Making Sense of AI: What Adobe Sensei Means for You

Arti­fi­cial intel­li­gence is trans­form­ing cre­ativ­i­ty and busi­ness — and Adobe Sen­sei is pro­vid­ing the tech­nol­o­gy to do it.

There’s an inevitabil­i­ty that comes with any “cut­ting-edge” tech­nol­o­gy — with­in a few years or, even, months, that once enve­lope-push­ing advance­ment is going to be seen as com­plete­ly com­mon­place. We’ve seen this for decades, espe­cial­ly when it comes to defin­ing and fram­ing arti­fi­cial intel­li­gence (AI).

As Wired’s found­ing edi­tor Kevin Kel­ley once explained, this isn’t a new phe­nom­e­non. In the 1960s, pro­gram­ming was con­sid­ered AI. Fast for­ward to the 1980s, and pro­gram­ming was the norm while data­bas­es and more mod­ernised approach­es to sta­tis­tics were con­sid­ered AI. Soon enough, though, these for­mer advances became par for the course — basic tech­no­log­i­cal appli­ca­tions that didn’t war­rant any sort of fan­fare, let alone the AI title. “Every achieve­ment in arti­fi­cial intel­li­gence rede­fines that suc­cess as ‘not AI,’” he wrote.

What artificial intelligence looks like today

AI today con­tin­ues to shift and evolve. How­ev­er, there does seem to be more of a pop­u­lar con­sen­sus when it comes to defin­ing present-day AI. Very sim­ply, most experts see AI as pro­gram­ming com­put­ers so they’re able to per­form tasks pre­vi­ous­ly done by humans. Under this umbrel­la, AI enables the com­put­er to “rea­son” so it can iden­ti­fy insights and make pre­dic­tions based on com­pli­cat­ed information.

At Adobe, we have a sim­i­lar anchor for AI, look­ing at advances with a three-part cri­te­ria that looks at the tech­nol­o­gy used, the val­ue it deliv­ers to cus­tomers, and the align­ment with our cor­po­rate val­ues. Sim­ply put —

Adobe Sen­sei deliv­ers the mag­ic of arti­fi­cial intel­li­gence and machine-learn­ing tech­nol­o­gy by enabling cus­tomers to:

• Dis­cov­er what is hidden,
• Accel­er­ate what is slow,
• Decide when it matters

while it serves the cre­ator and respects the consumer.

In-line with this def­i­n­i­tion, we intro­duced Adobe Sen­sei in Octo­ber 2016. Adobe Sen­sei brings togeth­er two unique Adobe capa­bil­i­ties — a mas­sive vol­ume of con­tent and data, plus deep under­stand­ing of how cus­tomers work.

This paired with the lat­est advance­ments in AI, machine learn­ing, deep learn­ing, and relat­ed fields enable cus­tomers to dis­cov­er what is hid­den, accel­er­ate what is slow, and decide when it mat­ters. All of this hap­pens while still align­ing to our cor­po­rate val­ues of serv­ing the cre­ator and respect­ing the con­sumer by check­ing these three boxes:

#1. Delivering artificial intelligence and machine learning at its core

Sci­ence fic­tion writer Arthur C. Clarke was clear­ly peer­ing into the future when he wrote, “Any suf­fi­cient­ly advanced tech­nol­o­gy is indis­tin­guish­able from magic.”

Ulti­mate­ly, we want to bring the mag­ic of tech­nol­o­gy to the user expe­ri­ence. To do this, Adobe Sen­sei inte­grates all branch­es of arti­fi­cial intel­li­gence, includ­ing machine learn­ing or deep learn­ing. This enables com­plex prob­lems and process­es to be bro­ken down into much sim­pler terms so bet­ter solu­tions can be eval­u­at­ed and act­ed on.

Giv­en Adobe Sensei’s scale and scope, this tech­nol­o­gy does tasks that seem unre­al — mag­ic, almost. From pre­dict­ing cus­tomer churn to adding a sun­set to the back­ground image of a pho­to, Adobe Sen­sei lever­ages machine learn­ing, algo­rithms, and out­puts to improve process­es with­out requir­ing a human pro­gram­mer inter­vene. The more data received, the bet­ter the performance.

#2. Promoting customer-first value

Adobe Research is tasked with imag­in­ing and invent­ing the future, with an eye on exper­i­men­ta­tion and col­lab­o­ra­tion. Last year, the team pub­lished more than 170 tech­ni­cal papers, filed for more than 130 patents, and trans­ferred near­ly 60 new tech­nolo­gies to exist­ing Adobe products.

With every­thing we do, we’re laser-focused on adding “mag­ic” to our col­lec­tive tool­box, ensur­ing we can cre­ate a bet­ter, more sophis­ti­cat­ed, and more high-val­ue future for our cus­tomers and their cus­tomers. Late­ly this has includ­ed AI and machine-learn­ing capa­bil­i­ties, which help our cus­tomers deliv­er cus­tomer-first expe­ri­ences with greater speed and rel­e­vance than ever before.

That said, not all AI syncs with Adobe Sensei’s goals and val­ue-cen­tric approach. For us to incor­po­rate new AI and machine-learn­ing exten­sions, they must pro­vide clear-cut cus­tomer val­ue in at least one of three ways:

Discover what is hidden

A 2018 Gallup sur­vey found that, despite liv­ing in a peri­od of infor­ma­tion over­load, the explo­sion of infor­ma­tion makes most peo­ple feel it is more dif­fi­cult to stay “in the know.” In the same way, mar­keters have access to so much infor­ma­tion about cus­tomers that it can be hard­er to under­stand those cus­tomers — it’s dif­fi­cult to sep­a­rate the impor­tant find­ings from the noise.

