懂你L7U2part3Word文档下载推荐.docx
《懂你L7U2part3Word文档下载推荐.docx》由会员分享,可在线阅读,更多相关《懂你L7U2part3Word文档下载推荐.docx(33页珍藏版)》请在冰点文库上搜索。
![懂你L7U2part3Word文档下载推荐.docx](https://file1.bingdoc.com/fileroot1/2023-5/6/fba91fa1-a18a-4111-a218-e641116c3890/fba91fa1-a18a-4111-a218-e641116c38901.gif)
发现本质;
嫌疑犯emotionalstatesandevenlying
fromprocessinghumanfaces.
13.Advertisersandevengovernmentsareveryinterested.
14.Ihadbecomeacomputerprogrammer
becauseIwasoneofthosekidscrazyaboutmathandscience.
15.ButsomewherealongthelineI'
dlearnedaboutnuclearweapons,
andI'
dgottenreallyconcernedwiththeethics伦理学;
道德学ofscience.
Iwastroubled.
16.However,becauseoffamilycircumstances,
Ialsoneededtostartworkingassoonaspossible.
17.SoIthoughttomyself,hey,letmepickatechnicalfield
whereIcangetajobeasily
andwhereIdon'
thavetodealwithanytroublesomequestionsofethics.
18.SoIpickedcomputers.
19.Well,ha,ha,ha!
Allthelaughsareonme.
20.Nowadays,computerscientistsarebuildingplatforms平台
thatcontrolwhatabillionpeopleseeeveryday.
21.They'
redevelopingcarsthatcoulddecidewhotorunover.
22.They'
reevenbuildingmachines,weapons,
thatmightkillhumanbeingsinwar.
23.It'
sethicsallthewaydown.
练
1.Theidiom“somewherealongtheline”meansatsomepointduringaprocess.
OnMachineIntelligence2
1.Machineintelligenceishere.
2.We'
renowusingcomputationtomakeallsortofdecisions,
butalsonewkindsofdecisions.
3.We'
reaskingquestionstocomputationthathavenosinglerightanswers,
thataresubjective
andopen-endedandvalue-laden偏向于接受特定价值观念的.
4.We'
reaskingquestionslike,
Whoshouldthecompanyhire?
5."
Whichupdatefromwhichfriendshouldyoubeshown?
Whichconvict宣判…有罪ismorelikelytoreoffend再冒犯;
再犯罪?
7."
Whichnewsitemormovieshouldberecommendedtopeople?
8.Look,yes,we'
vebeenusingcomputersforawhile,
butthisisdifferent.
9.Thisisahistoricaltwist欺骗;
欺诈,
becausewecannotanchor主持人;
顶梁柱;
刹车;
computationforsuchsubjectivedecisions
thewaywecananchorcomputationforflyingairplanes,buildingbridges,
goingtothemoon.
10.Areairplanessafer?
11.Didthebridgesway摇摆;
摆动andfall?
12.There,wehaveagreed-upon一致同意的;
约定,fairlyclearbenchmarks标准检查程序;
基准,
andwehavelawsofnaturetoguideus.
13.Wehavenosuchanchorsandbenchmarks
fordecisionsinmessyhumanaffairs.
B
1.Tomakethingsmorecomplicated,oursoftwareisgettingmorepowerful,
butit'
salsogettinglesstransparent透明的;
受公众监督的;
易察觉的andmorecomplex.
2.Recently,inthepastdecade,
complexalgorithms运算法则;
计算程序havemadegreatstrides阔步行进;
大踏步走.
3.Theycanrecognizehumanfaces.
4.Theycandecipher破译;
译解;
handwriting.
5.Theycandetectcreditcardfraud欺骗
andblockspam垃圾邮件
andtheycantranslatebetweenlanguages.
Theycandetecttumors肿瘤inmedicalimaging.
6.TheycanbeathumansinchessandGo.
7.Muchofthisprogresscomesfromamethodcalled"
machinelearning."
8.Machinelearningisdifferentthantraditionalprogramming,whereyougivethecomputerdetailed,exact,painstaking刻苦的;
下苦功的instructions.
9.It'
smorelikeyoutakethesystemandyoufeeditlotsofdata,
includingunstructureddata,
likethekindwegenerateinourdigitallives.
10.Andthesystemlearnsbychurning搅拌;
搅动;
翻腾throughthisdata.
11.Andalso,crucially,
thesesystemsdon'
toperateunderasingle-answerlogic.
12.Theydon'
tproduceasimpleanswer;
it'
smoreprobabilistic可能性的;
或然性的:
Thisoneisprobablymorelikewhatyou'
relookingfor."
13.Now,theupside积极面is:
thismethodisreallypowerful.
14.TheheadofGoogle'
sAIsystemscalledit,
theunreasonableeffectivenessofdata."
15.Thedownsideis,
wedon'
treallyunderstandwhatthesystemlearned.
Infact,that'
sitspower.
16.Thisislesslikegivinginstructionstoacomputer;
it'
smoreliketrainingapuppy-machine-creature
treallyunderstandorcontrol.
17.Sothisisourproblem.
18.It'
saproblemwhenthisartificialintelligencesystemgetsthingswrong.
19.It'
salsoaproblemwhenitgetsthingsright,becausewedon'
tevenknowwhichiswhichwhenit'
sasubjectiveproblem.
20.Wedon'
tknowwhatthisthingisthinking.
C
1.So,considerahiringalgorithm--
asystemusedtohirepeople,usingmachine-learningsystems.
2.Suchasystemwouldhavebeentrainedonpreviousemployees'
data
andinstructedtofindandhire
peopleliketheexistinghighperformersinthecompany.
3.Soundsgood.
