"GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models"
[Submitted on 17 Mar 2023] GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models Tyna Eloundou <https://arxiv.org/search/econ?searchtype=author&query=Eloundou%2C+T>, Sam Manning <https://arxiv.org/search/econ?searchtype=author&query=Manning%2C+S>, Pamela Mishkin <https://arxiv.org/search/econ?searchtype=author&query=Mishkin%2C+P>, Daniel Rock <https://arxiv.org/search/econ?searchtype=author&query=Rock%2C+D> We investigate the potential implications of Generative Pre-trained Transformer (GPT) models and related technologies on the U.S. labor market. Using a new rubric, we assess occupations based on their correspondence with GPT capabilities, incorporating both human expertise and classifications from GPT-4. Our findings indicate that approximately 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GPTs, while around 19% of workers may see at least 50% of their tasks impacted. The influence spans all wage levels, with higher-income jobs potentially facing greater exposure. Notably, the impact is not limited to industries with higher recent productivity growth. We conclude that Generative Pre-trained Transformers exhibit characteristics of general-purpose technologies (GPTs), suggesting that as these models could have notable economic, social, and policy implications. Subjects: General Economics (econ.GN); Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: arXiv:2303.10130 <https://arxiv.org/abs/2303.10130> [econ.GN] (or arXiv:2303.10130v1 <https://arxiv.org/abs/2303.10130v1> [econ.GN] for this version) https://doi.org/10.48550/arXiv.2303.10130
Grazie JC, l'avevo visto e mi promettevo di commentarlo... La lista delle limitazioni di questo articolo è lunga quanto la somma di tutte le braccia rubate all'agricoltura dei suoi autori (tranne Daniel Rock, che qualche anno fa aveva posizioni molto meno oltranziste sulla questione, cf. https://www.nber.org/papers/w24001). Questo documento è un ennesimo caso di LaTeX-driven advertising, nel senso che un paper con parvenza di scientificità viene usato per fare pubblicità o assecondare le operazioni di marketing di un'azienda. Un po' come il caso storico di Eytan Bakshy, Solomon Messing, Lada Adamic, “Exposure to ideologically diverse news and opinion on Facebook [archive]”, Science, 7, 2015, che scagionava l'algoritmo di NewsFeed di Facebook e dava la colpa agli utilizzatori per la creazione di “echo chambers”. Da accogliere, come tutte le analisi task-based, con un sonoro "meh". Cheers, ---a ----- Original Message ----- From: "J.C. DE MARTIN" <juancarlos.demartin@polito.it> To: "Nexa" <nexa@server-nexa.polito.it> Sent: Monday, March 20, 2023 12:41:41 PM Subject: [nexa] "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models" [Submitted on 17 Mar 2023] GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models [ https://arxiv.org/search/econ?searchtype=author&query=Eloundou%2C+T | Tyna Eloundou ] , [ https://arxiv.org/search/econ?searchtype=author&query=Manning%2C+S | Sam Manning ] , [ https://arxiv.org/search/econ?searchtype=author&query=Mishkin%2C+P | Pamela Mishkin ] , [ https://arxiv.org/search/econ?searchtype=author&query=Rock%2C+D | Daniel Rock ] We investigate the potential implications of Generative Pre-trained Transformer (GPT) models and related technologies on the U.S. labor market. Using a new rubric, we assess occupations based on their correspondence with GPT capabilities, incorporating both human expertise and classifications from GPT-4. Our findings indicate that approximately 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GPTs, while around 19% of workers may see at least 50% of their tasks impacted. The influence spans all wage levels, with higher-income jobs potentially facing greater exposure. Notably, the impact is not limited to industries with higher recent productivity growth. We conclude that Generative Pre-trained Transformers exhibit characteristics of general-purpose technologies (GPTs), suggesting that as these models could have notable economic, social, and policy implications. Subjects: General Economics (econ.GN) ; Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: [ https://arxiv.org/abs/2303.10130 | arXiv:2303.10130 ] [econ.GN] (or [ https://arxiv.org/abs/2303.10130v1 | arXiv:2303.10130v1 ] [econ.GN] for this version) [ https://doi.org/10.48550/arXiv.2303.10130 | https://doi.org/10.48550/arXiv.2303.10130 ] _______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa
Grazie, Antonio! jc On 20/03/23 13:36, Antonio Casilli wrote:
Grazie JC,
l'avevo visto e mi promettevo di commentarlo... La lista delle limitazioni di questo articolo è lunga quanto la somma di tutte le braccia rubate all'agricoltura dei suoi autori (tranne Daniel Rock, che qualche anno fa aveva posizioni molto meno oltranziste sulla questione, cf. https://www.nber.org/papers/w24001).
