《AI Now研究所：2024生成式AI時代的商業模式：開啟AGI崛起與營收模式迭代的“狂飆之旅”研究報告（英文版）（30頁）.pdf》由會員分享，可在線閱讀，更多相關《AI Now研究所：2024生成式AI時代的商業模式：開啟AGI崛起與營收模式迭代的“狂飆之旅”研究報告（英文版）（30頁）.pdf（30頁珍藏版）》請在三個皮匠報告上搜索。

1、Brian MerchantDecember 2024AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model2TableofContentsIntroduction3OpenAI and the Generative AI Boom6Silicon Valley Mythology,Distilled and Accelerated7From“SafeAI”toAGIandtheHype-LedBusinessModelGenesis12Marketing AGI,Shipping Co

2、mmercial AI21The Dream of AGI and the Fully AutomatedOrganization24Acknowledgements30AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model3IntroductionIn the spring of 2019,at a live event for StrictlyVC,the technology journalist Connie Loizosasked Sam Altman how the unus

3、ually structured company he ran planned on generatingrevenue.“OpenAI is so amorphous,”she said.“But it is a business.”Was the aim,shewondered,to license its technology,or customize algorithms for paying clients?“How is itgoing to work?”1“The honest answer is that we have no idea,”Altman replied.“We

4、have never made anyrevenue,we have no current plans to make revenue,we have no idea how we may one daygenerate revenue.We have made a soft promise to investors that once weve built thissort of generally intelligent system,basically we will ask it to figure out a way to generatean investment return f

5、or you.”2The room rippled with laughter,but Altman was serious.“It sounds like an episode ofSilicon Valley,it really does,I get it,you can laughits all right,”he said.“But it is what Iactually believe is gonna happen.”3Altmans comment is telling.It contains at least two key truths about OpenAI,theco

6、mpany that has spearheaded the generative AI boom now dominant in Silicon Valleyand the broader business world.First,before OpenAI released DALL-E and ChatGPT,thetwo buzzy products that would transform it into a household name,it had no businessmodel at all.Second,its executive leadership believed t

7、hat the development of its ultimatecore product,an artificial general intelligence(AGI),would solve the issue;profits wouldfollow.This combinationa lack of a premeditated business model and an insistence onreferring the public,partners,and investors to its pursuit of AGI,an aspirationaltechnological

8、 construct,as its raison dtre,in its steadhas shaped OpenAIs path tobecoming the definitive Silicon Valley unicorn of the 2020s.This report set out to investigate and elucidate the business models behind the generativeAI companies that are drawing hundreds of billions of dollars in investment.Such m

9、odelshave pushed companies like Nvidia,which supplies the chips necessary for AI3Ibid.2Ibid.1“Sam Altman in Conversation with StrictlyVC,”interview by Connie Loizos,YouTube,53:25,May 18,2019.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model4computation,to double and t

10、hen triple its$418 billion valuation in 2022 to a historicmarket capitalization in excess of$3 trillion in 2024.4It soon became clear thatunderstanding the composition of those business models meant understanding thedeployment and evolution of the concept of“AGI”as a lodestar for generative AIcompan

11、ies.Any e?ort to understand OpenAIs business modeland that of its emulators,peers,andcompetitorsmust thus begin with the understanding that they have been developedrapidly,even haphazardly,and out of necessity,to capitalize on the popularity ofgenerative AI products,to fund growing compute costs,and

12、 to pacify a growing portfolioof investors and stakeholders.Equally crucial is understanding how“AGI”operates in amaterial context,and how it serves as a driver of continued investment and enterprisesales,a marketing and recruitment tool,and a framework for bolstering the companysinfluence and cultu

13、ral footprint.That OpenAI had no discernible business model upon its inception does not mean thatprofit potential wasnt a consideration from the beginning.While the headlines announcingOpenAIs launch reliably painted the project as Elon Musk and Sam Altmans humanitariane?ort to protect the world fro

14、m a malignant,superpowerful AI,it was from the start adensely corporatized undertaking,established in a posh hotel in the middle of SiliconValley with seed money from tech billionaires,Amazon,and top venturecapitalistsdespite being labeled a“nonprofit.”Theres no reason to doubt that Altman was answe

15、ring earnestly when he declared toLoizos that OpenAI had no business model to speak of.It wasnt a pressing issue.Why?Forone thing,this was the end of the 2010s,a decade in which the tech sector enjoyedrock-bottom federal interest rates and it was never easier for founders and startups toattract fund

16、ing;as a result,much of Silicon Valleys investor class had grown accustomedto making bets that promised scale,grandeur,and“unicorn”status over reliable marginalreturns and sound business plans.In fact,we might consider OpenAI the apotheosis of atrend that rose to prominence with startups like Uber a

17、nd WeWork,formed in the headypost-Web 2.0 era.It was a time when highly salable products like the iPhone and ad-basedGoogle search ruled the tech sector,and when it was assumed that software would eatthe world with such ferocity that few people worried about overhead or operating costs.That picture

18、had changed dramatically by the 2020s.The federal reserve began raisinginterest rates in December 2015incidentally,the same month that Altman announcedthe launch of OpenAIafter seven years of keeping them near zero in the wake of the4“Nvidias Stock Is Expensive.A Look at Why,and How That Should Chan

19、ge,by the Numbers,”Associated Press,June 3,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model5Great Recession.5The hikes ended the era of free money for Silicon Valley,and investmentin startups began to slow.The higher credit rates quickly made clear that tech had

20、another issue on its hands:it hadnt had a truly hit product category in years.Facebook,Google,and the iPhone had all risen to prominence more than a decade ago;Uber,Lyft,and other gig-economy companies were still not profitable,and a cascade of broadlyhyped tech trendscrypto,Web3,the metaversecrashe

21、d and burned in rapidsuccession.Waves of layo?s hit the industry in 2022 and 2023,and Silicon Valley was,forthe first time in a generation,at risk of entering into a recession.These are the conditions into which the AI boom was born.And that boom,launched by acompany built on the assumptions and log

22、ics of the zero-percent-interest-rate era,withpresumed access to vast stores of investment,soon acquired an urgent need for aproduct to sell to users and corporate clients.In the rush to capitalize on its rise toprominence,AI companies,especially OpenAI,would wield a story about a technology sopower

23、ful it would soon have a life of its ownso powerful you could simply ask it how tomake more money,and it would oblige.All of this converged to give rise to the moderngenerative AI trend.The business model,it could safely be assumed,would come later.5Jon Hilsenrath and Ben Leubsdorf,“Fed Raises Rates

24、 After Seven Years Near Zero,Expects Gradual Tightening Path,”Wall StreetJournal,December 16,2015.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model6OpenAI and the Generative AI BoomTo understand the thrust and structure of the business models taking shape in thegenera

25、tive AI space,we need to understand OpenAI,the driving force that mostif notallof the new companies are reacting to,emulating,or feeding o?.OpenAI is at thecenter of nearly every conversation about the technology,both inside the industry andout.With a self-reported$3.4 billion in annualized revenue

26、projected this year,6its thelargest company operating in the space,with the highest valuation:$157 billion,based onits most recent funding round.7Meanwhile,OpenAIs most similarly constituted competitor,Anthropic(which says its ontrack to make$850 million in revenue this year8),was founded by former

27、OpenAIemployees,as was the much-hyped“answer engine”Perplexity.Microsoft is OpenAIsclosest partner,lending it$10 billion in funds and cloud compute,and getting a8“Anthropic Forecasts More than$850 Mln in Annualized Revenue Rate by 2024-End Report,”Reuters,December 26,2023.7Cade Metz,“OpenAI Complete

28、s Deal That Values Company at$157 Billion,”New York Times,October 2,2024.6Stephanie Palazzolo and Erin Woo,“OpenAIs Annualized Revenue Doubles to$3.4 Billion Since Late 2023,”The Information,June 12,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model7ChatGPT-powere

29、d Copilot out of the deal,while Google and Facebook hastily launchedproduct development cycles of their own in response to the shock waves sent out byChatGPT in late 2022 and early 2023.We lead with the case study of OpenAI,and interrogate how its a rather immaculateproduct of its time and place:uni

