win, af­ter all

Cosmos - - Spectrum -

IT IS POS­SI­BLE TO ar­gue that peo­ple in­volved in the minu­tiae of a par­tic­u­lar field are among the least likely to pro­vide a co­gent anal­y­sis of its broader im­pli­ca­tions.

A me­chan­i­cal en­gi­neer whose day job in­volves seek­ing in­cre­men­tal im­prove­ments in high-per­for­mance mo­tor­cars, for in­stance, might not be a cred­i­ble source re­gard­ing chal­lenges to trans­port in­fra­struc­ture.

This po­ten­tial for macro­cos­mic opac­ity to arise from mi­cro­cos­mic clar­ity is nowhere more present than within the area of ar­ti­fi­cial in­tel­li­gence. Thus, when a book writ­ten by an em­i­nent pro­fes­sor in the field cites in­ac­cu­rate in­for­ma­tion about AI re­layed by “peo­ple, es­pe­cially those out­side the field”, and in­forms the reader, via foot­note, that the book “is a mo­ment where I look up from my work”, fears of a text too nar­rowly fo­cused seem rea­son­able.

Thank­fully, how­ever, in the case of It’s Alive by Toby Walsh, pro­fes­sor of AI at the Univer­sity of New South Wales, such con­cerns are quickly dis­pelled.

Walsh de­liv­ers an ac­ces­si­ble and con­cise his­tory of ar­ti­fi­cial in­tel­li­gence, with Alan Tur­ing as its ful­crum but stretch­ing back to Got­tfried Leib­niz. He fol­lows with a longer ex­plo­ration of the cur­rent state of play within the dis­ci­pline, and then ends with a se­ries of cau­tious pre­dic­tions about the near and more dis­tant fu­ture.

Three things need to be said by way of global ob­ser­va­tions here. First, Walsh delves into the tech­ni­cal­i­ties of his world only suf­fi­ciently to pro­vide the de­tail his read­ers re­quire to com­pre­hend his points.

Sec­ond, he is adept at care­fully tread­ing a mid­dle path, steer­ing clear of the tech­noutopi­anism that glosses over the man­i­fest chal­lenges pre­sented by AI, on one hand, and the doom-laden keen­ing of pes­simistic fu­tur­ists on the other. Third, Walsh’s abil­ity to achieve what he sets out to do with the first two points means that this book, with text cov­er­ing just 294 pages, could eas­ily have been much longer with­out ei­ther baf­fling or bor­ing the reader.

In some ways Walsh is at his most in­ter­est­ing when his fo­cus is at its widest. In this mode, he pro­vides in­for­ma­tive in­sight into some AI de­vel­op­ments rolling out now, and oth­ers that re­main merely prom­ises – or night­mares, de­pend­ing on your per­spec­tive.

When ad­dress­ing the ad­vent of au­ton­o­mous ve­hi­cles, he is re­fresh­ingly free of the gung-ho free-mar­ket de­ter­min­ism that runs through much of the dis­course sur­round­ing the brave new world of ar­ti­fi­cial in­tel­li­gence and ro­bot­ics.

Rather than see­ing the de­vel­op­ment of driver­less taxis and robo-trucks as an ex­cit­ing and in­evitable con­se­quence of Sil­i­con Val­ley lais­sez faire, he calls for a strong reg­u­la­tory role for gov­ern­ment.

“We do not let drug com­pa­nies test their prod­ucts freely on the gen­eral pub­lic,” he states. “Sim­i­larly, we should not let tech­nol­ogy com­pa­nies test au­ton­o­mous cars freely on pub­lic roads with­out strong over­sight.”

(He is even sterner when it comes to au­ton­o­mous weapons, and in 2015 helped draft an open let­ter call­ing for a per­ma­nent ban on their de­vel­op­ment. To­day that let­ter con­tains 20,000 sig­na­tures, in­clud­ing those of Stephen Hawk­ing, Elon Musk and Noam Chom­sky.)

One of the com­pa­nies lung­ing to­wards an au­ton­o­mous fu­ture is the crowd-shar­ing ride ser­vice Uber. Be­ing an Uber driver, com­ments Walsh dryly, is one of the world’s new­est jobs, and is also very likely to be one of the short­est-lived.

The wide­spread loss of em­ploy­ment – and whole em­ploy­ment cat­e­gories – is a sub­ject fre­quently raised by technopes­simists, such as the fu­tur­ist Noah Yu­val Harari. There are good rea­sons for this, given pretty much ev­ery­one in the field agrees job losses will be mas­sive. Walsh, how­ever, at­tempts to bring to calm to the sub­ject by re­view­ing the find­ings of a key 2013 Ox­ford Univer­sity re­port that broke down likely re­dun­dan­cies sec­tor by sec­tor.

He has good and bad news. Cre­atives, such as nov­el­ists, will be largely safe, he says. The pre­dic­tion that bi­cy­cle re­pair will be al­most com­pletely au­to­mated is “rub­bish”, be­cause bi­cy­cles are fid­dly, bike

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.