Buf­falo News se­lected for Google news dig­i­tal subscriptions lab

The Buffalo News - - BUSINESS - By David Robinson

The Buf­falo News is one of 10 news­pa­per pub­lish­ers picked to be part of a new pro­gram backed by Google aimed at help­ing lo­cal me­dia build on­line subscriptions.

The ini­tia­tive, called the Google News Ini­tia­tive Subscriptions Lab, is a six-month pro­gram aimed at help­ing those news­pa­pers re­vamp and im­prove their dig­i­tal sub­scrip­tion process.

With the news­pa­per in­dus­try grap­pling with de­clin­ing print rev­enue in the in­ter­net age, dig­i­tal subscriptions are viewed as a key el­e­ment for lo­cal me­dia com­pa­nies, both now and in the fu­ture.

“This ef­fort will give us ac­cess to these part­ners’ tremen­dous re­sources as well as in­sight from peers at other par­tic­i­pat­ing news­pa­pers,” said Buf­falo News Pub­lisher War­ren T. Colville. “I ex­pect it to be a sig­nif­i­cant boost to our work cre­at­ing a dig­i­tal­sub­scrip­tion busi­ness that sup­ports our jour­nal­ism for years to come.”

The pro­gram was de­vel­oped as a part­ner­ship be­tween the Google News Ini­tia­tive and the Lo­cal Me­dia As­so­ci­a­tion, an in­dus­try trade group. As part of the pro­gram, con­sul­tants from FTI Con­sult­ing will do an eval­u­a­tion of each par­tic­i­pat­ing pub­lisher’s dig­i­tal sub­scrip­tion ef­forts, while also con­duct­ing mar­ket re­search to help de­ter­mine the size and scope of the mar­ket and also the will­ing­ness of read­ers to pay.

The pub­lish­ers also will be able to tap into sup­port from Google teams that have ex­per­tise in ar­eas such as data, tech­nol­ogy, prod­uct and subscriptions.

“We know that reader rev­enue is our fu­ture,” said Brian J. Con­nolly, The News’ vice pres­i­dent for in­no­va­tion and busi­ness de­vel­op­ment. “That reader rev­enue is likely to come from dig­i­tal sub­scribers. This is go­ing to give us the re­sources to jump­start those ef­forts.”

The Google pro­gram in­cludes 10 pub­lish­ers of vary­ing sizes across the United States and Canada. The other pub­lish­ers, se­lected out of 21 that were in­vited to ap­ply, in­clude: El Nuevo Día, the Bal­ti­more Sun, the Colum­bus Dis­patch, the Hous­ton Chron­i­cle, the Idaho Press, the Port­land (Me.) Press Her­ald, the Post and Courier in Charleston, S.C., the South­east Mis­sourian and the Toronto Star.

The pro­gram is ex­pected to be

a French-born re­searcher, was a key fig­ure in the de­vel­op­ment of Google Brain, the com­pany’s cen­tral ar­ti­fi­cial in­tel­li­gence lab. His team re­cently moved into a new lab on Google’s main cam­pus in Moun­tain View.

While the ma­chines may not be as eye-catch­ing as hu­manoid ro­bots, Google re­searchers be­lieve that the sub­tly more ad­vanced tech­nol­ogy in­side them gives them more po­ten­tial in the real world. The com­pany is de­vel­op­ing ways for these ro­bots to learn skills on their own, like sort­ing through a bin of un­fa­mil­iar ob­jects or nav­i­gat­ing a ware­house filled with un­ex­pected ob­sta­cles.

Google’s new lab is in­dica­tive of a broader ef­fort to bring machine learn­ing to ro­bot­ics. Re­searchers are ex­plor­ing sim­i­lar tech­niques at places like the Univer­sity of Cal­i­for­nia, Berkeley, and OpenAI, the ar­ti­fi­cial in­tel­li­gence lab founded by the Silicon Val­ley king­pins Elon Musk and Sam Alt­man. In re­cent months, both places have spawned startups try­ing to com­mer­cial­ize their work.

