You are on page 1of 38

1

LeL me begln by Lelllng you a blL abouL my background. lm a user experlence deslgner. l
was one of Lhe rsL professlonal Web deslgners. l've now been deslgnlng Web slLes for
nearly 20 years. 1hls ls Lhe navlgauon for a hoL sauce shopplng slLe l deslgned ln 1994.
2
l'm proud of Lhe facL LhaL 16 years laLer Lhey were sull uslng Lhe same vlsual ldenuLy. 1hese
were some of Lhe oldesL plxels on Lhe Web.

l had Lhe prlvllege of deslgnlng Lhe ul of some of Lhe rsL ecommerce slLes, onllne chaL
sysLems, and search englnes. l also goL Lo waLch Lwo doLcom crashes and saw many of my
cllenLs fall, wlLh years of Lhe work dlsappearlng nearly overnlghL
3
Slnce Lhen l've consulLed on Lhe user experlence deslgn of dozens, maybe hundreds of web
slLes. Pere's one for credlL.com, who were fanLasuc cllenLs a couple of years ago.
4
l've also worked on Lhe user experlence deslgn of a loL of consumer elecLronlcs producLs
from companles you've probably heard of.
5
l've wrluen a couple of books based on my experlence as a deslgner. Cne ls a cookbook of
user research meLhods, and Lhe second descrlbes whaL l Lhlnk are some of Lhe core
concerns when deslgnlng neLworked compuLauonal devlces. Much of Lhls Lalk ls based ln
work ln LhaL second book.
6
7
l also sLarLed a couple of companles. 1he rsL, Adapuve aLh, was prlmarlly focused on Lhe
web, buL wlLh Lhe second one, 1hlngM, l goL deep lnLo developlng hardware.
1hlngM prlmarlly deslgns and manufacLure smarL LLus for archlLecLs, lndusLrlal deslgners
and hackers,
8
8uL we also spenL some ume explorlng Lhe poLenual uses of embedded compuung as a
deslgn maLerlal for Lhlngs LhaL could be classled ln Lhe lnLerneL of 1hlngs.

1hls ls an 8llu wlne rack LhaL's faceLed classlcauon browser Lo a cloud-based servlce. lor
example, lL can show you all Lhe 2003 napa CaberneL Sauvlgnons whlch are ready Lo drlnk
and whose currenL markeL prlce ls under $30. 1hls ls a capacluve senslng klLchen cablneL
knob LhaL glows when you Louch lL. lL's an explorauon ln maklng a dlglLal producL LhaL
would sull be useful 20 years aer lL was made.

9
LasL year l goL Lo work exLenslvely wlLh Sleo's on Lhe ux of Lhelr game plauorm, whlch
was LoLally fun. 1hanks, uave!
10
A couple of monLhs ago l [olned A8C, Lhe famous research lab, full ume as a prlnclpal ln lLs
lnnovauon Servlces group.
11
A loL of my pro[ecLs broadly fall under Lhe descrlpuon of Lhe lnLerneL of 1hlngs, buL LhaL's a
really challenglng name Lo work wlLh.