30、corn-hunting-era Silicon Valley and the decade ofzero-percent interest rates.Silicon Valley Mythology,Distilled andAccelerated“If you think about the things that are most important to the future of the world,I thinkgood AI is probably one of the highest things on that list.So we are creating OpenAI,

31、”SamAltman told veteran tech journalist Steven Levy in 2015,in an interview announcing thecompany.“The organization is trying to develop a human positive AI.And because its anon-profit,it will be freely owned by the world.”9To anyone unfamiliar with the historical context of OpenAIs founding mission

32、,the notionthat the$157 billion company,which has raised the largest VC round in history,10andwhose revenue comes from selling enterprise software and API access to corporations likePwC,is a“nonprofit”that will be“freely owned by the world”may seem jarring.Therefore,its worth charting its evolution

33、from nonprofit to very-much-for-profitan importantpiece of the puzzle because this evolution is what has e?ectively eroded the boundarybetween mythology and business model.OpenAI began as a grandiose,science fiction-adjacent,and high-concept venture whoseaim was explicit in its ambitions:to ensure t

34、hat superpowerful AGI was developed safely,in a way that would benefit all humanity.The purported gravity of this mission was suchthat,for years,it eschewed a business model altogether(an approach also subsidized bythe beneficiaries of the unicorn era of heavy VC investment).As such,OpenAIs eventual

35、$157 billion valuation marks a milestone:It is the first modern tech giant birthed with apromise that what it was doing was so momentous that it need not consider revenue atallit would be a nonprofit.Silicon Valley has long had a penchant for grandiosity,but if were charting the ratio of itsworld-ch

36、anging talk to demonstrable business feasibility,from,say,Apple,which helped10Ina Fried,“OpenAI Raises$6.6 billion in largest VC round ever.”Axios,October 2,2024.9Steven Levy,“How Elon Musk and Y Combinator Plan to Stop Computers From Taking Over,”Medium,December 11,2015.AI Generated Business:The Ri

37、se of AGI andthe Rush to Find a Working Revenue Model8mint the practice of cosmic sales pitches yet still sold hardware in retail shops;to Uber,whose talk of disruption was ubiquitous,whose app was popular,but whose bottom-linemath was always fuzzy;to OpenAI,then perhaps we can see the moment when t

38、he cart isplaced fully before the horse:when the story of a venture,the story of a new technologywith commercial potential,eclipsed the need for a business model at all.But the story starts in earnest before OpenAI is even founded,when“superintelligent AI”is a trending area of interest and concern a

39、mong Silicon Valley elites11and,as well see,arguably a way for late entrants in the space to gain competitive advantage.In the 2010s,the commercial AI space was dominated by Google and Facebook,which were eachinvesting in machine learning technologies and competing to acquire promising startupslike

40、Deepmind and to hire top talent like Geo?rey Hinton.Analysts like Musks biographer,Bloomberg reporter Ashlee Vance,have argued that this made Musk professionallyjealous.Publicly,Musk may have worried about AI safety;privately,his concerns may havebeen more about missing out on a major developing tec

41、h trend and businessopportunity.12A widely documented argument between Musk and former Google CEOSergey Brin over whether AI would benefit or threaten humanity is sometimes cited as thestarting point for OpenAI.13Yet it is important to note that Musk had personal and financialreasons to oppose the c

42、ommercial development of AI by Google or the other tech giants.To understand the genesis of OpenAI,we consider the likelihood that the company wasnever the altruistic venture it portrayed itself as to the mediadespite being formed as anonprofitbut rather a vehicle through which Musk and Altman could

43、 achieve variouspersonal and business objectives,including stalling Googles progress on AI andpositioning OpenAI to be a more attractive alternative in the eyes of top research talent,consumers,and thus investors.13See Cade Metz,Karen Weise,Nico Grant,and Mike Isaac,“Ego,Fear and Money:How the A.I.F

44、use Was Lit,”New York Times,December3,2023.12Ashlee Vance,Elon Musk:Tesla,SpaceX,and the Quest for a Fantastic Future(New York:Ecco Press,2015).11“Superintelligence on NYT Bestseller List,”Future of Humanity Institute,University of Oxford,August 18,2024.AI Generated Business:The Rise of AGI andthe R

45、ush to Find a Working Revenue Model9A Timeline of Crucial Events Regarding theFormation of OpenAI,and the Influence of Muskand the Specter of“AGI”in Its Organizational DNA2012Elon Musk meets the founder of Deepmind,Demis Hassabis,who,accordingto multiple accounts,sparks his interest in AI by detaili

46、ng the risks it posesto humanity.14Musk invests$5 million in the company.Musk meets SamAltman,then a partner at influential startup incubator Y Combinator.2013Google expresses interest in acquiring Deepmind.Musk discusses AI at hisbirthday party in Napa with Google cofounder Larry Page.15Musk isrepo

47、rtedly envious of Googles dominance in AI,and his formerly closerelationship with Page is fraying.He reportedly argues that Page is notcautious enough with AI;the two have a falling-out.Musk attempts todissuade Hassabis from selling to Google.162014Google announces it has acquired Deepmind for$500 m

48、illion.17Musk makesheadlines by declaring AI to be“our biggest existential threat”during aninterview at MIT,among the first of his documented statements about thepower of malign AI.18Altman,now president of Y Combinator,beginsexchanging emails with Musk about AI.2015“Development of superhuman machin

49、e intelligence,”Altman writes on hispersonal blog in February,“is probably the greatest threat to the continuedexistence of humanity.”19Altman emails Musk in May proposing they startan AI lab together.“Thinking a lot about whether its possible to stophumanity from developing AI.I think the answer is

50、 almost definitely not.Ifits going to happen,it seems like it would be good for someone other than19Sam Altman,“Machine Intelligence,Part 1,”February 25,2015.18Samuel Gibbs,“Elon Musk:Artificial Intelligence Is Our Biggest Existential Threat,”Guardian,October 27,2014.17Catherine Shu,“Google Acquires

51、 Artificial Intelligence Startup DeepMind for More than$500M,”TechCrunch,January 26,2014.16Kai Xiang Teo,“Musk Once Tried to Stop Googles DeepMind Acquisition in 2014,Saying the Future of AI Shouldnt Be Controlled byLarry Page,”Yahoo News,September 7,2023.15Isaacson,“Inside Elon Musks Struggle for t

52、he Future of AI.”14See Walter Isaacson,“Inside Elon Musks Struggle for the Future of AI,”Time,September 6,2023;and Metz,Weise,Grant,and Isaac,“Ego,Fear and Money.”AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model10Google to do it first.”20Musk and Altman organize a di

53、nner at the SiliconValley dealmaking hot spot the Rosewood Hotel with leading AI researchersand founders to discuss starting that lab.Musk leads an open letter callingfor a ban on AI weapons.21Altman and Musk recruit Stripe CTO GregBrockman and notable AI researcher Ilya Sutskever,among others,andbe

54、gin forming OpenAI.Nov.2015In an email discussing the announcement of OpenAI,Elon Musk writes:“Idfavor positioning the blog to appeal a bit more to the general publicthereis a lot of value to having the public root for us to succeedand thenhaving a longer,more detailed and inside-baseball version fo

55、r recruiting,with a link to it at the end of the general public version.”22Dec.2015OpenAI publicly launches with a stated mission of developing AI“in a waymost likely to benefit humanity as a whole,”to widespread media interest.23From the outset,while publicly positioned as a nonprofit,OpenAI was st

56、ructured andanticipated by its founders as having the potential for profit,or at least the potential towield influence.While there was no business model,OpenAI served a distinct businesspurpose:suppressing Musks competitors and o?ering the ambitious Altman a bridge toMuskone of the wealthiest and mo

57、st influential figures in Silicon Valleyand his circle ofallies.One lens through which to view the emergence of“AGI”or“human-level AI”or“superintelligent AI”as a widespread discussion point is thus as the result of a businessand retaliatory tactic leveraged by Musk and embraced by Altman,as an e?ort

