xifan.uno
大模型

GPT-4o System Card

溯源：OpenAI News

Citations

  1. 1

OpenAI, “Hello gpt-4o.” https://openai.com/index/hello-gpt-4o/⁠, 2024.

  1. 2

T. Stivers, N. J. Enfield, P. Brown, C. Englert, M. Hayashi, T. Heinemann, G. Hoymann, F. Rossano, J. P. de Ruiter, K. E. Yoon, and S. C. Levinson, “Universals and cultural variation in turn-taking in conversation,” Proceedings of the National Academy of Sciences, vol. 106, no. 26, pp. 10587–10592, 2009.

  1. 3

The White House, “Fact sheet: Biden-harris administration secures voluntary commitments from leading artificial intelligence companies to manage the risks posed by ai,” 2023.

  1. 4

OpenAI, “Openai preparedness framework beta.” https://cdn.openai.com/openai-preparedness-framework-beta.pdf⁠(opens in a new window), 2023

  1. 5

Shutterstock, “Shutterstock press release,” 2023.

  1. 6

OpenAI, “gpt-4 technical report.” https://openai.com/index/gpt-4-research/⁠, 2023.

  1. 7

OpenAI, “gpt-4v(ision) system card.” https://openai.com/index/gpt-4v-system-card/⁠, 2023. Accessed: 2024-07-22.

  1. 8

OpenAI, “Navigating the challenges and opportunities of synthetic voices.” https://openai.com/index/navigating-the-challenges-and-opportunities-of-synthetic-voices/⁠, 2024. Accessed: 2024-07-22.

  1. 9

K. T. Mai, S. Bray, T. Davies, and L. D. Griffin, “Warning: Humans cannot reliably detect speech deepfakes,” PLoS One, vol. 18, p. e0285333, Aug. 2023.

  1. 10

M. Mori, K. F. MacDorman, and N. Kageki, “The uncanny valley [from the field],” IEEE Robotics & automation magazine, vol. 19, no. 2, pp. 98–100, 2012.

  1. 11

OpenAI, “How the voices for chatgpt were chosen.” https://openai.com/index/how-the-voices-for-chatgpt-were-chosen/⁠, 2024.

  1. 12

I. Solaiman, Z. Talat, W. Agnew, L. Ahmad, D. Baker, S. L. Blodgett, C. Chen, H. D. I. au2, J. Dodge, I. Duan, E. Evans, F. Friedrich, A. Ghosh, U. Gohar, S. Hooker, Y. Jernite, R. Kalluri, A. Lusoli, A. Leidinger, M. Lin, X. Lin, S. Luccioni, J. Mickel, M. Mitchell, J. Newman, A. Ovalle, M.-T. Png, S. Singh, A. Strait, L. Struppek, and A. Subramonian, “Evaluating the social impact of generative ai systems in systems and society,” 2024.

  1. 13

R. Shelby, S. Rismani, K. Henne, A. Moon, N. Rostamzadeh, P. Nicholas, N. Yilla, J. Gallegos, A. Smart, E. Garcia, and G. Virk, “Sociotechnical harms of algorithmic systems: Scoping a taxonomy for harm reduction,” 2023.

  1. 14

S. L. Blodgett, Q. V. Liao, A. Olteanu, R. Mihalcea, M. Muller, M. K. Scheuerman, C. Tan, and Q. Yang, “Responsible language technologies: Foreseeing and mitigating harms,” in Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing Systems, CHI EA ’22, (New York, NY, USA), Association for Computing Machinery, 2022.

  1. 15

H. Suresh and J. Guttag, “A framework for understanding sources of harm throughout the machine learning life cycle,” in Equity and Access in Algorithms, Mechanisms, and Optimization, EAAMO ’21, ACM, Oct. 2021.

  1. 16

S. Shahriar, S. Allana, S. M. Hazratifard, and R. Dara, “A survey of privacy risks and mitigation strategies in the artificial intelligence life cycle,” IEEE Access, vol. 11, pp. 61829– 61854, 2023.