Adobe Sen­sei can help cre­ate bet­ter, more effec­tive, and engag­ing expe­ri­ences by dis­cov­er­ing the hid­den jew­els in the great mounds of data. Anom­aly Detec­tion in Adobe Ana­lyt­ics, for exam­ple, uses machine learn­ing to assess and ana­lyze cus­tomer behav­ior. By mon­i­tor­ing past behav­ior and com­par­ing find­ings to real-time pat­terns, Anom­aly Detec­tion can spot behav­iour­al out­liers, sep­a­rat­ing the norm from “the noise.”

Accelerate what is slow

Got an idea for an image for a mar­ket­ing cam­paign — an idea that’s so pow­er­ful you can lit­er­al­ly see it in your head? Find­ing the right asset can be a time-con­sum­ing and, often, tedious chore.

To sim­pli­fy the process, Adobe Sen­sei pow­ers the Smart Tags fea­ture in Adobe Expe­ri­ence Man­ag­er. By auto­mat­i­cal­ly adding meta­da­ta tags to images, Smart Tags make it easy to find the right images even with­in bulk files.

That’s just the begin­ning. To help increase effi­cien­cy and effec­tive­ness, Adobe Sensei’s “smart crop­ping” fea­ture auto­mat­i­cal­ly crops images to spe­cif­ic dimen­sions while keep­ing the salient object in the picture.

These are both exam­ples of activ­i­ties humans could do man­u­al­ly with con­sid­er­able time and effort. What’s more, these tasks are chal­leng­ing if not impos­si­ble to do at the scale when nec­es­sary. Adobe Sen­sei tack­les these smart process­es — and takes these crit­i­cal steps to get the job done — automatically.

Decide when it matters

Adobe Sen­sei makes deci­sions at a speed and scale that are oth­er­wise impos­si­ble to do man­u­al­ly. By turn­ing over essen­tial tasks like Auto­mat­ed Per­son­al­iza­tion in Adobe Tar­get, Adobe Sen­sei can analyse pat­terns and end results, opti­mis­ing expe­ri­ences for every cus­tomer and seg­ment. In look­ing at every­thing — real-time deci­sions, cus­tomer traits, tem­po­ral con­sid­er­a­tions, and more — Adobe Sen­sei can opti­mise expe­ri­ences in the moment and over time.

#3. Aligning with Adobe’s core values and customer-first commitment

From an AI per­spec­tive, our over­rid­ing phi­los­o­phy is to “serve the cre­ator and respect the con­sumer” — and, lever­aged prop­er­ly, it does:

• Think about those repet­i­tive tasks that can now be han­dled by a machine mar­ket­ing partner.
• Think about the enhanced data capa­bil­i­ties and the bet­ter cus­tomer expe­ri­ences now available.
• Think about the improved
• human-machine inter­ac­tions
• that enable greater productivity.

A good exam­ple is Pre­dic­tive Fatigue Man­age­ment, an upcom­ing fea­ture in Adobe Cam­paign, which applies email fre­quen­cy con­trols to lim­it the num­ber of emails a cus­tomer receives. Reduc­ing the num­ber of unwant­ed emails serves both the cus­tomer and mar­keter, elim­i­nat­ing irri­ta­tion for the recip­i­ents while boost­ing the con­ver­sion rate. Reduc­ing spam merges busi­ness ben­e­fits with data ethics.

And that inter­sec­tion is at the heart of how we view arti­fi­cial intel­li­gence. Used effec­tive­ly, it can pro­vide huge pro­duc­tiv­i­ty ben­e­fits for cre­ative and mar­keters, giv­ing them the time and head­space to find the next big idea.

Like any pow­er­ful tech­nol­o­gy, though, ours must be used respon­si­bly. We at Adobe remain com­mit­ted to serv­ing the needs of cus­tomers and con­sumers in an eth­i­cal man­ner, in align­ment with our own core values.

Adobe Sen­sei is a pow­er­ful tool — and, like any pow­er­ful tech­nol­o­gy, we must always be mind­ful of the impli­ca­tions and keen­ly aware of how they impact the cus­tomer. For this rea­son, we’ve formed a work­ing group to exam­ine the impact of tech­nol­o­gy, eval­u­at­ing and cre­at­ing best prac­tices. This scope includes data pri­va­cy, data gov­er­nance, and data diver­si­ty — mak­ing sure data used to devel­op AI-dri­ven algo­rithms doesn’t inad­ver­tent­ly pro­duce biased algorithms.

The Adobe Research team recent­ly trained a deep learn­ing neur­al net­work to iden­ti­fy image manip­u­la­tion — some­thing foren­sic experts used to devote hours to, but now AI tech­nol­o­gy can iden­ti­fy in sec­onds. This, among oth­er strate­gies, can help curb every­thing from image tam­per­ing to image noise, ensur­ing greater authen­tic­i­ty in cre­ative and content.

“It’s impor­tant to devel­op tech­nol­o­gy respon­si­bly, but ulti­mate­ly these tech­nolo­gies are cre­at­ed in ser­vice to soci­ety,” says Jon Brandt, senior prin­ci­pal sci­en­tist and direc­tor for Adobe Research. “Con­se­quent­ly, we all share the respon­si­bil­i­ty to address poten­tial neg­a­tive impacts of new tech­nolo­gies through changes to our social insti­tu­tions and conventions.”

Read more sto­ries from our arti­fi­cial intel­li­gence series by vis­it­ing this page, and for more infor­ma­tion about Adobe Sen­sei, click here.