58、 to lash outagainst or even attempt to suppress development and/or deployment of commercial AIproducts by Google in particular.This positioning also benefits the institution or companywhose public aim is to develop and deploy AI“safely”which happens to be OpenAIsfounding mission.This tactic may have

59、 proved successful.Google,after all,had developed many keyframeworks and technologies,like TensorFlow,an open source machine learning softwarelibrary;and the“transformer”that gives the General Pre-Trained Transformer in ChatGPT23“Introducing OpenAI,”OpenAI,December 11,2015.22“OpenAI and Elon Musk,”O

60、penAI,March 5,2024.21Samuel Gibbs,“Musk,Wozniak and Hawking Urge Ban on Warfare AI and Autonomous Weapons,”Guardian,July 27,2015.20Nick Robins-Early,“The Feud Between Elon Musk and Sam Altman Explained,”Guardian,March 9,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue

61、 Model11its name.It had a yearslong head start on OpenAI on LLM technology:in 2013 Hintonbrought his machine learning acumen to Google,which acquired DeepMind the followingyear,introduced LaMDA in 2020,and developed Bard and Gemini in 2023.In 2016,justmonths after becoming CEO,Sundar Pichai announce

62、d that Google would become“anAI-first”company.24Yet OpenAI ate its lunch.This is partly because Google had stepped on a series of rakes in its commercial AIdevelopment,gaining a reputation for releasing“creepy”products and silencing critical AIresearchers,both internally and externally.In 2020,the c

63、ompany pushed out star AIethicist Timnit Gebru after preventing her now-seminal paper“On the Dangers ofStochastic Parrots:Can Language Models Be Too Big?”from being published.25In 2022,itfired software engineer Blake Lemoine,who went public with his fears that Googlesinternal chatbot products had be

64、come sentient.26As a result,the company did not releasea similar chatbot product to OpenAI,in part out of fear that such a product would result inpublic anxiety and bad press.27Google issued a“code red”internally after OpenAIsChatGPT exploded in popularity at the end of 2022,and raced to catch up by

65、 releasing acompeting product.There is also a lesson here about the importance of developing acorporate narrative:by the mid 2010s,Google,the object of public ire and rounds of badpublicity for releasing Google Glass and for revelations that it crawled your inbox toimprove ad suggestions(both of whi

66、ch the public felt violated privacy concerns),had areputation for being invasive;when it released an early voice-activated AI chatbot productcalled Duplex,it caused controversy and backlash.28After years of nurturing its image asconcerned custodians of safe AI,OpenAI su?ered little such blowback.Wit

67、h this context,that OpenAIs founding mission was guided by distinct businessobjectives,we can consider the nature of that founding missionto build a safe,human-level AI that would benefit everyoneand how it evolved.Or at least,we canconsider the story of that founding mission.Musk certainly understo

68、od the benefit ofhaving a narrative behind his companies:SpaceX would catapult humanity into amultiplanetary species;Tesla was supposed to help save the world from climate change;OpenAI would save humanity from superpowerful AI.And at a time when investors werebankrolling otherwise underwater but ex

69、tremely ambitious startups like Uber,Lyft,Instacart,and WeWorkcompanies that lost money every year but promised massive scaleand disruptionthe story was crucial.28Nick Statt,“Google Now Says Controversial AI Voice Calling System Will Identify Itself to Humans,”Verge,May 10,2018.27Richard Nieva,Alex

70、Konrad and Kenrick Cai,“AI First to Last:How Google Fell Behind in the AI Boom,”Forbes,February 9,2023.26Jon Brodkin,“Google Fires Blake Lemoine,the Engineer Who Claimed AI Chatbot Is a Person,”Ars Technica,July 25,2022.25Karen Hao,“We Read the Paper That Forced Timnit Gebru out of Google.Heres What

71、 It Says,”MIT Technology Review,December 4,2020.24Jillian DOnfro,“Googles CEO Is Looking to the Next Big Thing Beyond Smartphones,”Business Insider,April 21,2016.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model12“There is a lot of value,”as Musk put it in one of his

72、emails to fellow OpenAI founders,“tohaving the public root for us to succeed.”29*Next,well examine the recent historical context of OpenAIs founding mission,whichappears so jarring now,and chart its evolution from nonprofit to very-much-for-profit.Its an evolution that has been bu?eted by low intere

73、st rates that infused the SVecosystem with its most adventuring mindset,and eroded the boundary betweenmythology and business model.29“Open AI and Elon Musk.”AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model13From“SafeAI”toAGIandtheHype-LedBusinessModelGenesisWhen Ope

74、nAI launched,it described itself as a“nonprofit artificial intelligence researchcompany”whose mission was“to advance digital intelligence in the way that is most likelyto benefit humanity as a whole,unconstrained by a need to generate financial return.”30Its initial financial backers included Musk,3

75、1Altman,Brockman,Peter Thiel,LinkedInfounder Reid Ho?man,the Indian IT company Infosys,Y Combinator,and Amazon WebServices.Its stated aim was to be able to hire top AI researchers from high-paying gigslike Googleto conduct research like a nonprofit while paying for-profit salaries.OpenAIs launch ann

76、ouncement raised the alarm,once again,of an incipientsuperintelligent AI,though this time it referred to the concept as“human-level AI.”Its veryfirst post declared:“Its hard to fathom how much human-level AI could benefit society,and its equally hard to imagine how much it could damage society if bu

77、ilt or usedincorrectly.”32From the outset,OpenAIs dramatic mission struck a chord with the tech press,in no smallpart due to Musks involvement,and,to a lesser extent,Altmans.Taking cues from theorganizations first press release and interviews with key players given at the time,themedia described the

78、 project as a heroic,even utopian,e?ort.“Elon Musk and PartnersForm Nonprofit to Stop AI from Ruining the World,”read a headline in the Verge.33“Tech33Russell Brandom,“Elon Musk and Partners Form Nonprofit to Stop AI from Ruining the World,”Verge,December 11,2015.32“Introducing OpenAI.”31Though the

79、dollar amounts have not been made public,Musk is understood to be the largest and most important initial donor,withreported funding in the$45 million to“hundreds of millions”range.30“Introducing OpenAI.”AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model14Giants Pledge$

80、1Bn for Altruistic AI Venture,OpenAI,”said another on the BBCswebsite.34Vanity Fair described OpenAI as“a non-profit company to save the world from adystopian future.”35And Levys interview with Altman was titled“How Elon Musk and YCombinator Plan to Stop Computers from Taking Over.”Altman and Musk t

81、hus imprinted onto the tech press and the general public the notionthat OpenAI would be dedicated to the cause of developing a“safe”superintelligent AI,free of a profit motive.36It made for a compelling narrativeunlike the gargantuan Googleor monolithic Microsoft,here were researchers uninterested i

82、n making money but all in onprotecting the world from a technology so powerful it could lead to humanitys ruin ifpoorly stewarded.Altman even told Levy that OpenAI would be“open source and usableby everyone.”37It is remarkable,with the gift of hindsight,how little Musks or Altmansmotives were interr

83、ogated at the time.This also marks the early deployment of OpenAIs key marketing terms,like“safe AI,”“open,”“beneficial to all of humanity,”and“superpowerful”in hindsight,we can nowconsider these marketing terms from the outset.Especially pertinent is the shift in focusfrom“open source”and“safe”AI,t

84、erms handy for spinning up headlines and attractinggoodwill and idealistic,talented researchers,to“creating AGI,”which would come to becentral to OpenAIs language,and which is a mission that promises a more proactivedevelopment of powerful technologyand one that might benefit your business.38It may

85、be useful to consider the central term of art here.Artificial intelligence was itselforiginally coined as a marketing phrase.“I invented the term artificial intelligence,”AIpioneer John McCarthy said in the 1973 Lighthall debate,“and I invented it because wehad to do something when we were trying to

86、 get money for a summer study in 1956.”39The term was devised,in other words,to appeal to funders.It is also worth noting that AIresearchers in the 1960s and 70s already meant“AGI”when they said“AI”manyluminaries in the field made predictions that an artificial intelligence that could doanything a h