Many be­lieve that machine learn­ing – not ex­trav­a­gant new de­vices – will be the key to de­vel­op­ing ro­bot­ics for man­u­fac­tur­ing, ware­house au­to­ma­tion, trans­porta­tion and many other tasks.

“Ro­bot­ics has long held the pop­u­lar imag­i­na­tion, but what is eas­ily the most im­por­tant change is the ap­pli­ca­tion of machine learn­ing,” said Su­nil Dhali­wal, a gen­eral part­ner with Am­plify Part­ners, a Silicon Val­ley ven­ture cap­i­tal firm. “The util­ity is in the soft­ware.”

Ro­bots are al­ready used in ware­houses and on fac­tory floors, but they can han­dle only spe­cific tasks, like pick­ing up a par­tic­u­lar ob­ject or turn­ing a screw. Google wants the ma­chines it is work­ing with to learn on their own.

On a re­cent af­ter­noon in­side Google’s new lab, a ro­botic arm hov­ered over a bin filled with Ping-Pong balls, wooden blocks, plas­tic bananas and other ran­dom ob­jects. Reach­ing into this pile of clut­ter, the arm grabbed a ba­nana be­tween two fin­gers and, with a gen­tle flick of the wrist, tossed it into a smaller bin sev­eral feet away.

For a ro­bot, it was a re­mark­able trick. When first pre­sented with that pile of clut­ter, the arm did not know how to pick up a sin­gle ob­ject. But equipped with a cam­era that looked down into the bin, the Google sys­tem an­a­lyzed its own progress dur­ing about 14 hours of trial and er­ror.

The arm even­tu­ally learned to toss items into the right bins about 85 per­cent of the time. When the re­searchers tried the same task, their ac­cu­racy rate was about 80 per­cent.

It may sound sim­ple enough, but writ­ing com­puter code to tell a machine how to do that would be ex­tremely dif­fi­cult. “It is learn­ing more com­pli­cated things than I could ever think about,” said Shu­ran Song, one the pri­mary re­searchers on the project.

Re­searchers be­lieve these ma­chines could work in ware­houses and dis­tri­bu­tion cen­ters run by com­pa­nies like Ama­zon and UPS. To­day, hu­mans sort through items that move in and out of dis­tri­bu­tion cen­ters. A sys­tem like Google’s could au­to­mate at least part of the process, though it is unclear when it will be ready for com­mer­cial use. Ama­zon, which has al­ready de­ployed other kinds of ro­bot­ics in its dis­tri­bu­tion cen­ters, is in­ter­ested in this kind of tech­nol­ogy.

But many ro­bot­ics ex­perts warn that mov­ing this kind of machine learn­ing into the real world will be dif­fi­cult. Tech­nol­ogy that does well in the lab of­ten breaks down in­side a dis­tri­bu­tion cen­ter be­cause it can’t deal with un­ex­pected ob­jects it hasn’t seen be­fore or tasks that re­quire movements it has never tried.

“This is not the right so­lu­tion for all prob­lems,” said Leif Jentoft, chief ex­ec­u­tive of the Mas­sachusetts com­pany Right­Hand Ro­bot­ics and a sea­soned ro­bot­ics re­searcher. “These technologies can some­times seem more pow­er­ful than they are.”

In an­other cor­ner of Google’s lab, re­searchers are train­ing ro­botic hands to ma­nip­u­late ob­jects – push, pull and spin them in sub­tle ways.

The three-fin­gered hands are hardly com­plex, at least in the phys­i­cal sense. The soft­ware help­ing them learn is the break­through, and re­searchers hope the hands can even­tu­ally learn to use tools and other equip­ment.

Google is tak­ing a sim­i­lar ap­proach with all its ro­botic hard­ware. The arm that tosses ob­jects into a bin is not an elab­o­rate machine de­signed by Google en­gi­neers. Built by Universal Ro­bots, it is com­monly used for man­u­fac­tur­ing and other tasks. Google is train­ing it to do things it couldn’t oth­er­wise do.

“Learn­ing is ac­tu­ally help­ing us over­come the chal­lenges of low-cost ro­bots,” said Vikash Ku­mar, the Google re­searcher who over­sees this project.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.