埃維諾(Avanade)首席人工智能官弗洛里安·羅塔爾(Florian Rotar)警告稱,精通生成式人工智能技術(shù)的企業(yè)董事會與需要迎頭趕上的企業(yè)董事會之間的差距正在不斷擴(kuò)大。
羅塔爾在《財富》雜志與Diligent公司合作舉辦的“現(xiàn)代董事會”系列線上對話中表示:"我有些憂慮,我們可能會目睹發(fā)展上的差異,這可能導(dǎo)致一些公司落后。”
信息技術(shù)服務(wù)和咨詢公司埃維諾曾與數(shù)百家企業(yè)合作,在與這些企業(yè)的對話中,羅塔爾發(fā)現(xiàn)有些董事會“在人工智能應(yīng)用方面已經(jīng)形成相當(dāng)成熟的機(jī)制”。它們已經(jīng)實(shí)施了一些用例,例如運(yùn)用生成式人工智能來更好地為董事會會議做準(zhǔn)備,進(jìn)行激進(jìn)投資者試點(diǎn)和原型模擬演練,以及開展人工智能輔助桌面演練,以更有效地規(guī)劃業(yè)務(wù)風(fēng)險。
缺乏適當(dāng)?shù)娜斯ぶ悄苤卫頃l(fā)風(fēng)險
然而,隨著董事會成員在工作流程中引入生成式人工智能,若缺乏適當(dāng)?shù)娜斯ぶ悄苤卫恚赡軙构久媾R風(fēng)險。這包括如何在不暴露公司敏感信息的情況下遵守安全、政策和程序的明確指導(dǎo)方針。過去兩年,尤其是在聊天機(jī)器人ChatGPT首次亮相后,消費(fèi)者對人工智能的興趣激增,迫使雇主迅速制定有關(guān)人工智能安全使用的政策。
當(dāng)時的想法是,無論是否得到管理層的認(rèn)可,員工都會使用生成式人工智能,因此人力資源和信息技術(shù)團(tuán)隊必須制定限制措施、開設(shè)技能提升課程和其他形式的培訓(xùn)項目,并在內(nèi)部設(shè)立人工智能試驗區(qū),以允許員工進(jìn)行安全探索。專家表示,這一邏輯同樣適用于董事會成員。
治理、風(fēng)險和合規(guī)軟件即服務(wù)公司Diligent的首席法務(wù)官兼首席行政官尼提亞·達(dá)斯(Nithya Das)表示:“我認(rèn)為我們看到的情況是,董事會無疑需要更深入地了解基本知識。你應(yīng)該預(yù)設(shè)他們會找到自己的工具??紤]到董事會工作和相關(guān)材料的敏感性,這可能會給你所在的企業(yè)帶來不同的安全和隱私問題?!?/p>
達(dá)斯表示,培訓(xùn)課程能夠助力董事會成員迅速掌握人工智能的相關(guān)知識,這與近年來網(wǎng)絡(luò)安全威脅受到關(guān)注時必須進(jìn)行的教育項目類似。羅塔爾推薦的其中一門課程是斯坦福大學(xué)的“人工智能覺醒:對經(jīng)濟(jì)和社會的影響”。
人工智能正日益成為企業(yè)董事會成員的優(yōu)先事項
Diligent預(yù)覽了研究部門即將發(fā)布的一項調(diào)查報告,該報告顯示,到2025年,生成式人工智能將在美國上市公司董事會成員的優(yōu)先事項列表中排名第六,這一優(yōu)先級僅次于追求增長和優(yōu)化財務(wù),但高于網(wǎng)絡(luò)安全和人力資源規(guī)劃。
雖然排在第六位聽起來并不是很靠前,但達(dá)斯指出,這表明人工智能已成為備受關(guān)注的議題。領(lǐng)導(dǎo)者們?nèi)栽谂υu估他們的管理團(tuán)隊在人工智能應(yīng)用方面的熟練程度,同時解決對數(shù)據(jù)隱私的擔(dān)憂,以及對幻覺(人工智能模型基于不可靠數(shù)據(jù)生成誤導(dǎo)性信息)的關(guān)切。
達(dá)斯表示:“我們確實(shí)認(rèn)為,大多數(shù)董事會和公司目前正處于探索人工智能的初級階段,但他們無疑對人工智能抱有濃厚的興趣。我們預(yù)計這將是2025年持續(xù)關(guān)注的重點(diǎn)領(lǐng)域?!?/p>
家具和家居用品電子商務(wù)零售商Wayfair的首席技術(shù)官費(fèi)奧納·譚(Fiona Tan)表示,即使是在數(shù)字原生企業(yè),管理層也必須向董事會闡明生成式人工智能技術(shù)與已經(jīng)部署的人工智能和機(jī)器學(xué)習(xí)的傳統(tǒng)用例之間的區(qū)別。
譚表示:“董事會實(shí)際上需要意識到預(yù)測性、生成能力、大型語言模型能力以及風(fēng)險之間的一些細(xì)微差別?!被谶@些認(rèn)識,他們可以考慮在什么環(huán)節(jié)部署生成式人工智能。對于Wayfair這樣的公司來說,這可能包括內(nèi)容生成和針對每位特定購物者的需求制作更個性化的內(nèi)容。
譚指出,管理團(tuán)隊肩負(fù)著尋找各種機(jī)會,利用生成式人工智能提升業(yè)務(wù),并向董事會闡明這一愿景的責(zé)任。這還包括密切關(guān)注新興的人工智能初創(chuàng)公司,并構(gòu)建解決方案,優(yōu)先考慮通過收購的方式,而非從頭開始在內(nèi)部構(gòu)建。
尋找顛覆自己公司的方法
譚表示:“對于董事會而言,我們正致力于采用一種由外而內(nèi)的策略。我們需要探索在哪些領(lǐng)域進(jìn)行自我顛覆?”