1alklng abouL 1he lnLerneL of 1hlngs ls a hard because Lhere are so many dlerenL
denluons. 1hls ls 1lme Magazlne's lllusLrauon of Lhe lnLerneL of 1hlngs for Lhelr 8esL
lnvenuons of 2008" edluon. l love Lhls lllusLrauon ls because lL makes no sense no mauer
how you Lhlnk abouL lL, whlch ls acLually qulLe an accuraLe represenLauon of how confuslng
Lhe many denluons of Lhe lnLerneL of 1hlngs are rlghL now.
12
LeL me glve you my denluon, whlch ls preuy broad. lor me Lhe lnLerneL of 1hlngs ls Lhe
comblnauon of dlsLrlbuLed lnformauon processlng, pervaslve wlreless neLworklng and
auLomauc ldenucauon, deployed lnexpenslvely and wldely. 1he underlylng Lechnologles
and Lhe appllcauons LhaL are Lradluonally dlscussed don'L mauer much, because lL ls Lhls
comblnauon of facLors LhaL deeply aecLs people and lndusLrles, and lL does lL by
connecung people's lmmedlaLe experlences Lo Lhe power of dlglLally aggregaLed and
analyzed lnformauon. ln oLher words, Lhe lnLerneL Lo 1hlngs Lurns physlcal acuons lnLo
knowledge ln Lhe cloud and knowledge ln Lhe cloud lnLo physlcal acuon ln a way LhaL's
never exlsLed before.
13
So, for example, l counL Lhe ledLx Sensaware smarL Lag and Lhe ?ouamark Lracklng sysLem
Lo be roughly ldenucal. 1he Sensaware Lag has a bunch of sensors, a CS and Lhe
equlvalenL of a phone ln lL. lL's used Lo Lrack hlgh value lLems, such as human organs, LhaL
need Lo be shlpped under preclsely malnLalned condluons. 1he ?ouamark sysLem uses
suckers, readers and a wlred neLwork servlce Lo Lrack Lhlngs such as produce and car parLs.
1echnologlcally Lhere's almosL no overlap, buL Lhey boLh glve people Lhe ablllLy Lo LreaL
physlcal ob[ecLs llke Lhey have been Lreaung daLa packeLs. 1hey brlng Lhe power and ldeas
of Lhe lnLerneL Lo physlcal Lhlngs.
14
ln oLher words, alLhough lL's dlscussed as an emerglng Lechnology, l belleve LhaL 1he
lnLerneL of 1hlngs ls acLually a comblnauon of maLure Lechnologles, much more maLure
Lhan people glve lL credlL for. 1he reason lL's cllmblng Lhe CarLner Pype Cycle ls because
Lhose maLure Lechnologles are now cheap, as Lhe rlse of smarLphones pushed Lhe prlce of
componenLs down and made people more aware whaL happens when you Lake a small blL
of funcuonallLy, whlch ls whaL an app ls, and dlsLrlbuLe lL Lhrough Lhe world. l belleve LhaL
people are looklng aL apps and Lhlnklng Lo Lhemselves Why do l need LhaL expenslve
phone, wlLh all lLs capablllues, Lo do Lhls one Lhlng? Why can'L l [usL Lake LhaL app, pluck lL
o Lhe screen, and puL lL lnLo a dedlcaLed plece of hardware LhaL only does LhaL one
Lhlng?"
13
Powever, lf you look aL whaL appllcauons are currenLly glven as examples of Lhe lnLerneL of
1hlngs, you'll see LhaL Lhey're mosLly Lop-down large-scale cenLrallzed lnfrasLrucLural
appllcauons. Pere's San lranclsco's parklng sysLem. lL uses sensors ln Lhe sLreeL Lo see
whaL spaces cars are parked ln. lL can Lell you where Lhere are empLy parklng spoLs and can
dlspaLch meLer malds Lo wrlLe uckeLs more emclenLly.

8uL Lhese pro[ecLs are noL Lhe ones LhaL l belleve wlll have Lhe greaLesL lmpacL on Lhe
world, nor where Lhe greaLesL lnnovauon wlll lle. l belleve LhaL Lhe greaLesL lnLerneL of
1hlngs lnnovauon, and Lhe deepesL lmpacL, wlll come from small, rlsky pro[ecLs underLaken
by enLrepreneurs worklng wlLh exlsung lnfrasLrucLures.
16
And l belleve LhaL Lhls wlll happen as people brlng onllne servlces lnLo Lhe physlcal world as
speclallzed devlces. LeL me show you a couple of Lrends l see, and how l Lhlnk Lhey wlll
lnLeracL Lo brlng Lhls abouL.

1he rsL ls a shl from generlc devlces and soware Lo speclallzed devlces and soware.
When compuung was expenslve, you had general purpose devlces and general purpose
soware LhaL had deal wlLh almosL every slLuauon. 1hls necesslLaLed deslgn compromlses
LhaL resulLed ln devlces and soware LhaL could do almosL everyLhlng, buL dld none of lL
well.

now LhaL processlng ls so cheap, you can have a comblnauon of 10, 20, or 30 compuung
devlces and apps for Lhe prlce of LhaL one devlce, and you can acqulre new funcuonallLy as
needed. 1hls means LhaL every devlce and soware package can have a narrower purpose.
17
1he second Lrend ls LhaL much of Lhe value of compuung ls now remoLe. 1here's no longer
a need Lo pack everyLhlng lnLo a slngle plece of soware, and Lhere's no expecLauon LhaL
everyLhlng wlll be Lhere.
18
lf we charL Lhese lasL Lwo Lends, Lwo broad classes of dlglLal producLs emerge.

lf we follow Lhe local Lo remoLe axls, we nd general-purpose devlces LhaL do roughly Lhe
same seL of Lhlngs, and dler prlmarlly ln slze. 1hey exlsL Lo provlde access Lo onllne
servlces, ln a form facLor LhaL's approprlaLe Lo Lhe conLexL ln whlch Lhey're used.
l call Lhese devlces Lermlnals.