87、uman could do would emerge within a generation at the latest.40Thelooseness of this terminology was capitalized on by early computer firms like IBM,whichinvested in the space;and was promulgated in pop culture,perhaps most famously in2001:A Space Odyssey,which took place in the year AI scientists be

88、lieved human-level AIwould arise.40Je?rey Funk and Gary Smith,“Why A.I.Moonshots Miss,”Slate,May 4,2021.39Pierre-Yves Oudeyer,“The Lighthill Debate on Artificial Intelligence:The General Purpose Robot Is a Mirage,”YouTube,1:23:11,March 7,2018.38More recently,an OpenAI representative gave a demo wher

89、e he claimed that Sora is an important stepping stone on the road to AGI,because in order to create good videos of people,you have to know how people think.37Levy,“How Elon Musk and Y Combinator Plan to Stop Computers from Taking Over.”36See Leslie Hook,“Silicon Valley in Move to Keep AI Safe,”Finan

90、cial Times,December 11,2015.35Emily Jane Fox,“Sam Altman on His Plan to Keep A.I.out of the Hands of the Bad Guys,”Vanity Fair,December 15,2015.34“Tech Giants Pledge$1Bn for Altruistic AI Venture,OpenAI,”BBC,December 12,2015.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue

91、 Model15AI has in other words been deployed as a marketing term for decades,as computationallinguistics scholar Emily Bender and others have argued at length.41But if it is primarily amarketing termone used to describe a category of technical sciences and productdevelopment so broad it approaches me

92、aninglessnessthen its also a familiar one,andone less capable of exciting the interests of investors and the public.It was in need of anupgradeand AGI,not just artificial intelligence,but a powerful general intelligence thatcan do whatever a human can,fit the bill.The term“artificial general intelli

93、gence”wascoined by Mark Gubrud in a 1997 paper as a means of describing an AI distinct from thenarrower“expert systems”that had come to dominate the field after those earlier loftypredictions of intelligent software failed to come to pass.42The term was revived in the2010s.43Eventually,OpenAIalong w

94、ith much of the tech industrywould seize on the term andrender it central to their marketing e?orts.Heres an overview of this evolution.Timeline of OpenAIs Deployment of AI and AGITermsDecember2015The company launches,describing itself as“a non-profit artificialintelligence research company.Our goal

95、 is to advance digitalintelligence in the way that is most likely to benefit humanity as awhole,unconstrained by a need to generate financial return.Sinceour research is free from financial obligations,we can better focuson a positive human impact.“We believe AI should be an extension of individual

96、human wills and,inthe spirit of liberty,as broadly and evenly distributed as possible.Theoutcome of this venture is uncertain and the work is di?cult,but webelieve the goal and the structure are right.We hope this is whatmatters most to the best in the field.Its hard to fathom how muchhuman-level AI

97、 could benefit society,and its equally hard to imaginehow much it could damage society if built or used incorrectly.”4444“Introducing OpenAI.”My emphasis.43David Deutsch,“Philosophy Will Be the Key That Unlocks Artificial Intelligence,”Guardian,October 3,2012.42Mark Gubrud(mgubrud),“This is the orig

98、inal(first published)definition of AGI.It was not a pejorative against anything.It was anattempt to set a milestone for future AI that would have qualitatively distinct(from then-AI)impacts esp.re international security,”Twitter(now X),July 25,2019.41See Emily M.Bender,“Opening Remarks on AI in the

99、Workplace:New Crisis or Longstanding Challenge,”Medium,October 1,2023.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model16July 2016In the initial announcement detailing OpenAIs specific goalsthe firstis to measure progress,the second to build a household robot,thethird

100、 to build an agent with“natural language understanding,”and thefourth to have an agent“solve a wide variety of games”the companyreiterates its mission:“OpenAIs mission is to build safe AI,andensure AIs benefits are as widely and evenly distributed aspossible.”45November2016OpenAI and Microsoft annou

101、nce a deal:OpenAI will use MicrosoftsAzure cloud service for its experiments.“Its great to work withanother organization that believes in the importance ofdemocratizing access to AI.Were looking forward to accelerating theAI community through this partnership.”46Early 2017According to Altman,he,Musk

102、,and the executive leadership ofOpenAI begin discussing plans to restructure the company to acquireadditional funding.“We spent a lot of time trying to envision aplausible path to AGI.In early 2017,we came to the realization thatbuilding AGI will require vast quantities of compute.We begancalculatin

103、g how much compute an AGI might plausibly require.We allunderstood we were going to need a lot more capital to succeed atour missionbillions of dollars per year,which was far more than anyof us,especially Elon,thought wed be able to raise as thenon-profit.”47Elon Musk suggests folding OpenAI into Te

104、sla,or that he have fullcontrol over a newly structured for-profit venture,with a majoritystake and CEO title.OpenAI declines,and elects to install Altman asCEO.48February2018OpenAI announces new investments and Musks departure:“Werebroadening our base of funders to prepare for the next phase ofOpen

105、AI,which will involve ramping up our investments in our people48Ibid.47“OpenAI and Elon Musk.”46“OpenAI and Microsoft,”OpenAI,November 15,2016.My emphasis.45“OpenAI Technical Goals,”OpenAI,June 20,2016.My emphasis.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model17and

106、 the compute resources necessary to make consequentialbreakthroughs in artificial intelligence.”49April 2018OpenAI releases its charter,centering for the first time the term“AGI.”The charter,which still appears on OpenAIs website today,states:“OpenAIs mission is to ensure that artificial general int

107、elligence(AGI)by which we mean highly autonomous systems thatoutperform humans at most economically valuable workbenefits allof humanity.We will attempt to directly build safe and beneficial AGI,but will also consider our mission fulfilled if our work aids others toachieve this outcome.”The core pla

108、nks of the charter include“broadly distributed benefits,”“long-term safety,”“technicalleadership,”and“cooperative orientation.”The statement notes:“Ourprimary fiduciary duty is to humanity.”50June2018OpenAI releases an early GPT language model.51February2019OpenAI releases a research paper detailing

109、 the results of its latesttext-generation model,yet allows tech reporters to demo thetechnology.The result is a round of PR for OpenAI that furtherpositions AI as dangerous,and the company as an unusually ethicalsteward of said technology.A representative headline,from Wired,reads:“The AI Text Gener

110、ator Thats Too Dangerous to Make Public.”The article begins:“In 2015,car-and-rocket man Elon Musk joinedwith influential startup backer Sam Altman to put artificial intelligenceon a new,more open course.They cofounded a research institutecalled OpenAI to make new AI discoveries and give them away fo

111、r thecommon good.Now,the institutes researchers are su?cientlyworried by something they built that they wont release it to thepublic.”52This announcement generated significantly more interest inOpenAI than any previous news cycle,besides the companysfounding.52Tom Simonite,“The AI Text Generator Tha

112、ts Too Dangerous to Make Public,”Wired,February 14,2019.51“Improving Language Understanding with Unsupervised Learning,”OpenAI,June 11,2018.50“OpenAI Charter,”OpenAI,April 9,2018;accessed November 1,2024.My emphasis.49“OpenAI Supporters,”OpenAI,February 20,2018.AI Generated Business:The Rise of AGI

113、andthe Rush to Find a Working Revenue Model18March2019The next month,OpenAI announces it is transitioning from a nonprofitresearch company into“a new capped-profit company that allows usto rapidly increase our investments in compute and talent whileincluding checks and balances to actualize our miss

114、ion.”53Theannouncement elaborates:“Our mission is to ensure that artificialgeneral intelligence(AGI)benefits all of humanity,primarily byattempting to build safe AGI and share the benefits with the world.Weve experienced firsthand that the most dramatic AI systems usethe most computational power in

115、addition to algorithmic innovations,and decided to scale much faster than wed planned when startingOpenAI.Well need to invest billions of dollars in upcoming years intolarge-scale cloud compute,attracting and retaining talented people,and building AI supercomputers.We want to increase our ability to