數(shù)據(jù)和人工智能軟件公司Databricks的首席信息和安全官奧馬爾·卡瓦吉(Omar Khawaji)表示,董事會成員和管理層不應(yīng)將成為人工智能的狂熱用戶與深刻理解這些系統(tǒng)如何運(yùn)作以及如何應(yīng)用于業(yè)務(wù)混為一談。
卡瓦吉說:"事實(shí)上,我經(jīng)常觀察到董事會和其他領(lǐng)導(dǎo)者陷入一個誤區(qū),‘我用過人工智能,我了解它的工作原理。都已經(jīng)過去三個月了,為什么你們尚未魔法般地解決X、Y和Z問題?’”
他將這種對人工智能準(zhǔn)備就緒程度的常見誤判比作在TikTok上觀看烹飪視頻。觀看網(wǎng)紅制作一道菜肴可能僅需幾分鐘,然而在家完成同樣的菜肴卻可能需要耗費(fèi)數(shù)小時之久。
卡瓦吉說:“管理和治理、治理和策劃以及整理數(shù)據(jù)是90%的工作的挑戰(zhàn)所在。其余的挑戰(zhàn)與模型訓(xùn)練和確定合適的用例有關(guān)?!保ㄘ敻恢形木W(wǎng))
譯者:中慧言-王芳
埃維諾(Avanade)首席人工智能官弗洛里安·羅塔爾(Florian Rotar)警告稱,精通生成式人工智能技術(shù)的企業(yè)董事會與需要迎頭趕上的企業(yè)董事會之間的差距正在不斷擴(kuò)大。
羅塔爾在《財富》雜志與Diligent公司合作舉辦的“現(xiàn)代董事會”系列線上對話中表示:"我有些憂慮,我們可能會目睹發(fā)展上的差異,這可能導(dǎo)致一些公司落后?!?/p>
信息技術(shù)服務(wù)和咨詢公司埃維諾曾與數(shù)百家企業(yè)合作,在與這些企業(yè)的對話中,羅塔爾發(fā)現(xiàn)有些董事會“在人工智能應(yīng)用方面已經(jīng)形成相當(dāng)成熟的機(jī)制”。它們已經(jīng)實(shí)施了一些用例,例如運(yùn)用生成式人工智能來更好地為董事會會議做準(zhǔn)備,進(jìn)行激進(jìn)投資者試點(diǎn)和原型模擬演練,以及開展人工智能輔助桌面演練,以更有效地規(guī)劃業(yè)務(wù)風(fēng)險。
缺乏適當(dāng)?shù)娜斯ぶ悄苤卫頃l(fā)風(fēng)險
然而,隨著董事會成員在工作流程中引入生成式人工智能,若缺乏適當(dāng)?shù)娜斯ぶ悄苤卫?,可能會使公司面臨風(fēng)險。這包括如何在不暴露公司敏感信息的情況下遵守安全、政策和程序的明確指導(dǎo)方針。過去兩年,尤其是在聊天機(jī)器人ChatGPT首次亮相后,消費(fèi)者對人工智能的興趣激增,迫使雇主迅速制定有關(guān)人工智能安全使用的政策。
當(dāng)時的想法是,無論是否得到管理層的認(rèn)可,員工都會使用生成式人工智能,因此人力資源和信息技術(shù)團(tuán)隊必須制定限制措施、開設(shè)技能提升課程和其他形式的培訓(xùn)項目,并在內(nèi)部設(shè)立人工智能試驗區(qū),以允許員工進(jìn)行安全探索。專家表示,這一邏輯同樣適用于董事會成員。
治理、風(fēng)險和合規(guī)軟件即服務(wù)公司Diligent的首席法務(wù)官兼首席行政官尼提亞·達(dá)斯(Nithya Das)表示:“我認(rèn)為我們看到的情況是,董事會無疑需要更深入地了解基本知識。你應(yīng)該預(yù)設(shè)他們會找到自己的工具。考慮到董事會工作和相關(guān)材料的敏感性,這可能會給你所在的企業(yè)帶來不同的安全和隱私問題?!?/p>