lf we follow Lhe general Lo speclc axls, we see a shl ls Lo more narrow-funcuon devlces
LhaL are deslgned Lo do a small, speclc seL of Lhlngs really well. 1hey prlmarlly dler ln
whaL Lhose speclc Lhlngs are. l call Lhese devlces appllances.
19
1he mosL lnLeresung Lhlng, however, Lo me ls LhaL Lhese shls are parL of an even larger
Lransluon, one where devlces are slmulLaneously speclc Anu deeply ued Lo onllne
servlces. ln Lhls model, Lhe servlce provldes Lhe ma[orlLy of Lhe value, and can be
represenLed elLher as a dedlcaLed appllance, an app runnlng on a Lermlnal, or anyLhlng ln
beLween.

l call Lhese devlces servlce avaLars. And Lhls ls where Lhe lnLerneL of 1hlngs lles, aL leasL for
consumer producLs.
20
As value shls Lo servlces, Lhe devlces, soware appllcauons and webslLes used Lo access lL
-lLs avaLars-become secondary. A camera becomes a really good appllance for Laklng
phoLos for lllckr, whlle a 1v becomes a nlce lllckr dlsplay LhaL you don'L have Lo log lnLo
every ume, and a phone becomes a convenlenL way Lo Lake your lllckr plcLures on Lhe
road.

Pardware becomes slmulLaneously more speclallzed and devalued as users see Lhrough"
each devlce Lo Lhe servlce lL represenLs.
21
My favorlLe example of Lhls ls neullx. ?ou can geL neullx on vlrLually any Lermlnal LhaL has
a screen and a neLwork connecuon. ?ou can pause a neullx movle on one Lermlnal and
Lhen upause lL on anoLher.
22
23
8ecause Lo Lhe neullx cusLomer, any devlce used Lo waLch a movle on neullx ls [usL a hole
ln space Lo Lhe neullx servlce. lL's a shorL-Lerm manlfesLauon of a slngle servlce. 1he value,
Lhe brand loyalLy, and Lhe focus ls on Lhe servlce, noL Lhe frame around lL. 1he Lechnology
exlsLs Lo enable Lhe servlce, noL as an end Lo lLself.
neullx appllances are creaLed for a slngle reason: Lo make lL easler Lo access neullx. 1haL's
whaL 8oku does. lL Lurns every Lermlnal LhaL's noL already neullx enabled lnLo a neullx
Lermlnal.
24
This to me is why an Internet of Things is primarily based around services in the
cloud, but services that talk to hardware devices as one of their many avatars. This is
already happening.

Let me use Withings as an example. If you dont know what this is, google it on your
pocket terminal.
23
WlLhlngs aL rsL lL was klnd of a glmmlck. ?ou can LweeL your welghL Lo your frlends!" was
one of Lhe ways lL was orlglnally plLched. 1haL's of course noL parucularly lnLeresung, buL
LhaL was noL Lhe purpose of Lhe devlce. 1he devlce ls Lhe avaLar Lo a healLh servlce. 1he
scale ls Lhe way Lhe servlce dlerenuaLes lLself from oLher healLh Lracklng servlces, buL Lhe
value ls noL ln Lhe scale, buL ln Lhe servlce, whlch ls fully experlenced uslng oLher avaLars,
such as Lhe ones deplcLed on Lhe rlghL.
26
WlLhlngs has now expanded Lhe servlce Lo lnclude a blood pressure cu and acuvlLy
Lracker. Agaln, Lhe value ls noL ln Lhe devlces, buL ln Lhe knowledge LhaL Lhey creaLe by
collecung slmple pleces of lnformauon and Lhen provldlng users wlLh Lhe full power of
cloud-based servlces Lo make use of LhaL plece of lnformauon. WlLhlngs can keep addlng
avaLars, new sensors and new ways Lo dlsplay Lhe lnformauon Lhe sensors collecL, wlLhouL
fundamenLally changlng Lhe promlse of Lhe servlce.
27
1here ls a whole class of such devlces LhaL are essenually pro[ecuons of a cloud servlce
Lhrough a llmlLed funcuonallLy hardware producL. 1he quanued self folks love Lhem.
1hese are of course sensor-based devlces, buL whaL Lhey're selllng ls noL Lhe capablllues of
Lhe sensor, buL of Lhe cloud-based servlce Lhe sensor connecLs Lo.

28
vlLallLy Clowcap ls one of Lhe mosL maLure of Lhese servlces wlLh speclallzed avaLars. lL's
now been around for abouL ve years. 1hls ls a wlreless neLwork-connecLed plll boule
LhaL's an avaLar Lo vlLallLy's servlce for lncreaslng compllance Lo medlclne prescrlpuons.
When you close Lhe cap, lL sends a packeL of lnformauon Lhrough a moblle phone-based
base sLauon Lo a cenLral server and lL sLarLs counung down Lo when you nexL need Lo Lake
your medlclne. When lL's ume, lL llghLs up Lhe LLu on Lhe Lop of Lhe boule.