Questo documento è un ennesimo caso di LaTeX-driven advertising, nel senso che un paper con parvenza di scientificità viene usato per fare pubblicità o assecondare le operazioni di marketing di un'azienda. Un po' come il caso storico di Eytan Bakshy, Solomon Messing, Lada Adamic, “Exposure to ideologically diverse news and opinion on Facebook [archive]”, Science, 7, 2015, che scagionava l'algoritmo di NewsFeed di Facebook e dava la colpa agli utilizzatori per la creazione di “echo chambers”.
Da accogliere, come tutte le analisi task-based, con un sonoro "meh". Cheers, ---a
----- Original Message ----- From: "J.C. DE MARTIN" <juancarlos.demartin@polito.it> To: "Nexa" <nexa@server-nexa.polito.it> Sent: Monday, March 20, 2023 12:41:41 PM Subject: [nexa] "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models"
[Submitted on 17 Mar 2023] GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models [ https://arxiv.org/search/econ?searchtype=author&query=Eloundou%2C+T | Tyna Eloundou ] , [ https://arxiv.org/search/econ?searchtype=author&query=Manning%2C+S | Sam Manning ] , [ https://arxiv.org/search/econ?searchtype=author&query=Mishkin%2C+P | Pamela Mishkin ] , [ https://arxiv.org/search/econ?searchtype=author&query=Rock%2C+D | Daniel Rock ]
We investigate the potential implications of Generative Pre-trained Transformer (GPT) models and related technologies on the U.S. labor market. Using a new rubric, we assess occupations based on their correspondence with GPT capabilities, incorporating both human expertise and classifications from GPT-4. Our findings indicate that approximately 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GPTs, while around 19% of workers may see at least 50% of their tasks impacted. The influence spans all wage levels, with higher-income jobs potentially facing greater exposure. Notably, the impact is not limited to industries with higher recent productivity growth. We conclude that Generative Pre-trained Transformers exhibit characteristics of general-purpose technologies (GPTs), suggesting that as these models could have notable economic, social, and policy implications.
