International Scientific Report on the Safety of Advanced AI: interim report
Yohsua Bengio
(1)
,
Soren Mindermann
(2)
,
Daniel Privitera
(3)
,
Tamay Besiroglu
(4)
,
Rishi Bommasani
(5)
,
Stephen Casper
(6)
,
Yejin Choi
(7)
,
Danielle Goldfarb
(1)
,
Hoda Heidari
(8)
,
Leila Khalatbari
(9)
,
Shayne Longpre
(6)
,
Vasilios Mavroudis
(10)
,
Mazeika Mantas
(11)
,
Kwan Yee Ng
(12)
,
Chinasa T. Okolo
(13)
,
Deborah Raji
(14)
,
Theodora Skeadas
(15)
,
Florian Tramèr
(16)
,
Bronwyn Fox
(17)
,
Andre Carlos Ponce de Leon Ferreira de Carvalho
(18)
,
Mona Nemer
(19)
,
Raquel Pezoa Rivera
(20)
,
Yi Zeng
(21)
,
Juha Heikkilä
(22)
,
Guillaume Avrin
(23)
,
Antonio Krüger
(24)
,
Balaraman Ravindran
(25)
,
Hammam Riza
(26)
,
Ciarán Seoighe
(27)
,
Ziv Katvir
(28)
,
Andrea Monti
(29)
,
Hiroaki Kitano
(30)
,
Mary Kerema
(31)
,
José Ramón
(32)
,
Haroon Sheikh
(33)
,
Gill Jolly
(34)
,
Olubunmi Ajala
(35)
,
Dominic Ligot
(36)
,
Kyoung Mu Lee
(37)
,
Ahmet Halit
(38)
,
Crystal Rugege
(39)
,
Fahad Albalawi
(40)
,
Denise Wong
(41)
,
Nuria Oliver
(42)
,
Christian Busch
(43)
,
Oleksii Molchanovskyi
(44)
,
Marwan Alserkal
(45)
,
Saif Khan
(46)
,
Angela Mclean
(47)
,
Amandeep Gill
(48)
,
Bayo Adekanmbi
(49)
,
Paul Christiano
(50)
,
David Dalrymple
(51)
,
Thomas G. Dietterich
(52)
,
Edward Felten
(53)
,
Pascale Fung
(9)
,
Pierre-Olivier Gourinchas
(54)
,
Nick Jennings
(55)
,
Andreas Krause
,
Percy Liang
(5)
,
Teresa Ludermir
,
Vidushi Marda
(56)
,
Helen Margetts
(57)
,
John A. Mcdermid
,
Arvind Narayanan
(53)
,
Alondra Nelson
(58)
,
Alice Oh
(59)
,
Gopal Ramchurn
(60)
,
Stuart Russell
(61)
,
Marietje Schaake
(5)
,
Dawn Song
(61)
,
Alvaro Soto
(62)
,
Lee Tiedrich
(63)
,
Gaël Varoquaux
(64)
,
Andrew Yao
(65)
,
Yaqin Zhang
(65)
1
Mila - Quebec AI Institute
2 Mila - Institut québécois d’intelligence artificielle = Quebec Artificial Intelligence Institute
3 KIRA Center
4 Epoch AI
5 Stanford University
6 MIT - Massachusetts Institute of Technology
7 University of Washington [Seattle]
8 CMU - Carnegie Mellon University [Pittsburgh]
9 HKUST - Hong Kong University of Science and Technology
10 The Alan Turing Institute
11 UIUC - University of Illinois at Urbana-Champaign [Urbana]
12 Concordia AI
13 The Brookings Institution
14 Mozilla [Berlin]
15 Humane Intelligence
16 ETH Zürich
17 CSIRO - Commonwealth Scientific and Industrial Research Organisation [Australia]
18 USP - Universidade de São Paulo = University of São Paulo
19 Government of Canada
20 Federico Santa María Technical University
21 Institute of Automation - Chinese Academy of Sciences
22 DG Connect
23 DGE Direction Générale des Entreprises
24 DFKI - Deutsches Forschungszentrum für Künstliche Intelligenz GmbH = German Research Center for Artificial Intelligence
25 IIT Madras - Indian Institute of Technology Madras
26 KORIKA
27 Science Foundation Ireland
28 Israel Innovation Authority
29 Ud'A - Università degli studi "G. d'Annunzio" Chieti-Pescara [Chieti-Pescara]
30 Sony Corporation
31 Ministry of Information Communications Technology and Digital Economy
32 Q Element
33 Netherlands’ Scientific Council for Government Policy
34 Ministry of Business, Innovation and Employment
35 Innovation and Digital Economy
36 CirroLytix
37 SNU - Seoul National University [Seoul]
38 Turkish Ministry of Industry and Technology
39 National Center for AI and Innovation Policy
40 Saudi Authority for Data and Artificial Intelligence
41 Data Innovation and Protection Group, Infocomm Media Development Authority
42 ELLIS Alicante
43 Federal Department of Economic Affairs, Education and Research
44 Expert Committee on the Development of Artificial intelligence in Ukraine
45 Ministry of Cabinet Affairs, Prime Minister’s Office
46 departement of commerce
47 UK Government
48 ONU - United Nations = Organisation des Nations unies
49 Data Science Nigeria
50 AI Safety Institute
51 ARIA
52 OSU - Oregon State University
53 Princeton University
54 International Monetary Fund (IMF)
55 Loughborough University
56 REAL ML
57 University of Oxford
58 IAS - Institute for Advanced Study [Princeton]
59 CS - Department of Computer Science [KAIST]
60 University of Southampton
61 UC Berkeley - University of California [Berkeley]
62 Pontifica Universidad de Chile
63 Duke University [Durham]
64 SODA - Méthodes computationnelles et mathématiques pour comprendre la société et la santé à partir de données
65 Tsing Hua University
2 Mila - Institut québécois d’intelligence artificielle = Quebec Artificial Intelligence Institute
3 KIRA Center