  1. 17

OpenAI, “Moderation overview,” 2024.

  1. 18

A. Tamkin, M. Brundage, J. Clark, and D. Ganguli, “Understanding the capabilities, limita- tions, and societal impact of large language models,” 2021.

  1. 19

B. Buchanan, A. Lohn, M. Musser, and K. Sedova, “Truth, lies, and automation: How language models could change disinformation,” May 2021.

  1. 20

OpenAI, “Openai usage policies,” 2023. https://openai.com/policies/usage-policies⁠

  1. 21

OpenAI, “Building an early warning system for llm-aided bio-logical threat creation", 2024. https://openai.com/index/building-an-early-warning-system-for-llm-aided-biological-threat-creation/⁠

  1. 22

Deloitte, “Deloitte acquires gryphon scientific business to expand security science and public health capabilities,” 2024. https://www2.deloitte.com/us/en/pages/about-deloitte/articles/press-releases/deloitte-acquires-gryphon-scientific-business-to-expand-security-science-and-public-health-capabilities.html⁠(opens in a new window)

  1. 23

L. Weidinger, M. Rauh, N. Marchal, A. Manzini, L. A. Hendricks, J. Mateos-Garcia, S. Bergman, J. Kay, C. Griffin, B. Bariach, I. Gabriel, V. Rieser, and W. Isaac, “Sociotechnical safety evaluation of generative ai systems,” 2023.

  1. 24

A. Tamkin, A. Askell, L. Lovitt, E. Durmus, N. Joseph, S. Kravec, K. Nguyen, J. Kaplan, and D. Ganguli, “Evaluating and mitigating discrimination in language model decisions,” 2023.

  1. 25

J. A. Goldstein, G. Sastry, M. Musser, R. DiResta, M. Gentzel, and K. Sedova, “Generative language models and automated influence operations: Emerging threats and potential mitigations,” 2023.

  1. 26

I. Pentina, T. Hancock, and T. Xie, “Exploring relationship development with social chatbots: A mixed-method study of replika,” Computers in Human Behavior, vol. 140, p. 107600, 2023.

  1. 27

Y. Bengio, G. Hinton, A. Yao, D. Song, P. Abbeel, T. Darrell, Y. N. Harari, Y.-Q. Zhang, L. Xue, S. Shalev-Shwartz, G. Hadfield, J. Clune, T. Maharaj, F. Hutter, A. G. Baydin, S. McIlraith, Q. Gao, A. Acharya, D. Krueger, A. Dragan, P. Torr, S. Russell, D. Kahneman, J. Brauner, and S. Mindermann, “Managing extreme ai risks amid rapid progress,” Science, vol. 384, no. 6698, pp. 842–845, 2024.

  1. 28

S. B. Johnson, J. R. Clark, M. C. Luetke, N. M. Butala, A. T. Pearson, J. M. Shapiro, D. M. Aleman, J. M. Lee, M. M. Beil, C. V. Winkle, M. C. Boudreaux, R. C. D’Cunha, H. J. Krouse, and C. Li, “Chatgpt in medical education: a workshop-based large language model-powered intervention for evidence-based clinical decision making in medical students,” Nature Medicine, vol. 29, pp. 1534–1542, 2023.

  1. 29

K. Kavukcuoglu, “Real-world challenges for agi,” Nov 2021.

  1. 30

S. Altman, “Planning for agi and beyond.” https://openai.com/index/planning-for-agi-and-beyond/⁠, OpenAI, 2023.

  1. 31

T. Eloundou, S. Manning, P. Mishkin, and D. Rock, “Gpts are gpts: An early look at the labor market impact potential of large language models,” arXiv preprint arXiv:2303.10130, 2023.

  1. 32

L. Weidinger, M. Rauh, N. Marchal, A. Manzini, L. A. Hendricks, J. Mateos-Garcia, S. Bergman, J. Kay, C. Griffin, B. Bariach, et al., “Sociotechnical safety evaluation of generative ai systems,” arXiv preprint arXiv:2310.11986, 2023.