116、raise capital while still serving our mission,and no pre-existing legalstructure we know of strikes the right balance.Our solution is tocreate OpenAI LP as a hybrid of a for-profit and nonprofit.”54To recap:OpenAI publicly launched as a nonprofit with theself-described mission of developing a“safe,”

117、“human-level,”open-sourced AI that would“benefit all of humanity,”emphasizing ithad no financial incentive.Just before it announces a restructuringinto a for-profit entity,OpenAI publishes a charter centering itsmission as bringing about“AGI,”explicitly casting this as the primarygoal.After the spec

118、ter of AGI is raised,OpenAI releases to the press atext model it says is too dangerous to release to the public,taking fulladvantage of the social imaginary of a looming AGI.The presspublicizes both the model,which is impressive,and OpenAIsprudence in exercising caution over its own creation,making“

119、the AIthats too dangerous to release”a phenomenon in the tech media.Months later,OpenAI announces a$1 billion deal with Microsoft,againtouting its pursuit of AGI.July 2019By the summer,“AGI”is being splashed in the headline of one ofOpenAIs highest-profile press releases yet.Announcing a$1 billionpa

120、rtnership with Microsoft,OpenAI says:“Microsoft invests in andpartners with OpenAI to support us building beneficial AGI.”5555“Microsoft Invests in and Partners with OpenAI to Support Us Building Beneficial AGI,”OpenAI,July 22,2019.My emphasis.54Ibid.My emphasis.53“OpenAI LP,”OpenAI,March 11,2019.AI

121、 Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model19July 2019Note the language,and the shift away from centering“safety”toemphasizing the power of a(still speculative)AGI:AGI will be a system capable of mastering a field of study to theworld-expert level,and mastering mo

122、re fields than any onehumanlike a tool which combines the skills of Curie,Turing,and Bach.An AGI working on a problem would be able to seeconnections across disciplines that no human could.We wantAGI to work with people to solve currently intractablemulti-disciplinary problems,including global chall

123、enges suchas climate change,a?ordable and high-quality healthcare,andpersonalized education.We think its impact should be to giveeveryone economic freedom to pursue what they find mostfulfilling,creating new opportunities for all of our lives that areunimaginable today.OpenAI is producing a sequence

124、 of increasingly powerful AItechnologies,which requires a lot of capital for computationalpower.The most obvious way to cover costs is to build aproduct,but that would mean changing our focus.Instead,weintend to license some of our pre-AGI technologies,withMicrosoft becoming our preferred partner fo

125、r commercializingthem.We believe that the creation of beneficial AGI will be the mostimportant technological development in human history,withthe potential to shape the trajectory of humanity.We have ahard technical path in front of us,requiring a unified softwareengineering and AI research e?ort of

126、 massive computationalscale,but technical success alone is not enough.To accomplishour mission of ensuring that AGI(whether built by us ornot)benefits all of humanity,well need to ensure that AGIis deployed safely and securely;that society is well-preparedfor its implications;and that its economic u

127、pside is widelyshared.If we achieve this mission,we will have actualizedMicrosoft and OpenAIs shared value of empowering everyone.5656Ibid.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model20December2020OpenAI co-founder Dario Amodei leaves the company,reportedly overc

128、oncerns about the Microsoft partnership.In an update announcingAmodeis departure,Altman notes that“OpenAIs mission is tothoughtfully and responsibly develop general-purpose artificialintelligence,and as we enter the new year our focus onresearchespecially in the area of safetyhas never been stronger

129、.Making AI safer is a company-wide priority,and a key part of MiraMuratis new role.”57January2021OpenAI releases the text-to-image generator DALL-E.Theannouncement states:“We recognize that work involving generativemodels has the potential for significant,broad societal impacts.In thefuture,we plan

130、to analyze how models like DALLE relate to societalissues like economic impact on certain work processes andprofessions,the potential for bias in the model outputs,and thelonger term ethical challenges implied by this technology.”58May 2021Amodei announces his new company Anthropic with a historic$1

131、24million investment,the largest sum thus far raised for a companytrying to build all-purpose AI.The company describes itself as“an AIsafety and research company thats working to build reliable,interpretable,and steerable AI systems.”59Amodei says he aims to“deploy these systems in a way that benefi

132、ts people.”60As Anthropicbecomes OpenAIs largest non-FAANG competitor,it uses“AI safety”as a guiding principle and marketing term.61April2022OpenAI releases DALL-E 2.“This is not a product,”Mira Murati,OpenAIs head of research,tells the New York Times.“The idea isunderstand sic capabilities and limi

133、tations and give us the61This might be seen as what will become something of an“AI safety”arms race,in which new entrants compete to declare theirproduct the safest vehicle for achieving AI.For the culmination of this trend,see Ilya Sutskever,Daniel Gross,and Daniel Levy,SafeSuperintelligence Inc.,J

134、une 19,2024,accessed November 1,2024.60“Anthropic Raises$124 Million to Build More Reliable,General AI Systems,”Anthropic,May 28,2021.59“Anthropic:An AI Safety and Research Company Thats Working to Build Reliable,Interpretable,and Steerable AI Systems,”WelcomeAI,May 2021,accessed November 1,2024.58“

135、DALL-E:Creating Images from Text,”OpenAI,January 5,2021.57“Organizational Update from OpenAI,”OpenAI,December 29,2020.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model21opportunity to build in mitigation.”62The AI image generator iscelebrated in the press.63November20

136、22OpenAI releases ChatGPT.The language in the launch announcementis more retail-orientedthe culmination,perhaps,of a yearslongmarketing e?ort to generate interest in what is now clearly touted asa commercial product:“We are excited to introduce ChatGPT to getusers feedback and learn about its streng

137、ths and weaknesses.Duringthe research preview,usage of ChatGPT is free.”64The companys reputation as the vanguard of responsible AIdevelopmentthis was the company that was built to protecthumankind from malicious AI,after allmeant that the release of theproduct was treated with extra gravity by the

138、press.“We are notready,”Kevin Roose of the New York Times writes about thetext-generating chatbot product.65Marketing AGI,ShippingCommercial AIOpenAI launched as a nonprofit positioned to fulfill a business objective,was from thestart rich with profit-making potential,and deployed marketing terms to

139、 create a narrativeabout its altruistic pursuit of building a“safe”“AGI.”It is immaterial whether or not theparticipants in this venture believe in the risks of AImany clearly dobut these risksserve marketing purposes regardless.If an AI product may be all-powerful,it is all themore alluring to exec

140、utives and managers who hope to harness it to increase productivityand save labor costs.From the start,OpenAI has positioned itself to be the only reliablesteward of this dangerous AGI-adjacent technologyand so we should trust only themwhen the time comes to sell it.Tracing the usage of the term“AGI

141、”in OpenAIs marketing materials,patterns emerge.Theterm is most often deployed at crucial junctures in the companys fundraising history,orwhen it serves the company to remind the media of the stakes of its mission.OpenAI firstmade AGI a focus of o?cial company business in 2018,when it released its c

142、harter,just65Kevin Roose,“The Brilliance and Weirdness of ChatGPT,”New York Times,December 5,2022.64“Introducing ChatGPT,”OpenAI,November 30,2022.63For examples,complete with adulatory headlines,see Casey Newton,“How DALL-E Could Power a Creative Revolution,”Verge,June10,2022;and Mark Sullivan,“Open

143、AIs DALL-E AI Is Becoming a Scary-Good Graphic Artist,”Fast Company,April 6,2022.62Cade Metz,“Meet DALL-E,the A.I.That Draws Anything at Your Command,”New York Times,April 6,2022.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model22after it announced Elon Musks departur

144、e;and again as it is negotiating investment fromMicrosoft and preparing to restructure as a for-profit company.In January 2023,just over a month after the launch of ChatGPT,when the technology wasstill rapidly ascendentit would be the fastest app to climb to 100 million users,accordingto the company

145、OpenAI and Microsoft announced a$10 billion partnership that OpenAIstated“will allow us to continue our independent research and develop AI that isincreasingly safe,useful,and powerful.”66There is no mention of AGI explicitly.Twomonths later,when ChatGPT was su?ciently publicized and perhaps even be