Subjects: General Economics (econ.GN) ; Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: [ https://arxiv.org/abs/2303.10130 | arXiv:2303.10130 ] [econ.GN] (or [ https://arxiv.org/abs/2303.10130v1 | arXiv:2303.10130v1 ] [econ.GN] for this version) [ https://doi.org/10.48550/arXiv.2303.10130 | https://doi.org/10.48550/arXiv.2303.10130 ]
_______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa _______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa
Si dà per scontato che il business model del LLM sia la vendita di servizi presso imprese. Credo varrebbe la pena interrogarsi tempestivamente sui business models che verranno usati verso il pubblico per non ripetere l'esperienza fatta coi motori di ricerca e con i SN, specie in vista di una regolazione. Sappiamo che nei motori di ricerca la risposta a una query dipende da 1) dal contenuto della domanda, 2) dal profilo dell'utente e 3) dall'esistenza di offerte commerciali di inserzionisti compatibili con i primi due elementi. Vista la lucratività, non vi sono motivi di pensare che il modello di business di un servizio linguistico debba essere diverso, tanto più che la profilazione è facilmente giustificata dalla possibilità di generare testi che rispecchino il punto di vista individuale e contengano elementi stilistici personali. Con un modello di business di questo tipo la risposta del servizio conterrà elementi determinati da inserzionisti per favorire la formazione di testi con contenuti specifici. Ad esempio i testi generati per i giornalisti potrebbero contenere o meno fatti, collegamenti, paragoni, esempi, o usare certe espressioni, ricordare certi eventi, far prevalere certi punti di vista, in base a quanto richiesto da parti interessate alla presenza o assenza di certi contenuti. Al contrario della veridicità dei fatti presenti nei testi, questi elementi (specie se assenti) non possono essere verificati. Questo offrirà possibilità di manipolazione ancora più forti e soprattutto subdole rispetto a quelle già spaventose dei motori di ricerca. Alberto On 3/20/23 13:36, Antonio Casilli wrote:
Grazie JC,
l'avevo visto e mi promettevo di commentarlo... La lista delle limitazioni di questo articolo è lunga quanto la somma di tutte le braccia rubate all'agricoltura dei suoi autori (tranne Daniel Rock, che qualche anno fa aveva posizioni molto meno oltranziste sulla questione, cf. https://www.nber.org/papers/w24001).
Questo documento è un ennesimo caso di LaTeX-driven advertising, nel senso che un paper con parvenza di scientificità viene usato per fare pubblicità o assecondare le operazioni di marketing di un'azienda. Un po' come il caso storico di Eytan Bakshy, Solomon Messing, Lada Adamic, “Exposure to ideologically diverse news and opinion on Facebook [archive]”, Science, 7, 2015, che scagionava l'algoritmo di NewsFeed di Facebook e dava la colpa agli utilizzatori per la creazione di “echo chambers”.
Da accogliere, come tutte le analisi task-based, con un sonoro "meh". Cheers, ---a
----- Original Message ----- From: "J.C. DE MARTIN" <juancarlos.demartin@polito.it> To: "Nexa" <nexa@server-nexa.polito.it> Sent: Monday, March 20, 2023 12:41:41 PM Subject: [nexa] "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models"
[Submitted on 17 Mar 2023] GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models [ https://arxiv.org/search/econ?searchtype=author&query=Eloundou%2C+T | Tyna Eloundou ] , [ https://arxiv.org/search/econ?searchtype=author&query=Manning%2C+S | Sam Manning ] , [ https://arxiv.org/search/econ?searchtype=author&query=Mishkin%2C+P | Pamela Mishkin ] , [ https://arxiv.org/search/econ?searchtype=author&query=Rock%2C+D | Daniel Rock ]
We investigate the potential implications of Generative Pre-trained Transformer (GPT) models and related technologies on the U.S. labor market. Using a new rubric, we assess occupations based on their correspondence with GPT capabilities, incorporating both human expertise and classifications from GPT-4. Our findings indicate that approximately 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GPTs, while around 19% of workers may see at least 50% of their tasks impacted. The influence spans all wage levels, with higher-income jobs potentially facing greater exposure. Notably, the impact is not limited to industries with higher recent productivity growth. We conclude that Generative Pre-trained Transformers exhibit characteristics of general-purpose technologies (GPTs), suggesting that as these models could have notable economic, social, and policy implications.
Subjects: General Economics (econ.GN) ; Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: [ https://arxiv.org/abs/2303.10130 | arXiv:2303.10130 ] [econ.GN] (or [ https://arxiv.org/abs/2303.10130v1 | arXiv:2303.10130v1 ] [econ.GN] for this version) [ https://doi.org/10.48550/arXiv.2303.10130 | https://doi.org/10.48550/arXiv.2303.10130 ]
_______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa _______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa
participants (3)
-
Alberto Cammozzo -
Antonio Casilli -
J.C. DE MARTIN