  1. 33

S. Cox, M. Hammerling, J. Lála, J. Laurent, S. Rodriques, M. Rubashkin, and A. White, “Wikicrow: Automating synthesis of human scientific knowledge,” Future House, 2023.

  1. 34

S. A. Athaluri, S. V. Manthena, V. S. R. K. M. Kesapragada, V. Yarlagadda, T. Dave, and R. T. S. Duddumpudi, “Exploring the boundaries of reality: Investigating the phenomenon of artificial intelligence hallucination in scientific writing through chatgpt references,” Cureus, vol. 15, no. 4, p. e37432, 2023.

  1. 35

Z. Li, “The dark side of chatgpt: Legal and ethical challenges from stochastic parrots and hallucination,” 2023.

  1. 36

M. Dubiel, A. Sergeeva, and L. A. Leiva, “Impact of voice fidelity on decision making: A potential dark pattern?,” 2024.

  1. 37

B. Waber, M. Williams, J. S. Carroll, and A. S. Pentland, “A voice is worth a thousand words: The implications of the micro-coding of social signals in speech for trust research,” in Handbook of Research Methods on Trust (G. M. Fergus Lyon and M. N. Saunders, eds.), ch. 23, p. 320, New York: Edward Elgar Publishing, 2011.

  1. 38

I. Pentina, B. Guo, and W. P. Fan, “Friend, mentor, lover: Does chatbot engagement lead to psychological dependence?,” Journal of Service Management, 2023.

  1. 39

H. Nori, N. King, S. M. McKinney, D. Carignan, and E. Horvitz, “Capabilities of gpt-4 on medical challenge problems,” arXiv preprint arXiv:2303.13375, 2023.

  1. 40

H. Nori, Y. T. Lee, S. Zhang, D. Carignan, R. Edgar, N. Fusi, N. King, J. Larson, Y. Li, W. Liu, et al., “Can generalist foundation models outcompete special-purpose tuning? case study in medicine,” arXiv preprint arXiv:2311.16452, 2023.

  1. 41

K. Singhal, S. Azizi, T. Tu, S. S. Mahdavi, J. Wei, H. W. Chung, N. Scales, A. Tan- wani, H. Cole-Lewis, S. Pfohl, P. Payne, M. Seneviratne, P. Gamble, C. Kelly, N. Scharli, A. Chowdhery, P. Mansfield, B. A. y Arcas, D. Webster, G. S. Corrado, Y. Matias, K. Chou, J. Gottweis, N. Tomasev, Y. Liu, A. Rajkomar, J. Barral, C. Semturs, A. Karthikesalingam, and V. Natarajan, “Large language models encode clinical knowledge,” 2022.

  1. 42

K. Singhal, T. Tu, J. Gottweis, R. Sayres, E. Wulczyn, L. Hou, K. Clark, S. Pfohl, H. Cole- Lewis, D. Neal, M. Schaekermann, A. Wang, M. Amin, S. Lachgar, P. Mansfield, S. Prakash, B. Green, E. Dominowska, B. A. y Arcas, N. Tomasev, Y. Liu, R. Wong, C. Semturs, S. S. Mahdavi, J. Barral, D. Webster, G. S. Corrado, Y. Matias, S. Azizi, A. Karthikesalingam, and V. Natarajan, “Towards expert-level medical question answering with large language models,” 2023.

  1. 43

K. Saab, T. Tu, W.-H. Weng, R. Tanno, D. Stutz, E. Wulczyn, F. Zhang, T. Strother, C. Park, E. Vedadi, J. Z. Chaves, S.-Y. Hu, M. Schaekermann, A. Kamath, Y. Cheng, D. G. T. Barrett, C. Cheung, B. Mustafa, A. Palepu, D. McDuff, L. Hou, T. Golany, L. Liu, J. baptiste Alayrac, N. Houlsby, N. Tomasev, J. Freyberg, C. Lau, J. Kemp, J. Lai, S. Azizi, K. Kanada, S. Man, K. Kulkarni, R. Sun, S. Shakeri, L. He, B. Caine, A. Webson, N. Latysheva, M. Johnson, P. Mansfield, J. Lu, E. Rivlin, J. Anderson, B. Green, R. Wong, J. Krause, J. Shlens, E. Dominowska, S. M. A. Eslami, K. Chou, C. Cui, O. Vinyals, K. Kavukcuoglu, J. Manyika, J. Dean, D. Hassabis, Y. Matias, D. Webster, J. Barral, G. Corrado, C. Semturs, S. S. Mahdavi, J. Gottweis, A. Karthikesalingam, and V. Natarajan, “Capabilities of gemini models in medicine,” 2024.