4 Epoch AI
5 Stanford University
6 MIT - Massachusetts Institute of Technology
7 University of Washington [Seattle]
8 CMU - Carnegie Mellon University [Pittsburgh]
9 HKUST - Hong Kong University of Science and Technology
10 The Alan Turing Institute
11 UIUC - University of Illinois at Urbana-Champaign [Urbana]
12 Concordia AI
13 The Brookings Institution
14 Mozilla [Berlin]
15 Humane Intelligence
16 ETH Zürich
17 CSIRO - Commonwealth Scientific and Industrial Research Organisation [Australia]
18 USP - Universidade de São Paulo = University of São Paulo
19 Government of Canada
20 Federico Santa María Technical University
21 Institute of Automation - Chinese Academy of Sciences
22 DG Connect
23 DGE Direction Générale des Entreprises
24 DFKI - Deutsches Forschungszentrum für Künstliche Intelligenz GmbH = German Research Center for Artificial Intelligence
25 IIT Madras - Indian Institute of Technology Madras
26 KORIKA
27 Science Foundation Ireland
28 Israel Innovation Authority
29 Ud'A - Università degli studi "G. d'Annunzio" Chieti-Pescara [Chieti-Pescara]
30 Sony Corporation
31 Ministry of Information Communications Technology and Digital Economy
32 Q Element
33 Netherlands’ Scientific Council for Government Policy
34 Ministry of Business, Innovation and Employment
35 Innovation and Digital Economy
36 CirroLytix
37 SNU - Seoul National University [Seoul]
38 Turkish Ministry of Industry and Technology
39 National Center for AI and Innovation Policy
40 Saudi Authority for Data and Artificial Intelligence
41 Data Innovation and Protection Group, Infocomm Media Development Authority
42 ELLIS Alicante
43 Federal Department of Economic Affairs, Education and Research
44 Expert Committee on the Development of Artificial intelligence in Ukraine
45 Ministry of Cabinet Affairs, Prime Minister’s Office
46 departement of commerce
47 UK Government
48 ONU - United Nations = Organisation des Nations unies
49 Data Science Nigeria
50 AI Safety Institute
51 ARIA
52 OSU - Oregon State University
53 Princeton University
54 International Monetary Fund (IMF)
55 Loughborough University
56 REAL ML
57 University of Oxford
58 IAS - Institute for Advanced Study [Princeton]
59 CS - Department of Computer Science [KAIST]
60 University of Southampton
61 UC Berkeley - University of California [Berkeley]
62 Pontifica Universidad de Chile
63 Duke University [Durham]
64 SODA - Méthodes computationnelles et mathématiques pour comprendre la société et la santé à partir de données
65 Tsing Hua University
Andre Carlos Ponce de Leon Ferreira de Carvalho
- Fonction : Auteur
- PersonId : 955050
Antonio Krüger
- Fonction : Auteur
- PersonId : 1019931
Andreas Krause
- Fonction : Auteur
- PersonId : 1316284
Teresa Ludermir
- Fonction : Auteur
- PersonId : 955048
John A. Mcdermid
- Fonction : Auteur
- PersonId : 1016417
Gaël Varoquaux
- Fonction : Auteur
- PersonId : 5878
- IdHAL : gael-varoquaux
- ORCID : 0000-0003-1076-5122
- IdRef : 126239894
Résumé
We are in the midst of a technological revolution that will fundamentally alter the way we live, work, and relate to one another. Artificial Intelligence (AI) promises to transform many aspects of our society and economy.
There is broad scientific consensus that the capabilities of AI systems have progressed rapidly on many tasks in the last five years. Large Language Models (LLMs) are a particularly salient example. In 2019, GPT-2, then the most advanced LLM, could not reliably produce a coherent paragraph of text and could not always count to ten. At the time of writing, the most powerful LLMs like Claude 3, GPT-4, and Gemini Ultra can engage consistently in multi-turn conversations, write short computer programs, translate between multiple languages, score highly on university entrance exams, and summarise long documents.
This step-change in capabilities, and the potential for continued progress, could help advance the public interest in many ways. Among the most promising prospects are AI’s potential for education, medical applications, research advances in a wide range of fields, and increased innovation leading to increased prosperity. This rapid progress has also increased awareness of the current harms and potential future risks associated with the most capable types of AI.
Domaines
Informatique [cs]
Fichier principal
international_scientific_report_on_the_safety_of_advanced_ai_interim_report.pdf (4.05 Mo)
Télécharger le fichier
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|
