Thomas Goetz

Why you shouldn’t fall in love with new tech­nol­ogy


Iand dis­tract­ing to our team. It’s a bit em­bar­rass­ing to ad­mit this, but I’m hardly alone. In Sil­i­con Val­ley, swoon­ing star­tups are about as com­mon as Tes­las. Here, if you’re not in a swoon, you’re not re­ally try­ing.

Ev­ery­one is talk­ing about vir­tual re­al­ity and aug­mented re­al­ity and ar­ti­fi­cial in­tel­li­gence and deep learn­ing (not us­ing those ac­tual words, but rather “VR” and “AR” and “AI” and “DL”). One ac­count­ing puts the num­ber of AI star­tups in health care, where I com­pete, at more than 100. AI is everywhere else, too. It seems if you don’t have a chat­bot strat­egy in 2017, you’re so flip phone. Why are so many star­tups in an AI swoon? Tim­ing. It’s ex­cep­tion­ally hard to get the tim­ing right, es­pe­cially for a small com­pany with a tick­ing clock and shrink­ing cap­i­tal. A cou­ple of years ago, all the cool kids were chas­ing VR—rais­ing money, build­ing pro­to­types, wait­ing for the bil­lions to roll in. To­day, many of those star­tups have gone into “cock­roach mode” as they wait for some bona fide con­sumer de­mand for their toys.

Tech progress usu­ally moves along a path from sci­ence to tech­nol­ogy to in­dus­try to cul­ture. It typ­i­cally starts with a lab dis­cov­ery: the semi­con­duc­tor, an al­go­rithm. From there, it gets turned into a tech­nol­ogy—a tool that can be tested and de­vel­oped. Next, it moves to in­dus­trial ap­pli­ca­tions, and fi­nally, once it has con­sumer util­ity, it reaches the cul­ture at large. Com­put­ers took 40 years to move from lab to home; ro­bot­ics, though, have so far had only in­dus­trial im­pact. (No, Roomba doesn’t count.)

In the­ory, it’s pos­si­ble for a startup to cap­i­tal­ize on an in­no­va­tion at any point along this arc, so long as you know what you’re gun­ning for. But for star­tups chas­ing this dragon, it’s just so easy to be out­flanked and out­matched. In our case, we so loved the idea of of­fer­ing photo iden­ti­fi­ca­tion that we didn’t re­ally think through the con­se­quences. Given our size, we could ei­ther build a con­sumer web­site or a deep-learn­ing com­pany, but not both. And if we wanted to do the lat­ter, we should’ve not only started with a dif­fer­ent team but also cho­sen dif­fer­ent in­vestors, a dif­fer­ent busi­ness model, and so on. Thank­fully, we put the com­puter-vi­sion project on hold— and then we shelved it al­to­gether.

Shak­ing off the swoon, we fo­cused on a proven tech­nol­ogy called a web­site. Much less sexy, but it turns out that’s where our mar­ket is: mil­lions of peo­ple who just want bet­ter in­for­ma­tion about their med­i­ca­tions, at the right time. A su­per­in­tel­li­gent, com­puter-vi­sion tech­nol­ogy with ma­chine-learn­ing ca­pa­bil­i­ties? Sounds like a great idea. It’s all yours. MAGINE YOU’RE 60 YEARS OLD and take sev­eral pre­scrip­tions ev­ery day. Blue pill, red pill, a cou­ple of white pills, each at a dif­fer­ent hour. Pa­tients, par­tic­u­larly older ones, of­ten fail to take their meds—it’s a se­ri­ous health is­sue. Wouldn’t it be amaz­ing to just snap a pic­ture and have each pill auto-mag­i­cally iden­ti­fied (to make sure you’re tak­ing the right one) and logged (to make sure you’re tak­ing it at the right time)?

This is an idea we had at my startup, Io­dine. More than an idea, in fact. We ac­tu­ally hired a com­put­er­vi­sion PhD and re­cruited a cou­ple of other en­gi­neers to make this a re­al­ity. We spent about a year try­ing to work this out. But we com­mit­ted the mis­take of be­ing too early. Though we made some progress—it’s a re­ally hard chal­lenge—we couldn’t get more than 80 per­cent of the way there. And in soft­ware, that last 20 per­cent takes 80 per­cent of your ef­fort.

We’d been caught in a swoon— en­rap­tured by a shiny new tech­nol­ogy with­out un­der­stand­ing that, for our firm of just eight or nine peo­ple, it was be­yond our ca­pac­ity, tan­gen­tial to our mis­sion,

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.