現(xiàn)在,回避型高管們可以把人工智能加入其財(cái)報(bào)電話會(huì)議的審查者名單。
據(jù)位于倫敦的投資策略師約阿希姆·克萊門特報(bào)道,德國(guó)明斯特大學(xué)(University of Münster)和華盛頓大學(xué)圣路易斯分校(Washington University in St. Louis)的研究人員發(fā)現(xiàn),分析師可以利用大語言模型來推斷高管們?cè)谪?cái)報(bào)電話會(huì)議上有沒有開誠(chéng)布公。
在研究過程中,研究人員將財(cái)報(bào)電話會(huì)議記錄輸入OpenAI的GPT-4 Turbo,以判斷高管的評(píng)論是“尋?!边€是“不尋常”??巳R門特解釋說,當(dāng)人工智能識(shí)別出不尋常的財(cái)報(bào)電話會(huì)議記錄時(shí),它就會(huì)列出自己的理由。
克萊門特表示,在一個(gè)例子中,人工智能分析了會(huì)議記錄,并警告管理層可能在分析師的追問下仍在回避具體細(xì)節(jié)。當(dāng)發(fā)言人提供冗長(zhǎng)的回應(yīng)和長(zhǎng)時(shí)間討論非財(cái)務(wù)話題時(shí),人工智能最常標(biāo)記出“不尋?!保@正是高管們回避負(fù)面新聞的最常見方式。
自2022年ChatGPT發(fā)布以來,金融業(yè)已越來越多地將人工智能融入其日常運(yùn)營(yíng)。今年早些時(shí)候,摩根大通(JPMorgan Chase)推出了一款人工智能工具,幫助解讀美聯(lián)儲(chǔ)主席杰羅姆·鮑威爾口中模棱兩可的“美聯(lián)儲(chǔ)術(shù)語”。盡管如此,人工智能在選股方面產(chǎn)生重大影響的結(jié)果尚未得到證實(shí)。
最終,研究人員認(rèn)為,人工智能通過更好地理解財(cái)報(bào)電話會(huì)議,可以幫助分析師預(yù)測(cè)市場(chǎng)。
研究人員在研究摘要中寫道:“股票市場(chǎng)會(huì)對(duì)不尋常的金融信息做出負(fù)面反應(yīng),交易活動(dòng)增加。當(dāng)一家公司被識(shí)別出更多異常信息時(shí),這種反應(yīng)就會(huì)加劇。”
克萊門特指出,隨著5月份OpenAI推出GPT-4o,分析師也可以將財(cái)報(bào)電話會(huì)議音頻上傳到大語言模型,或在直播過程中進(jìn)行部署。
雖然分析師經(jīng)常會(huì)自己捕捉到不尋常的言論,但這項(xiàng)技術(shù)可以讓他們同時(shí)分析不同行業(yè)的多個(gè)財(cái)報(bào)電話會(huì)議。
克萊門特寫道:“就我個(gè)人而言,我喜歡它,但作為企業(yè)高管,我可能不會(huì)喜歡它,因?yàn)镃hatGPT讓欺騙和分散注意力變得更加困難?!保ㄘ?cái)富中文網(wǎng))
譯者:劉進(jìn)龍
審校:汪皓
現(xiàn)在,回避型高管們可以把人工智能加入其財(cái)報(bào)電話會(huì)議的審查者名單。
據(jù)位于倫敦的投資策略師約阿希姆·克萊門特報(bào)道,德國(guó)明斯特大學(xué)(University of Münster)和華盛頓大學(xué)圣路易斯分校(Washington University in St. Louis)的研究人員發(fā)現(xiàn),分析師可以利用大語言模型來推斷高管們?cè)谪?cái)報(bào)電話會(huì)議上有沒有開誠(chéng)布公。
在研究過程中,研究人員將財(cái)報(bào)電話會(huì)議記錄輸入OpenAI的GPT-4 Turbo,以判斷高管的評(píng)論是“尋?!边€是“不尋?!???巳R門特解釋說,當(dāng)人工智能識(shí)別出不尋常的財(cái)報(bào)電話會(huì)議記錄時(shí),它就會(huì)列出自己的理由。
克萊門特表示,在一個(gè)例子中,人工智能分析了會(huì)議記錄,并警告管理層可能在分析師的追問下仍在回避具體細(xì)節(jié)。當(dāng)發(fā)言人提供冗長(zhǎng)的回應(yīng)和長(zhǎng)時(shí)間討論非財(cái)務(wù)話題時(shí),人工智能最常標(biāo)記出“不尋?!?,而這正是高管們回避負(fù)面新聞的最常見方式。
自2022年ChatGPT發(fā)布以來,金融業(yè)已越來越多地將人工智能融入其日常運(yùn)營(yíng)。今年早些時(shí)候,摩根大通(JPMorgan Chase)推出了一款人工智能工具,幫助解讀美聯(lián)儲(chǔ)主席杰羅姆·鮑威爾口中模棱兩可的“美聯(lián)儲(chǔ)術(shù)語”。盡管如此,人工智能在選股方面產(chǎn)生重大影響的結(jié)果尚未得到證實(shí)。
最終,研究人員認(rèn)為,人工智能通過更好地理解財(cái)報(bào)電話會(huì)議,可以幫助分析師預(yù)測(cè)市場(chǎng)。
研究人員在研究摘要中寫道:“股票市場(chǎng)會(huì)對(duì)不尋常的金融信息做出負(fù)面反應(yīng),交易活動(dòng)增加。當(dāng)一家公司被識(shí)別出更多異常信息時(shí),這種反應(yīng)就會(huì)加劇。”
克萊門特指出,隨著5月份OpenAI推出GPT-4o,分析師也可以將財(cái)報(bào)電話會(huì)議音頻上傳到大語言模型,或在直播過程中進(jìn)行部署。
