Australian Transport News - - Operations + Strategy | Artificial Intelligence -

Core tech­no­log­i­cal ad­vances are cen­tral to the con­tin­ued devel­op­ment of AI. Sig­nif­i­cant progress has been made with all core AI tech­nolo­gies, and the lev­els of in­vest­ment and de­mand for on­go­ing im­prove­ment give good rea­son to ex­pect this growth will con­tinue well into the fu­ture. Tech­no­log­i­cal ad­vances can be clas­si­fied into three broad cat­e­gories: im­prov­ing com­puter pro­cess­ing speed and power, in­creas­ing AI sys­tem ac­cess to big data, and us­ing al­go­rith­mic im­prove­ments to en­able more com­plex AI ap­pli­ca­tions.

• Com­put­ing Power &Speed: AI is a com­puter pro­cess­ing in­ten­sive tech­nol­ogy – break­throughs in com­put­ing power and ef­fi­ciency have en­abled the ex­pan­sion and com­plex­ity of AI ap­pli­ca­tions.

In the tech­nol­ogy in­dus­try, Moore’s Law is used to show the re­la­tion­ship be­tween the cost and speed of com­puter pro­cess­ing power over time, the tra­jec­tory of which re­sults in an ex­po­nen­tial curve.

Un­til re­cently, a com­put­ing de­vice’s CPU, or cen­tral pro­cess­ing unit, typ­i­cally pro­vided the core func­tion of pro­cess­ing.

In re­cent years, GPUs, or graph­i­cal pro­cess­ing units, have be­gun to par­tially take over com­puter pro­cess­ing work­loads, con­tribut­ing sig­nif­i­cantly to the rise of AI.

Orig­i­nally de­signed for the much larger and more com­plex com­pu­ta­tional work­loads of ren­der­ing com­puter gam­ing graph­ics in real time, GPUs are de­signed to han­dle hun­dreds of tasks in par­al­lel, and to­day are suc­cess­fully be­ing used to en­able AI ap­pli­ca­tions.

Ad­vances in com­puter chip tech­nol­ogy are an im­por­tant part of the AI de­vel­op­men­tal story. Given the con­sis­tency of chip im­prove­ments and the like­li­hood that chip de­sign will con­tinue to im­prove, this is not the pri­mary rea­son for the ex­is­tence of AI but just one of the es­sen­tial en­ablers.

• Big Data: The ex­is­tence of plen­ti­ful and eas­ily ac­ces­si­ble data is not a new phe­nom­e­non, how­ever its ever-in­creas­ing vol­ume, ve­loc­ity, and va­ri­ety is a key part of the AI story. Even though AI could ex­ist on a smaller scale with­out these ad­vances, AI re­quires data to demon­strate its full power. While new types of data have emerged in the past few years, and while there is a sig­nif­i­cant in­crease in the pace at which data is cre­ated and changes, AI sys­tems are cur­rently con­sum­ing only a tiny frac­tion of avail­able data. This has been true for a long time. So even if data quan­ti­ties were to stag­nate, and the rates of data vol­ume and ve­loc­ity were to re­main con­stant, AI would still have a lot of data to in­gest, con­tex­tu­alise, and un­der­stand.

• Al­go­rith­mic Im­prove­ments: The in­creas­ing abun­dance of data be­ing cre­ated ev­ery day has in­vited re­searchers, data sci­en­tists, and soft­ware engi­neers to con­cep­tu­alise so­phis­ti­cated new al­go­rithms ca­pa­ble of in­gest­ing large vol­umes of com­plex data.

Be­cause of this, to­day AI is not merely ca­pa­ble of han­dling the rapid as­sem­bly of large and quickly chang­ing datasets but, in fact, thrives on this. These big datasets make the best con­tri­bu­tion to AI’s abil­ity to learn when they are com­plex, so the more di­ver­sity in the data do­main the bet­ter. This is an ad­van­tage AI sys­tems have over other data-pro­cess­ing meth­ods: whereas stan­dard sys­tems get bogged down with large com­plex datasets, al­go­rith­mic im­prove­ments in re­cent years have im­proved sig­nif­i­cantly to be able to han­dle large vol­umes of het­ero­ge­neous data, en­abling the detection of pat­terns and dis­cov­ery of cor­re­la­tions that might not be ob­vi­ous to hu­mans or to stan­dard rule-based sys­tems. While these three tech­nol­ogy ad­vances are the main driv­ers of AI, con­sid­er­ing the fu­ture of AI re­veals not just a sin­gle trend, but the con­flu­ence of many un­der­ly­ing tech­nol­ogy trends.

Si­mul­ta­ne­ously, other im­por­tant tech­nol­ogy trends are de­vel­op­ing along a path that com­ple­ments AI, namely cloud com­put­ing and con­nec­tiv­ity. As cloud com­put­ing ad­vances to be­come a new in­dus­try stan­dard, it al­lows cen­tral­ized pro­cess­ing of large datasets. And as con­nec­tiv­ity (through the in­ter­net and cel­lu­lar net­works) in­creases, it en­ables trans­mis­sion and con­trol of large datasets in real time.

This means more and more datasets are be­ing stored, pro­cessed, and ac­cessed through the cloud, and con­nec­tiv­ity to that in­for­ma­tion no longer lim­its over­all sys­tem per­for­mance. As a re­sult, the ac­cel­er­at­ing ad­vance­ment of data stor­age, ac­ces­si­bil­ity, and trans­mis­sion speed is cat­alyz­ing the fur­ther devel­op­ment of AI.

Tech­no­log­i­cal fac­tors pro­vide es­sen­tial sup­port to the progress of AI, but their con­tri­bu­tion pales in com­par­i­son to the way that so­cial and com­mer­cial fac­tors in­flu­ence the vi­a­bil­ity of AI tech­nol­ogy.

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.