热门搜索：

2023生成式人工智能的潜在有害影响与未来之路

信息技术2023-06-26-美国电子隐私信息中心任***

AI智能总结

EPIC Report: "Generating Harm: Generative AI's Impact and Paths Forward"

Introduction

The rapid deployment of generative AI systems, particularly in the form of large language models, has raised significant concerns about potential harms. This report serves as an update on the known and anticipated impacts of generative AI as of May 15, 2023. It highlights the increasing role of AI in society and the need for regulation to mitigate risks associated with AI technologies.

Turbocharging Information Manipulation

With widespread access to free and low-cost generative AI tools, there's a risk of rapid dissemination of high volumes of content, including text, images, videos, and audio. While some content may be beneficial, these tools can also facilitate the spread of false, misleading, biased, inflammatory, or dangerous information. As AI systems become more sophisticated, the ease and speed of producing such content increase, with existing harmful content serving as a foundation for creating more.

Case Study: Election Misinformation

Generative AI tools, such as GPT-4 and subsequent large language models, can create human-sounding scripts quickly. These scripts can be distributed through various channels, influencing voter behavior or spreading misinformation in multiple states. This type of election misinformation is becoming common, and generative AI tools amplify bad actors' ability to rapidly spread believable misinformation, potentially affecting democratic processes.

Scams

Scams involving phone calls, texts, and emails have been rampant, leading to significant financial losses for consumers. Generative AI can accelerate the creation of personalized and believable scams, making them harder to detect. The use of AI-generated voices to mimic loved ones seeking immediate financial assistance can lead to fraud, especially targeting vulnerable communities such as seniors and those in debt.

Disinformation

Generative AI tools can be manipulated to produce adaptable content that supports campaigns, political agendas, or hateful positions, spreading misinformation effectively. This poses a significant risk to public trust and the integrity of information in society.

Recommendations

To address these issues, companies should be required to demonstrate transparency in their AI development processes, clearly indicate when AI is used, and provide informed consent throughout the training, development, and application phases. Legislation should protect against deliberate voter intimidation, deterrence, or interference through false or misleading information, including false claims of endorsement. Clear prohibitions on high-risk uses, addressing the spread of disinformation, requiring meaningful disclosure, facilitating informed consent, and providing guidance on various issues related to generative AI use are essential steps forward.

Conclusion

As policymakers and industry worldwide focus on the substantial risks and opportunities posed by AI, there's an opportunity to ensure that this technology benefits society. By enacting clear regulations and fostering transparency, we can mitigate the potential harms of generative AI, ensuring it serves the public good while respecting privacy, democracy, and cybersecurity.