Powever, Lhe real power ls ln Lhe packeL of daLa lL sends.
29
1haL packeL leLs you geL a progress reporL LhaL can be used lnLeracuvely or dellvered by
emall. lL's llke Coogle Analyucs for your medlclne.
30
PealLh care pracuuoners geL yeL anoLher avaLar LhaL glves Lhem long-Lerm and longlLudlnal
analyucs abouL compllance across medlcauons and ume.

1o me, Lhls klnd of conversauon beLween devlces and neL servlces ls where Lhe real power
of 1he lnLerneL of 1hlngs beglns.

31
Creaung servlces llke Lhls ls becomlng lncreaslngly sLralghuorward. now we're seelng
Lechnologles LhaL make lL slmllarly easy Lo add speclallzed hardware devlces Lo servlces.

l grabbed Lhls lmage from ArrayenL, who ls a company LhaL makes a llule hardware blob
LhaL connecLs vlrLually anyLhlng, ln Lhls case a smoke deLecLor, Lo Lhelr cloud servlce. lL can
make any devlce look llke a Web slLe, and Lhere are oLher devlces llke lL on Lhe markeL. lf
you compare Lhls componenL Lo Lhe klnd of Lools LhaL made Web 2.0 posslble-8uby on
8alls, server vlrLuallzauon, eLc-you can see some key quallues LhaL made LhaL movemenL
successful: lL's noL LhaL Lhe core Lechnology ls novel, buL LhaL lL's been made cheap and you
don'L need Lo know how lL works Lo make a loL of problems dlsappear.

Source: ArrayenL
32
Servlces such as Cosm/achube, sen.se, 1hlngspeak, LvryLhng and Axeda are now servlng a
slmllar role by acung as daLa brokerages LhaL make arblLrary, dlerenL devlces acL
conslsLenLly. achube, for example, allows an arblLrary daLa sLream from any neL
connecLed devlce Lo share LhaL sLream wlLh any oLher devlce. 1he servlce wlll do Lhe
buerlng, Lhe proLocol Lranslauon, Lhe analyucs, everyLhlng. lL's a sysLem LhaL has lLs rooLs
ln Web mashups, now connecLed Lo hardware.

Connecung devlces Lo Lhe cloud allows for rapld lLerauon on feaLures, slnce mosL of Lhe
funcuonallLy of Lhose devlces lles ln Lhe cloud.
33
A8C's conLrlbuuon Lo Lhls ls a seL of Lechnologles focused on maklng elecLronlc devlces
cheaper, maklng Lhem communlcaLe ln more emclenL ways, creaung manufacLurlng Lools
LhaL make lL easler for people Lo make Lhem, and analyzlng all of Lhe daLa LhaL Lhey
produce.
34
8uL l wanL Lo leave you wlLh anoLher ldea. l Lhlnk LhaL Lhere's a fundamenLally dlerenL
way we need Lo Lhlnk abouL how Lo deslgn experlences LhaL lnvolved dozens or hundreds
of devlces. 1hls ls from 1lmo Arnall's greaL shorL lm, Wlreless ln Lhe World".
33
Cur currenL lnLerneL of Lhlngs Lools are rlgld and deLermlnlsuc. 1hese Lools work on one
devlce, or a couple, buL Lhls rlgld command and conLrol approach falls when you move up
Lo many devlces. l Lhlnk LhaL Lhe Lools we need are ones LhaL wlll help us dynamlcally shape
Lhe behavlor of many devlces acung slmulLaneously, ln real ume, worklng probablllsucally,
raLher Lhan Lhrough lncreaslng hlerarchles.


lmage from THE INTERNET OF THINGS: From RFID to the Next-Generation Pervasive
Networked Systems (Lu Yan et al, 2008)

36
l Lhlnk LhaL Lhe Lransformauonal polnL for Lhe lnLerneL of 1hlngs wlll come when Lhe Lools
for deslgnlng Lhe behavlor of Lhousands of devlces look llke Lhls. 1hey'll look llke SlmClLy or
Lemmlngs, or SLarCra or MlneCra. Sure, Lhese look llke games, buL really Lhey're Lools
for shaplng emergenL eecLs Lo Lrend Loward deslrable concluslons, raLher Lhan Lools LhaL
requlre you Lo focus on speclc rlgld behavlor.

1hese are Lhe Lools LhaL wlll help us bulld noL Lhe lnLerneL of 1hlngs, buL Lhe Coogle of
1hlngs, Lhe lacebook of 1hlngs, Lhe loursquare of 1hlngs, Lhe ayal of 1hlngs, Lhe
larmvllle of 1hlngs. 1haL's when we'll see Lhlngs really Lake o.
37
1hank you
38

You might also like