達(dá)斯表示,培訓(xùn)課程能夠助力董事會成員迅速掌握人工智能的相關(guān)知識,這與近年來網(wǎng)絡(luò)安全威脅受到關(guān)注時必須進(jìn)行的教育項目類似。羅塔爾推薦的其中一門課程是斯坦福大學(xué)的“人工智能覺醒:對經(jīng)濟(jì)和社會的影響”。
人工智能正日益成為企業(yè)董事會成員的優(yōu)先事項
Diligent預(yù)覽了研究部門即將發(fā)布的一項調(diào)查報告,該報告顯示,到2025年,生成式人工智能將在美國上市公司董事會成員的優(yōu)先事項列表中排名第六,這一優(yōu)先級僅次于追求增長和優(yōu)化財務(wù),但高于網(wǎng)絡(luò)安全和人力資源規(guī)劃。
雖然排在第六位聽起來并不是很靠前,但達(dá)斯指出,這表明人工智能已成為備受關(guān)注的議題。領(lǐng)導(dǎo)者們?nèi)栽谂υu估他們的管理團(tuán)隊在人工智能應(yīng)用方面的熟練程度,同時解決對數(shù)據(jù)隱私的擔(dān)憂,以及對幻覺(人工智能模型基于不可靠數(shù)據(jù)生成誤導(dǎo)性信息)的關(guān)切。
達(dá)斯表示:“我們確實(shí)認(rèn)為,大多數(shù)董事會和公司目前正處于探索人工智能的初級階段,但他們無疑對人工智能抱有濃厚的興趣。我們預(yù)計這將是2025年持續(xù)關(guān)注的重點(diǎn)領(lǐng)域。”
家具和家居用品電子商務(wù)零售商Wayfair的首席技術(shù)官費(fèi)奧納·譚(Fiona Tan)表示,即使是在數(shù)字原生企業(yè),管理層也必須向董事會闡明生成式人工智能技術(shù)與已經(jīng)部署的人工智能和機(jī)器學(xué)習(xí)的傳統(tǒng)用例之間的區(qū)別。
譚表示:“董事會實(shí)際上需要意識到預(yù)測性、生成能力、大型語言模型能力以及風(fēng)險之間的一些細(xì)微差別?!被谶@些認(rèn)識,他們可以考慮在什么環(huán)節(jié)部署生成式人工智能。對于Wayfair這樣的公司來說,這可能包括內(nèi)容生成和針對每位特定購物者的需求制作更個性化的內(nèi)容。
譚指出,管理團(tuán)隊肩負(fù)著尋找各種機(jī)會,利用生成式人工智能提升業(yè)務(wù),并向董事會闡明這一愿景的責(zé)任。這還包括密切關(guān)注新興的人工智能初創(chuàng)公司,并構(gòu)建解決方案,優(yōu)先考慮通過收購的方式,而非從頭開始在內(nèi)部構(gòu)建。
尋找顛覆自己公司的方法
譚表示:“對于董事會而言,我們正致力于采用一種由外而內(nèi)的策略。我們需要探索在哪些領(lǐng)域進(jìn)行自我顛覆?”
數(shù)據(jù)和人工智能軟件公司Databricks的首席信息和安全官奧馬爾·卡瓦吉(Omar Khawaji)表示,董事會成員和管理層不應(yīng)將成為人工智能的狂熱用戶與深刻理解這些系統(tǒng)如何運(yùn)作以及如何應(yīng)用于業(yè)務(wù)混為一談。
卡瓦吉說:"事實(shí)上,我經(jīng)常觀察到董事會和其他領(lǐng)導(dǎo)者陷入一個誤區(qū),‘我用過人工智能,我了解它的工作原理。都已經(jīng)過去三個月了,為什么你們尚未魔法般地解決X、Y和Z問題?’”