146、ginning toplateau in user growth,and when Microsoft had an interest in publicizing its ownChatGPT-powered products,its researchers published a paper announcing that ChatGPT4 showed“sparks”of AGI.67OpenAI is nearly the precise inverse of what it initially promised to beinstead of opensource and nonpr

147、ofit,structured to avoid any single person taking too much power overthis supposedly all-powerful technology,it has evolved into a company that is highlyproprietary and explicitly for-profit,with a single person,CEO Sam Altman,exertingnear-total control over the operation.Yet this is little remarked

148、 upon in the media,or bybusiness analysts,with the somewhat ironic exception of Elon Musk,who filed a suitalleging that by transitioning into a for-profit,non-open source entity,OpenAI violated theterms of his initial funding.OpenAI burst into public consciousness and into the headlines in late 2022

149、 after its publicrelease of ChatGPT,which would come to be synonymous with the new booming productcategory of generative AI.Throughout 2023,Silicon Valleyand many other industries,ledby consultancies and financereoriented itself around the most-hyped technology it hadseen in years:chatbots,image gen

150、erators,audio synthesizers,and other AI productsrolled out from startups and established giants.AI became the single dominant trend intech,and bent all other products to its arc.Generative AI caught the publics eye with flashy tech demos,which OpenAI executivesrapidly translated into hype for the ne

151、w services,and for the power of AI in general.Thehype reached apocalyptic levels,replete with Altman making a tour through Washingtonand European capitals,apparently leading a public charge for regulation;testifying to thepower of the technology;and making himself the point person for corporate AI p

152、olicy.Themythology constructed since OpenAIs founding can be seen paying dividends throughout2023.67See Chloe Xiang,“Microsoft Now Claims GPT-4 Shows Sparks of General Intelligence,”Motherboard,March 24,2023;and SbastienBubeck,Varun Chandrasekaran,Ronen Eldan,Johannes Gehrke,Eric Horvitz,Ece Kamar,P

153、eter Lee,Yin Tat Lee,Yuanzhi Li,ScottLundberg,Harsha Nori,Hamid Palangi,Marco Tulio Ribeiro,and Yi Zhang,“Sparks of Artificial General Intelligence:Early Experimentswith GPT-4,”arXiv,March 22,2023.66Cade Metz and Karen Weise,“Microsoft to Invest$10 Billion in OpenAI,the Creator of ChatGPT,”New York

154、Times,January 23,2023.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model23The AGI construct can also be used as a device for blunting scandals,or for remindinginvestors of OpenAIs ultimate business objective.For instance,in May 2024,when OpenAIendured a string of its m

155、ost stinging scandalsthe most notable being ScarlettJohansson accusing the company of mimicking her voice without consent68thecompany responded by announcing a new safety council and a new“frontier model”tohelp create AGI:“OpenAI has recently begun training its next frontier model and weanticipate t

156、he resulting systems to bring us to the next level of capabilities on our path toAGI.”69By 2024,the evolution of AGI into purely a marketing term,disconnected entirely from theresearch sphere,was complete.OpenAI hailed the launch of an automated customersupport company called MavenAGI,built with its

157、 GPT technology,on the o?cial companywebsite:“MavenAGI is a new software company for the AI era.”70Over the decades“AI”had become rote,familiar,and so broadly understood as to be meaningless.AGI could beseen as a marketing-friendly reclamation of the concept to di?erentiate a new raft ofproducts fro

158、m a decades-old discipline.And if AGI was the marketing driver,then generative AI was the demo ideally suited tocommunicate its potential.Its“magical”autogenerated images,uncanny-valley videos,and a chatbot that mimicked human user behavior in its presentation(with its flickeringcursor and sentences

159、 being typed out in real time as if by a supersmart person behind thescenes)gave the impression that this was humanity,augmented.The success of theDALL-E and ChatGPT demos meant that the early business interest would be inautomating text,code,and image generationand in automating creative labor.It w

160、ouldincentivize AI companies to shape their business models to capitalize on this interest,andto conform their pursuits of enterprise deals to this narrative.OpenAI helped construct an imaginary where an AGI of great power looms on thehorizonone that can do anything a human can,at least in the digit

161、al worldand o?ered asuite of demonstrations as evidence.For all the wonder they sparked,the clearest path toa business model seemed to be articulating the capacity for AGI to automate human labor.Indeed,that was how it was redefined in OpenAIs 2018 charter:AGI remains defined byOpenAI as“highly auto

162、nomous systems that outperform humans at most economicallyvaluable work.”71While it was still unclear as to what most generative AI companiesprecise business models would be,investment was unleashed nonetheless.71“OpenAI Charter.”My emphasis.70“MavenAGI Launches Automated Customer Support Agents Pow

163、ered by OpenAI,”OpenAI,n.d.,accessed November 1,2024.69“OpenAI Board Forms Safety and Security Committee,”OpenAI,May 28,2024.68Sarah Krouse,Deepa Seetharaman,and Joe Flint,“Behind the Scenes of Scarlett Johanssons Battle With OpenAI,”Wall StreetJournal,May 23,2024.AI Generated Business:The Rise of A

164、GI andthe Rush to Find a Working Revenue Model24Investment in“AI”from 2010 to 2015,according to CB Insights,totalled less than$1 billioncombined,though it did increase dramatically,from$45 million in 2010 to$310 million in2015.72Over the 2010s,investment in AI was ongoing,if targeted at“narrow”AI co

165、mpaniesaimed at generating production and manufacturing e?cienciesnotable startupsincluded Vicariousbut it was a mere fraction of what the boom of 2023 would unleash.According to an OECD report,“between 2015 and 2023,global VC investments in AIstart-ups tripled(from USD 31 billion to USD 98 billion)

166、,with investments in generative AIspecifically growing from 1%of total AI VC investments in 2022(USD 1.3 billion)to 18.2%(USD 17.8 billion)in 2023,despite cooling capital markets.”73High-profile investors include Andreessen Horowitz,or a16z,the largest VC firm by assetsunder management($42 billion).

167、A16z participated in OpenAIs$300 million fundraise inApril 2023,and has since started a$6 billion fund focused on AI companies and“Americandynamism.”74It has invested in twenty AI startups;according to Pitchbook,“a16z hasmade a play to own every part of the generative AI value chain.”75Most notably,

168、it hasinvested$415 million in a Series A round for French open source AI startup Mistral.The Valleys second-largest firm,Sequoia Capital,invested in OpenAI in 2021,and in 2023also unleashed a blitz of AI investment:$30 million in Collaborative Robotics Series A,$28million in Hex Technologies,$21 mil

169、lion in the legal AI firm Harvey,$15M in CloseFactor,$12.2 million in generative video startup Tavus,$5.5 million in the AI assistant companyDust,$5.3 in Replicate,and$3.2 in Helm,a health AI startup.The firm says it has seventyAI companies in its portfolio.76The Dream of AGI and the FullyAutomated

170、OrganizationIn sum,OpenAI was likely conceived as a strategic hedge by Elon Musk and Sam Altmanagainst Googles mounting claim to commercial AI dominance.It was positioned from thestart to communicate its altruistic mission contra the for-profit objectives of the techgiants of the day,despite itself

171、being backed by and employed by figures from the largestSilicon Valley firms and venture capital outlets.By deploying the use of“AGI”to elevate itsmission beyond the stale concern of generations-old“AI,”it was able to cultivate a cultural76Gen Teare,“Inside Sequoia Capitals Growing AI Portfolio,”Cru

172、nchbase News,October 23,2023.75Jacob Robbins,“Andreessen Horowitzs$415M Mistral investment Rounds Out AI Strategy,”PitchBook,December 10,2023.74Sam Blum,“A16z Targets AI as Part of New$6 Billion Fund,Days After Marc Andreessen Called OpenAIs Security Protocol SwissCheese,”Inc.,March 7,2024.73Celine

173、Caira,“The Future of Artificial Intelligence,”in OECD Digital Economy Outlook 2024(Volume 1):Embracing the TechnologyFrontier,OECD,2024.72“Deep Interest in AI:New High in Deals to Artificial Intelligence Startups in Q4 15,”CB Insights,February 4,2016.AI Generated Business:The Rise of AGI andthe Rush