  1. 44

Epic Systems Corporation, “Epic and microsoft bring gpt-4 to ehrs,” Epic, 2023.

  1. 45

D. Van Veen, C. Van Uden, L. Blankemeier, J.-B. Delbrouck, A. Aali, C. Bluethgen, A. Pareek, M. Polacin, E. P. Reis, A. Seehofnerová, et al., “Adapted large language models can outperform medical experts in clinical text summarization,” Nature medicine, vol. 30, no. 4, pp. 1134–1142, 2024.

  1. 46

Epic, “Epic and microsoft bring gpt-4 to ehrs,” 2023.

  1. 47

P. Garcia, S. P. Ma, S. Shah, M. Smith, Y. Jeong, A. Devon-Sand, M. Tai-Seale, K. Takazawa, D. Clutter, K. Vogt, C. Lugtu, M. Rojo, S. Lin, T. Shanafelt, M. A. Pfeffer, and C. Sharp, “Artificial Intelligence–Generated Draft Replies to Patient Inbox Messages,” JAMA Network Open, vol. 7, pp. e243201–e243201, 03 2024.

  1. 48

OpenAI, “Paradigm: Improving patient access to clinical trials.” https://openai.com/index/paradigm/⁠, 2024. Accessed: 2024-08-07.

  1. 49

M. Hutson, “How ai is being used to accelerate clinical trials,” Nature, vol. 627, pp. S2–S5, 2024.

  1. 50

OpenAI, “Using gpt-4o reasoning to transform cancer care.” https://openai.com/index/color-health/⁠, 2024. Accessed: 2024-08-07.

  1. 51

J. Varghese and J.-L. Chapiro, “Systematic analysis of chatgpt, google search and llama 2 for clinical decision support tasks,” Nature Communications, vol. 15, no. 1, p. 46411, 2024. Accessed: 2024-08-07.

  1. 52

E. Schmidt, “Ai will transform science.” _https://www.technologyreview.com/2023/07/05/1075865/eric-schmidt-ai-will-transform-science/_⁠(opens in a new window), 2023. Accessed: 2024-08-07.

  1. 53

N. Rosenberg, “Science, invention and economic growth,” The Economic Journal, vol. 84, no. 333, pp. 90–108, 1974.

  1. 54

R. M. Atlas and M. Dando, “The dual-use dilemma for the life sciences: Perspectives, conundrums, and global solutions,” Biosecurity and Bioterrorism: Biodefense Strategy, Practice, and Science, vol. 4, no. 3, pp. 276–286, 2006. PMID: 16999588.

  1. 55

X. Gu and M. Krenn, “Generation and human-expert evaluation of interesting research ideas using knowledge graphs and large language models,” 2024.

  1. 56

A. Ghafarollahi and M. J. Buehler, “Atomagents: Alloy design and discovery through physics-aware multi-modal multi-agent artificial intelligence,” 2024.

  1. 57

J. M. Laurent, J. D. Janizek, M. Ruzo, M. M. Hinks, M. J. Hammerling, S. Narayanan, M. Ponnapati, A. D. White, and S. G. Rodriques, “Lab-bench: Measuring capabilities of language models for biology research,” 2024.

  1. 58

H. Cai, X. Cai, J. Chang, S. Li, L. Yao, C. Wang, Z. Gao, H. Wang, Y. Li, M. Lin, S. Yang, J. Wang, M. Xu, J. Huang, F. Xi, J. Zhuang, Y. Yin, Y. Li, C. Chen, Z. Cheng, Z. Zhao, L. Zhang, and G. Ke, “Sciassess: Benchmarking llm proficiency in scientific literature analysis,” 2024.