他將這種對人工智能準(zhǔn)備就緒程度的常見誤判比作在TikTok上觀看烹飪視頻。觀看網(wǎng)紅制作一道菜肴可能僅需幾分鐘,然而在家完成同樣的菜肴卻可能需要耗費(fèi)數(shù)小時之久。
卡瓦吉說:“管理和治理、治理和策劃以及整理數(shù)據(jù)是90%的工作的挑戰(zhàn)所在。其余的挑戰(zhàn)與模型訓(xùn)練和確定合適的用例有關(guān)。”(財富中文網(wǎng))
譯者:中慧言-王芳
There’s a growing disparity in organizations with boardrooms that are well versed in generative artificial intelligence and those that need to play catchup, warns Florian Rotar, chief AI officer at Avanade.
“I’m a little bit worried that we’ll see this building divergence, and some will be left behind,” says Rotar, during a virtual conversation hosted by Fortune in partnership with Diligent for The Modern Board series.
Avanade, an IT services and consulting firm, has worked with hundreds of organizations and in those conversations found some boardrooms are getting “quite sophisticated in terms of using AI themselves,” Rotar says. Some use cases that have been deployed include relying on generative AI to better prepare for board meetings, piloting and prototyping simulated activist investor exercises, and AI-enabled tabletop exercise to better plan for risks to the business.
Risks without proper AI governance
But as board members dive into applying generative AI to their workflows, it could present some risks to companies if the proper AI governance isn’t in place. That includes clear guidelines on how to adhere to safety, policies, and procedures without exposing sensitive company information. Over the past two years, employers have had to quickly set up policies about safe AI use, especially after the explosion of consumer interest in AI following the debut of chatbot ChatGPT.
The thinking was that employees were going to use generative AI whether it was blessed by management or not, so HR and IT teams had to set up restrictions, upskill classes, and other forms of training, as well as internal AI playgrounds to allow for safe exploration. Experts say that the same logic should apply to board members too.
“I think what we’re seeing is there’s definitely a need for a more fundamental understanding of the basics with the board,” says Nithya Das, chief legal officer and chief administrative officer at Diligent, a governance , risk and compliance SaaS company. “You should assume that they are going to find their own tools, and that may raise different security and privacy concerns for you as an organization, just given the sensitivity of board work and board materials.”
Das says training classes can be helpful to get boards up to speed on AI, similar to the education that had to be done when cybersecurity threats came into focus in recent years. One such course, recommended by Rotar, is Stanford University’s “The AI Awakening: Implications for the Economy and Society.”
AI is a growing priority for corporate directors
Diligent previewed a survey it will soon publish from the company’s research arm showing that generative AI will rank sixth on the priority list for board directors at U.S.-based public companies in 2025, trailing behind pursuing growth and optimizing financials, but higher than cybersecurity and workforce planning.
Sixth may not sound very high, but Das says it is an indication that AI is top of mind. Leaders are still sorting out how well versed their management team is on AI, working through concerns about data privacy, and worries about hallucinations, which can occur when AI models generate incorrect information based on unsound data.
“We do think that most boards and companies are at the beginning of their AI journeys, but they’re definitely very AI curious,” says Das. “We expect to see this to be a continued area of focus for 2025.”
Fiona Tan, chief technology officer at Wayfair, an e-commerce furniture and home goods retailer, says even at digitally native companies, management had to explain to their boards the difference between generative AI technologies and more traditional use of AI and machine learning that was already deployed.
“For a board, it’s actually realizing some of the nuances between what was predictive…what are the generative capabilities, what are the large language model capabilities, and what are the risks,” says Tan. From that point, they can think through where to deploy generative AI. For a company like Wayfair, that could include content generation and making more personalized content for each specific shopper’s needs.
The management team, Tan says, should be responsible for looking at the various opportunities to enhance the business with generative AI and articulate that vision to the board. That should also include a close eye at AI startups that are emerging and building solutions that may be better to buy rather than build internally from scratch.
Looking for ways to disrupt your own company
“For the board, it is pushing to ensure that we are taking a little bit of an outside-in approach,” says Tan. “Where do we need to go in and disrupt ourselves?”
Omar Khawaji, chief information and security officer at Databricks, a data and AI software company, says board members and management should not conflate being an avid user of AI with actually understanding how these systems work and can be applied to the business.
“In fact, a trap that I often see boards and other leaders falling into is, ‘I’ve used AI, I know how it works, it’s been three months, why haven’t you magically solved problems x, y, and z,’” says Khawaji.
He likens this common miscalculation on AI readiness to watching a cooking video on TikTok. It may only take a few minutes to watch a dish get whipped up by an influencer, but to do the same task at home could take hours.
“The challenge of managing and governance, governing and curating, and organizing your data is where 90% of the work happens,” says Khawaji. The rest, he says, is related to training a model and leveraging it with the appropriate use cases.