epic.org ELECTRONICPRIVACYINFORMATIONCENTER GENERATNG HARMS GenerativeAl'sImpact& PathsForward MAY2023 CONTRIBUTIONSBY GrantFergussonTomMcBrien CaitrionaFitzgeraldCalliSchroeder ChrisFrascellaBenWinters MeganlorioEnidZhou EDITEDBY GrantFergusson,CalliSchroeder,BenWinters,andEnidZhou ThankyoutoSarahMyersWestandKatharinaKoppforyourgenerous commentsonanearlierdraftofthepaper. NotesonthisPaper: Thisisversion1ofthispaperandisreflectiveofdocumentedandanticipatedharmsofGenerativeAlasofMay15,2023.Duetothefast- changingpaceofdevelopment,use,andharmsofGenerativeAl,wewantto acknowledgethatthisisaninherentlydynamicpaper,subjecttochangesin thefuture. Throughoutthispaper,weuseastandardformattoexplainthetypologyof harmsthatgenerativeAlcanproduce.EachsectionfirstexplainsrelevantbackgroundinformationandpotentialrisksimposedbygenerativeAl,thenhighlightsspecificsharmsandinterventionsthatscholarsandregulators havepursuedtoremedyeachharm.Thispaperdrawsontwotaxonomiesof A.l.harmstoguideouranalysis: 1.DanielleCitron'sandDanielSolove'sTypologyofPrivacyHarms,comprisingphysical,economic,reputational,psychological,autonomy,discrimination,andrelationshipharms;land 2.JoyBuolamwini'sTaxonomyofAlgorithmicHarms,comprisinglossofopportunity,economicloss,andsocialstigmatization,includinglossofliberty,increasedsurveillance,stereotypereinforcement,andother dignitaryharms.2 ThesetaxonomiesdonotnecessarilycoverallpotentialAlharms,andouruseofthesetaxonomiesismeanttohelpreadersvisualizeandcontextualizeAlharmswithoutlimitingthetypesandvarietyofAlharmsthatreadersconsider. TableofContents Introduction. TurbochargingInformationManipulation. Harassment,Impersonation,andExtortion...9 Spotlight:Section230..19 ProfitsOverPrivacy:IncreasedOpagueDataCollection..24 IncreasingDataSecurityRisk..30 ConfrontingCreativity:ImpactonIntellectualPropertyRights..33 ExacerbatingEffectsofClimateChange.40 LaborManipulation,Theft,andDisplacement.44 Spotlight:Discrimination.53 ThePotentialApplicationofProductsLiabilityLaw.54 ExacerbatingMarketPowerandConcentration.57 Recommendations.60 AppendixofHarms.64 References68 Introduction OpenAl'sdecisiontoreleaseChatGPT,achatbotbuiltontheLargeLanguageModelGPT-3,lastNovemberthrustAltoolstotheforefrontofpublicconsciousness.Inthelastsixmonths,newAltoolsusedtogenerate text,images,video,andaudiobasedonuserpromptsexplodedin popularity.Suddenly,phraseslikeStableDiffusion,Hallucinations,andValue capabilitiesofgenerativeAl-andtheirpotentialforharm-emergedwithoutanyclearindicationofwhatwouldcomenextorwhatimpactsthesetools wouldhave. WhilegenerativeAlmaybenew,itsharmsarenot.AlscholarshavebeenwarningusoftheproblemsthatlargeAlmodelscancauseforyears.3Theseoldproblemsareexacerbatedbytheindustry'sshiftingoalsfromresearch andtransparencytoprofit,opacity,andconcentrationofpower.The widespreadavailabilityandhypeofthesetoolshasledtoincreasedharm bothindividuallyandonamassivescale.Alreplicatesracial,gender,anddisabilitydiscrimination,andtheseharmsareweavedinextricablythrougheveryissuehighlightedinthisreport. OpenAlandothercompanies'decisionstorapidlyintegrategenerativeAltechnologyintoconsumer-facingproductsandserviceshaveundermined longstandingeffortstomakeAldevelopmenttransparentandaccountable, leavingmanyregulatorsscramblingtopreparefortherepercussions.Andit isclearthatgenerativeAlsystemscansignificantlyamplifyriskstoboth individualprivacyandtodemocracyandcybersecuritygenerally.Inthe wordsoftheOpenAlCEO,whoindeedhadthepowernottoacceleratethereleaseofthistechnology,"I'mespeciallyconcernedthatthesemodelscouldbeusedforwidespreadmisinformation...[and]offensivecyberattacks." EPICIGeneratingHarm:GenerativeAl'sImpactandPathsForward Introduction ThisrapiddeploymentofgenerativeAlsystemswithoutadequatesafeguardsisclearevidencethatself-regulationhasfailed.Hundredsofentities,fromcorporationstomediaandgovernmententities,are developingandlookingtorapidlyintegratetheseuntestedAltoolsintoa widerangeofsystems.Andthisrapidrolloutwillhavedisastrousresults builtinfromthebeginning. Weareatacriticaljunctureaspolicymakersandindustryaroundtheglobe arefocusingonthesubstantialrisksandopportunitiesposedbyAl.Thereis anopportunitytomakethistechnologyworkforpeople.Companiesshouldberequiredtoshowtheirwork,makeitclearwhenAlisinuse,andoffer informedconsentthroughoutthetraining,development,anduseprocess. OnethreadofpublicconcernfocusesonAl's"existential"risks-speculativelong-termrisksinwhichrobotsreplacehumansatwork,socially,and ultimatelytakingover,ala“I,Robot."SomelegislatorsonthestateandfederallevelhavebeguntotaketheissueofaddressingAlmoreseriously- however,itremainstobeseeniftheirfocuswillbeonlyonsupportingcompanieswiththeirdevelopmentofAltoolsandreguiringmarginaldisclosureandtransparencyrequirements.Enactingclearprohibitionsonhigh-riskuses,addressingtheeasyspreadofdisinformation,requiring meaningfulandproactivedisclosuresthatfacilitateinformedconsent,and broadoutlineofdifferentissuesthattheuseofgenerativeAlbringsup, educatelawmakersandthepublic,andoffersomepathsforwardtomi

点击免费查看完整报告