174、 to Find a Working Revenue Model25mystique that translated into a marketing device that could be wielded to reliablygenerate press,bolster recruitment of top researchers,and attract investment at crucialcorporate junctures.Thanks to the investment climate and the zero-percent-interest-rate era,there

175、 was littleconcern over a business model being in place as OpenAIs leadership transitioned into afor-profitthe importance of a strong narrative and the involvement of central Valleyfigures eclipsed the need for a plan to generate revenue.A precedent had been set byUber and other“unicorns”that invest

176、ors and consumers would tolerate long-term e?ortsto scale and to capture markets.OpenAI was in command of easily the most compellingstory about AI,and thus when its product ChatGPT exploded in popularity,it was wellpoised to capitalize on its recent legacy of presenting itself as a steward of AGI,wh

177、ich thecompany was itself allegedly developing.Early competitors like Anthropic would have to compete on grounds of“safety,”but thismattered less as the products reached cultural saturation and market interest becameobvious in the generative AI product category,especially from enterprise clients.Now

178、,with widespread industry dedication and investment,and OpenAI successfully at thecenter of the generative AI economy,a rush to discern what generative AI is actuallycapable of in a business context unfoldedand continues to unfold even now.It appears certain that ChatGPTs rapid,whiplash success was

179、something of a surprise toboth the company and its partners.It was not part of a meticulous product roadmapinfact,OpenAI board member Helen Toner has revealed that the board was not eveninformed in advance about ChatGPTs release;They found out,to their chagrin,on Twitteralong with everyone else.77Th

180、e app quickly notched 100 million users,causing industryanalysts at banks and investment firms to swoon.“In 20 years following the internetspace,we cannot recall a faster ramp in a consumer internet app,”UBS analysts wrote in aFebruary 2023 report.78To keep up with demandespecially in server space,g

181、iven the taxing computerequirements of running AI systemsOpenAI and Microsoft expanded their partnership,with the software giant pledging an additional$10 billion in exchange for the right to useOpenAI technology in its platforms and software o?erings,and a cut of OpenAIs sales.79(To those keeping t

182、rack at home,that brought the total to around$13 billion,given pastinvestments and pledges.)79“Microsoft and OpenAI Extend Partnership,”Microsoft(blog),January 23,2023.78Krystal Hu,“ChatGPT Sets Record for Fastest-Growing User Base Analyst Note,”Reuters,February 2,2023.77Helen Toner,interview by Bil

183、awal Sidhu,The TED AI Show,May 2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model26The details of this partnership have not been made public,but it is known to stake muchof its value in the form of compute credits for Azure,Microsofts cloud compute service.80It al

184、so requires Open AI to use Azure as its compute provider.OpenAIs embrace ofMicrosoftand vice versaas it moved away from a research organization aimed at“protecting the world from AI”as per the early Altman-Musk iteration,signals the startupsambition to take a more capital-intensive approach than Mus

185、k was able to o?er.There aremany reasons why Microsoft might find a deal attractive,and why its ideally structured tobe tolerant of a startup that has no immediate or obvious roads to profitability.First,sinceits earliest investments,OpenAI can be seen as a hedge against Google and Facebook,each of

186、which were outspending Microsoft in-house to develop AI systems.Second,thestructure of the deal(or what we know of it)ensures that Microsoft will profit from nearlyany outcome.Microsoft benefits from using OpenAIs technology in its products,andreceives a cut of revenue from the products OpenAI sells

187、 as wellbu?ering any financialrisk,which is already curbed because,as noted above,much of the deal is understood totake the form of compute credits.As such,as recently as the summer of 2024,Microsoftwas able to publicly state that it was comfortable with Sam Altmans vision of an incipientAI,and was

188、not afraid to spend billions investing with no promise of profits in the neartermthe deals structure,and the dream of AGI,are in ideal symbiosis.In 2023,after the OpenAI-Microsoft deal was inked,interest in OpenAI,generative AI,andlarge language models truly exploded.Microsoft added GPT technology t

189、o Bing andCopilot;Google announced Bard,which later became Gemini;and Anthropic debutedClaude,its ChatGPT competitor,in March 2023.81(In Q3 2024,half of all venture capitalinvestment in the US went to AI,up from 15 percent in 2017.82)OpenAIs first major move to shape a business model was the obvious

190、 one:It opened apremium paid tier for ChatGPT called Plus,o?ering higher performance for fans andsuperusers of the app.OpenAI also started granting access to its API,allowing developersand companies to purchase API keys.Within months,the company began leaning intoenterprise sales,recognizing the pot

191、ential in touting an all-powerful technology thatcould also automate jobs.That spring,OpenAI employees coauthored a report with Cornellresearchers about job impacts;83While the paper was covered in the press as a warning,ithad the e?ect,as many such impact papers do,of bolstering the automation tech

192、nologysjob-replacing bona fides.Throughout the rest of 2023,OpenAI floated other ideas for generating revenue,somewith more conviction than others.It teased and eventually released an app store for GPT83Tyna Eloundou,Sam Manning,Pamela Mishkin,and Daniel Rock,“GPTs Are GPTs:An Early Look at the Labo

193、r Market Impact Potentialof Large Language Models,”arXiv,March 17,2023.82Rani Molla,“Watch AI Eat the VC World in One Chart,”Sherwood News,July 17,2024.81“Introducing Claude,”Anthropic,March 14,2023.80Zitron,Ed.How Does Open AI Survive?Newsletter,July 29,2024.AI Generated Business:The Rise of AGI an

194、dthe Rush to Find a Working Revenue Model27apps developed by independent programmers,as well as the AI video generator Sora.Altman nodded to the idea of selling ads against GPT output online.But the chief potentialfor revenue generation,and thus the core plank of its business model,remained tied tot

195、he same plank:selling enterprise-tier GPTs and selling API access.Now we might see whythe AGI mythology is so important to the formulation here:Managers can be assured thatthey are purchasing both the safest and the most powerful technology available as theyseek to cut costs in a tight labor market.

196、This is similar to the way in which the previous era of tech unicornsUber and the gigeconomy startupsultimately tried to attain profit by promising to reduce labor costs(inridehails case,by disrupting the taxi and black cab economy)despite the buzzy rhetoricand lofty corporate mythologies.In this ca

197、se,the talk of AGI supersedes discussions ofOpenAIs enterprise business in the press,making it a more palatable automationcompany.Reduce tasks,use as leverage,replace jobs,increase attrition.Nearly two years into the generative AI boom,this still holds true:Enterprise clients,bythe companys own esti

198、mation,are the most valuable line of its business by a considerablemargin.(OpenAIs sales figures arent public,but many of the major contracts inked so farare.)The largest client of OpenAI is the consulting and financial services firm PwC,whichhas purchased 101,000 seats.84The American biotech firm M

199、oderna,85and Klarna,aSwedish fintech startup,are among other leading users;OpenAI estimated 600,000 totalenterprise clients as of the spring of 2024.86By September,it reported one million.87InJune 2024,the company self-reported its annualized revenue as$3.4 billion.88Yet OpenAIs overhead is staggeri

200、ng,making it notably di?erent from the“zero-price”products that have otherwise dominated the modern tech industry.Last summer,thecompanys compute costs were reported to be$700,000 a day;89they are certainly muchlarger now.Sam Altman has said publicly he needs$7 trillion in investment for chips to ru

201、nwhat he had planned for his AI programs.90Meanwhile,OpenAIs workforce is large,expensive,and expanding.Research and development for LLMs requires vast investment,compute,and energyespecially as OpenAI pushes into video production.Licensing dealswith media companies for training data have cost the c

202、ompany hundreds of millions ofdollars.91Multiple legal battles,against journalists and creative workers who allege OpenAI91The licensing deal with NewsCorp alone was worth$250 million.See,e.g.,Alexandra Bruell,Schechner,and Deepa Seetharaman,“OpenAI,WSJ Owner News Corp Strike Content Deal Valued at