雖然分析師經(jīng)常會(huì)自己捕捉到不尋常的言論,但這項(xiàng)技術(shù)可以讓他們同時(shí)分析不同行業(yè)的多個(gè)財(cái)報(bào)電話會(huì)議。
克萊門特寫道:“就我個(gè)人而言,我喜歡它,但作為企業(yè)高管,我可能不會(huì)喜歡它,因?yàn)镃hatGPT讓欺騙和分散注意力變得更加困難?!保ㄘ?cái)富中文網(wǎng))
譯者:劉進(jìn)龍
審校:汪皓
Evasive C-suite execs can now add AI to the list of those scrutinizing their earnings calls.
A study from researchers at Germany’s University of Münster and Washington University in St. Louis found that analysts can use large language models to infer when executives aren’t being forthcoming on earnings calls, reported London-based investment strategist Joachim Klement.
During the study, the researchers fed earnings call transcripts into OpenAI’s GPT-4 Turbo to gauge whether an executive’s comments were “usual” or “unusual.” When the AI identifies an unusual earnings call transcript, it will lay out its reasoning, Klement explains.
In one example, the AI analyzed the transcript and flagged that management was possibly avoiding specifics, despite being pressed by analysts, according to Klement. The AI flagged “unusualness” most often when speakers offered lengthy responses and long discussions about non-financial topics, some of the most common ways execs deflect negative news.
Since the launch of ChatGPT in 2022, the finance industry has increasingly incorporated AI into more of its day-to-day operations. Earlier this year JPMorgan Chase unveiled an AI tool to help interpret the ambiguous “Fedspeak” spouted by Chairman Jerome Powell. Still, the results of AI making a significant difference in stock picking is yet unproven.
Ultimately, the researchers believe AI could help analysts predict the market, according to the researchers by getting a better handle on earnings call
“The stock market reacts negatively to unusual financial communication, with an elevated trading activity,” the researchers wrote in the study’s abstract. “This response is exacerbated when more dimensions of unusual communication are identified for a firm.”
Klement points out that with the introduction of OpenAI’s GPT-4o in May, analysts could also upload the earnings call audio to the LLM or deploy it during the livestream.
While analysts often pick up on unusual comments themselves, the technology will allow them to analyze multiple earnings calls at once across industries.
“Personally, I love it, but as a corporate executive I probably wouldn’t because chatGPT makes it much harder to get away with deception and distraction,” Klement wrote.