4.Ionceattendedaconference
thatbroughttogetherhumanresourcesmanagersandexecutives,
high-levelpeople,usingsuchsystemsinhiring.
5.Theyweresuperexcited.
6.Theythoughtthatthiswouldmakehiringmoreobjective,lessbiased有偏见的;
有偏的;
结果偏倚的,
andgivewomenandminorities少数;
未成年;
未及法定年龄abettershot
againstbiasedhumanmanagers.
7.Andlook--humanhiringisbiased.
8.Iknow.
Imean,inoneofmyearlyjobsasaprogrammer,
9.
Myimmediatemanager直属经理wouldsometimescomedowntowhereIwas
reallyearlyinthemorningorreallylateintheafternoon,
andshe'
dsay,"
Zeynep,let'
sgotolunch!
10.I'
dbepuzzledbytheweirdtiming.
It'
s4pm.Lunch?
11.Iwasbroke,sofreelunch.Ialwayswent.
12.Ilaterrealizedwhatwashappening.
13.Myimmediatemanagershadnotconfessed承认totheirhigher-ups高层;
大人物
thattheprogrammertheyhiredforaseriousjobwasateengirl
whoworejeansandsneakers运动鞋towork.
14.Iwasdoingagoodjob,Ijustlookedwrong
andwasthewrongageandgender.
15.Sohiringinagender-andrace-blind不分性别和种族way
certainlysoundsgoodtome.
16.Butwiththesesystems,itismorecomplicated,andhere'
swhy:
17.Currently,computationalsystemscaninferallsortsofthingsaboutyou
fromyourdigitalcrumbs细屑,
evenifyouhavenotdisclosed揭露;
使显露;
公开thosethings.
18.Theycaninferyoursexualorientation,
yourpersonalitytraits特征,
yourpoliticalleanings.
19.Theyhavepredictive预言性的;
预计的powerwithhighlevelsofaccuracy准确;
精确;
准确度.
20.Remember--forthingsyouhaven'
tevendisclosed.
Thisisinference.
练:
1.Therearenoguidelinesforsubjectiveissues.
2.AccordingtoTufekci,machineintelligenceshouldnotbetrustedtohirenewemployees.
3.Toanchorsomethingmeanstomakeitstableandsecure.
4.Itisoftenunclearhowitcomestoitsconclusions.
5.Itleadstoprobabilistic可能性的;
或然性的answers.
6.Ifamethodorargumentisprobabilistic,itisbasedonwhatismostlikelytobetrue.
7.Shedidn’tappearqualifiedforthejobduetoherageandgender.
8.Tomakeaninferencemeanstoformanopinionbasedontheavailableinformation.
9.Iwasdoingagoodjob,Iwasthewrongageandgender.
OnMachineIntelligence3
1.Ihaveafriendwhodevelopedsuchcomputationalsystems
topredictthelikelihoodofclinical临床的;
临诊的;
冷静能干的;
客观的;
冷淡疏远的;
不偏不倚的;
orpostpartum产后的depression
fromsocialmediadata.
2.Theresultsareimpressive.
3.Hersystemcanpredictthelikelihoodofdepression
monthsbeforetheonset开始ofanysymptoms症候;
症状;
病徵;
--
monthsbefore.
4.Nosymptoms,there'
sprediction.
5.Shehopesitwillbeusedforearlyintervention插入;
介入;
干预;
干涉;
措施;
做法.Great!
6.Butnowputthisinthecontextofhiring.
7.Soatthesehumanresourcesmanagers’conference,
Iapproachedahigh-levelmanagerinaverylargecompany,
8.AndIsaidtoher,"
Look,whatif,unbeknownst未知的toyou,
yoursystemisweeding淘汰outpeoplewithhighfuturelikelihoodofdepression?
9.They'
renotdepressednow,justmaybeinthefuture,morelikely.
10.Whatifit'
sweedingoutwomenmorelikelytobepregnant
inthenextyearortwobutaren'
tpregnantnow?
11.Whatifit'
shiringaggressivepeoplebecausethat'
syourworkplaceculture?
12.Youcan'
ttellthisbylookingatgenderbreakdowns.
Thosemaybebalanced.
13.Andsincethisismachinelearning,nottraditionalcoding编码,
14.thereisnovariable易变的;
多变的;
反复无定的therelabeled"
higherriskofdepression,"
higherriskofpregnancy,"
aggressiveguyscale."
15.Notonlydoyounotknowwhatyoursystemisselectingon,
youdon'
tevenknowwheretobegintolook.
sablackbox.
16.Ithaspredictivepower,butyoudon'
tunderstandit.
17."
Whatsafeguards,"
Iasked,"
doyouhave
tomakesurethatyourblackboxisn'
tdoingsomethingshady不正当的;
靠不住的;
可疑的?
18.ShelookedatmeasifIhadjuststeppedon10puppytails.
19.Shestaredatmeandshesaid,
Idon'
twanttohearanotherwordaboutthis."
20.Andsheturnedaroundandwalkedaway.
21.Mindyou--shewasn'
trude.
Itwasclearly:
whatIdon'
tknowisn'
tmyproblem,goaway,deathstare.
22.Look,suchasystemmayevenbelessbiased
thanhumanmanagersinsomeways.
23.Anditcouldmakemonetary金钱的;
货币的sense.
24.Butitcouldalsolead
toasteady稳定的butstealthy隐秘的shuttingoutofthejobmarket
ofpeoplewithhigherriskofdepression.
25.Isthisthekindofsocietywewanttobuild,
withoutevenknowingwe'
vedonethis,
26.
Becauseweturneddecision-makingtomachineswedon'
ttotallyunderstand?
1.Anotherproblemisthis:
thesesystemsareoftentrainedondatageneratedbyouractions,
humanimprints痕迹.
2.Well,theycouldjust