203、over$250 Million,”Wall Street Journal,May 22,2024.90Keach Hagey,“Sam Altman Seeks Trillions of Dollars to Reshape Business of Chips and AI,”Wall Street Journal,February 8,2024.89Aaron Mok,“ChatGPT Could Cost over$700,000 per Day to Operate.Microsoft Is Reportedly Trying to Make It Cheaper,”BusinessI

204、nsider,April 20,2023.88Palazzolo and Woo,“OpenAIs Annualized Revenue Doubles to$3.4 Billion Since Late 2023.”87“OpenAI Hits More than 1 Million Paid Business Users,”Reuters,September 5,2024.86Jackie Davalos,“OpenAI Sees Tremendous Growth in Corporate Version of ChatGPT,”Bloomberg,April 4,2024.85“Acc

205、elerating the Development of Life-Saving Treatments,”OpenAI,n.d.,accessed November 1,2024.84“PwC to Become OpenAIs Largest Enterprise Customer amid genAI Boom,”Reuters,May 29,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model28has used their work without consent o

206、r compensation,are ongoing.And competitors areeating into the companys market share.Taken together,we see a portrait of a company that wrapped itself in an altruisticnarrative mythology to attract researchers,investment,and press.It stumbled into a hitapp that opened a pathway to a new product categ

207、ory in commercial generative AI(something Silicon Valley had been pursuing unsuccessfully for years),ignited a gold rush,drew competitors,and wielded its unique legacy and relationship to AGI to di?erentiateitself.However,given that generative AI technology is so expensive to develop and run,aunique

208、 imperative to generate revenuelots of revenuein order to capitalize on itspopularity,cultural cachet,and market opportunity has become the companys dominantconcern.(In the past,as mentioned earlier,OpenAI has stated that its move away fromnonprofit status was necessitated by the need for more compu

209、te power if it were to makesatisfactory progress in creating AGI.This can be seen instead as a move towardpreparing for an era of commercial product releases,even if the company remainedunprepared for the success of ChatGPT when it arrived.)This is how a companytransforms from a nonprofit whose aim

210、is to be“owned by all of humanity”and“free of theprofit motive”to one whose board is purged of safety experts in favor of Larry Summers.This frenzy to locate and craft a viable business model has had other consequences.Ongoing and highly unresolved issues regarding copyright threaten the foundation

211、of theindustry:If content currently used in AI training models is found to be subject to copyrightclaims,top VCs investing in AI like Marc Andreessen say it could destroy the nascentindustry.Governments,citizens,and civil society advocates have had little time to prepareadequate policies for mitigat

212、ing misinformation,AI biases,and economic disruptionscaused by AI.Furthermore,the haphazard nature of the AI industrys rise means that byall appearances,another tech bubble is being rapidly inflated.92So much investment hasflowed into the space so fast,based on the bona fides of just a handful ofcom

213、paniesespecially OpenAI,Microsoft,Google,and Nvidiathat a crash could provehighly disruptive,and have a ripple e?ect far beyond Silicon Valley.This is especially concerning because so much of the business,as we have seen,relies onlarge enterprise contracts and on using generative AI in a capacity as

214、 automationsoftware for creative labor.Given the unreliability,hallucinatory output,and securityconcerns still posed by even enterprise editions of the software,this makes the long-termgambit of generative AI as e?ciency-generating,cost-cutting,andproductivity-improving software a deeply dubious one

215、.Goldman Sachs and SequoiaCapital have both authored reports suggesting that generative AI may not be worth thecurrent investment.Sequoia Capital partner David Cahn wrote that the generative AIindustry would have to generate$600 billion in revenue annually to sustain the current92Sam Blum,“Warnings

216、About an AI Bubble Are Growing.When Could It Burst?”Inc.,July 10,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model29rate of investmenta long way to go when the biggest company in the space is making$3.4 billion a year.93Goldman Sachs was even more blunt,finding t

217、hat there was simplytoo much spend and too little benefit to justify the technology in most corporateenvironments.94The Wall Street Journal reported in July 2024 that firms had switchedfrom discussing generative AI in terms of productivity gainsperhaps in part becausethose gains were dubious or hard

218、 to quantifyto analyzing its capacity for revenuegeneration.95If that doesnt pan outif companies find generative AI is coming up short onrevenueits easy to see a mass dismissal of the technology among corporate firms afterthe trial periods expire and the hype wears o?.Given that generative AI has en

219、tered a new and crucial stagewith large clientacquisitions and investor concerns colliding,critics voices growing,no new majortechnology advancements released for months,and a howling imperative to start makingmoneyperhaps its no surprise that,once again,OpenAI turned to centering its AGInarrative.I

220、n July 2024,OpenAI announced a five-level system for evaluating itstechnologies on the road to human-level intelligence.96According to Bloomberg,“thetiers,which OpenAI plans to share with investors and others outside the company,rangefrom the kind of AI available today that can interact in conversat

221、ional language withpeople(Level 1)to AI that can do the work of an organization(Level 5).”97Notably,OpenAIis still at Level 2its AI is a“reasoner”that can do“human-level problem solving.”This positioning can be viewed as a resetting of expectations to corporate clients whomight be getting itchy afte

222、r seeing a few months with enterprise GPT fail to provideimpressive results,a reminder that AGI is still coming,that the systems are improving allthe timeso dont cancel your contract with us just yetand a framework through whichthe company can whet the appetite of future clients.(Note that Altman ha

223、s had torecalibrate his AGI expectations before;in January 2024,he insisted that AGI is coming,but“will change the world much less than we think,”98perhaps another e?ort to tampdown expectations set by,well,himself,just a few months earlier,as corporate clientsbegan to express frustration with the s

224、low gains made by enterprise AI.)When AGI arrives,OpenAI promises,its AI wont just be able to do a single workers jobitwill be able to do all of their jobs.The generative AI on o?er at OpenAIs Level 5 is an“AIthat can do the work of an organization.”99And that,ultimately,is what executives andmanage

225、ment are pursuing,and have been pursuing,since the industrial revolution,when99Metz,“OpenAI Scale Ranks Progress Toward Human-Level Problem Solving.”98MacKenzie Sigalos and Ryan Browne,“OpenAIs Sam Altman Says Human-Level AI Is Coming but Will Change the World Much Lessthan We Think,”CNBC,January 16

226、,2024.97Ibid.96Rachel Metz,“OpenAI Scale Ranks Progress Toward Human-Level Problem Solving,”Bloomberg,July 11,2024.95Isabelle Bousquette,“Its Time for AI to Start Making Money for Businesses.Can It?”Wall Street Journal,July 10,2024.94“Gen AI:Too Much Spend,Too Little Benefit?”Goldman Sachs,June 27,2

227、024.93David Cahn,“AIs$600B Question,”Sequoia Capital,June 20,2024.AI Generated Business:The Rise of AGI andthe Rush to Find a Working Revenue Model30early entrepreneurs dreamt up the first fully mechanized factoriescompletely automatedoperations producing output for the profit of the one at the helm

228、,without the protestsand ine?ciencies of a human workforce.In fact,what AGIartificial general intelligencethat can carry out human-level task-making free of human ine?cienciesis to AIresearchers,the fully automated factory is to C-suite executives and managementconsultancies:an ultimate ideal that m

229、ay or may not be achievable,and yet serves as apowerful driver of interest and investment.This,perhaps,is also why the apocalyptic nature of the hype around generative AI or aprospective AGI does not seem to have bothered many corporate clients,and why it hasinstead proven to be such an e?ective mar

230、keting tool.It is in many ways the same dream:a“highly autonomous system that outperforms humans at most economically valuablework.”100As it turns out,Sam Altman didnt even have to build his AGI to ask it how it mightgenerate revenueit was clear from the start.AcknowledgmentsCopyediting by Caren Litherland.Design by Partner&Partners.Graph by Roberto Rocha.Cite as Brian Merchant,“AI Generated Business:The Rise of AGI and the Rush to Finda Working Revenue Model”,AI Now Institute,December 2024.100“OpenAI Charter.”