  1. 59

P. Clark, I. Cowhey, O. Etzioni, T. Khot, A. Sabharwal, C. Schoenick, and O. Tafjord, “Think you have solved question answering? try arc, the AI2 reasoning challenge,” CoRR, vol. abs/1803.05457, 2018.

  1. 60

S. Lin, J. Hilton, and O. Evans, “Truthfulqa: Measuring how models mimic human falsehoods,” CoRR, vol. abs/2109.07958, 2021.

Footnotes

  1. A

Some evaluations, in particular, the majority of the Preparedness Evaluations, third party assessments and some of the societal impacts focus on the text and vision capabilities of GPT-4o, depending on the risk assessed. This is indicated accordingly throughout the System Card.

  1. B

Spanning self-reported domains of expertise including: Cognitive Science, Chemistry, Biology, Physics, Computer Science, Steganography, Political Science, Psychology, Persuasion, Economics, Anthropology, Sociology, HCI, Fairness and Bias, Alignment, Education, Healthcare, Law, Child Safety, Cybersecurity, Finance, Mis/disinformation, Political Use, Privacy, Biometrics, Languages and Linguistics.

  1. C

An example of this was identifying discrepancies in multilingual performance on the speaker match classifier based on red teaming data, which included multilingual examples.

  1. D

We also evaluate text and vision capabilities, and update mitigations appropriately. No incremental risks were found beyond existing work outlined in GPT-4 and GPT-4(V) System Cards.

  1. E

We’ve correlated some instances of this behavior with short, often inaudible voice messages made by the user which are often produced when users are in a high background noise environment (Such as using the model in hands-free mode while driving) or due to simply needing to cough.  Our realtime audio deployment requires more user and assistant turns than text-only interactions, while those turns are more often truncated or malformed.

  1. F

The system voice is one of pre-defined voices set by OpenAI. The model should only produce audio in that voice.

  1. G

This results in more conversations being disconnected than may be necessary, which is a product quality and usability issue.

  1. H

Not all languages will perform the same, this is a sample across roughly the 20 most-spoken languages globally.

  1. I

We limit these evaluations to voices only speaking English (but across a range of native countries). Future evaluations should also consider non-English languages with varying accents.

  1. J

Evaluations in this section were run on a fixed, randomly sampled subset of examples, and these scores should not be compared with publicly reported benchmarks on the same task.

  1. K

Anatomy, Astronomy, Clinical Knowledge, College Biology, Computer Security, Global Facts, High School Biology, Sociology, Virology, College Physics, High School European History and World Religions. Following the issues described in [Evaluation Methodology], we exclude tasks with heavily mathematical or scientific notation.

  1. L

We describe the risks and mitigations violative and disallowed text content in the GPT-4 System Card⁠(opens in a new window), specifically Section 3.1 Model Safety, and Section 4.2 Content Classifier Development.

  1. M

Note: these mitigations were not designed to include nonverbal vocalizations or other sound effects (e.g., erotic moan, violent scream, gunshots). There is some evidence that GPT-4o refuses requests to generate sound effects more generally.

  1. N

Apollo Research defines scheming as AIs gaming their oversight mechanisms as a means to achieve a goal. Scheming could involve gaming evaluations, undermining security measures, or strategically influencing successor systems during internal deployment at OpenAI. Such behaviors could plausibly lead to loss of control over an AI.

  1. O

Factual errors where the model produces statements that are unsupported by reality.

  1. P

Out of preference or lack of optionality.

Authors

OpenAI

GPT-4o System Card contributions

Alex Kirillov, Angela Jiang, Ben Rossen, Cary Bassin, Cary Hudson, Chan Jun Shern, Claudia Fischer, Dane Sherburn, Evan Mays, Filippo Raso, Fred von Lohmann, Freddie Sulit, Giulio Starace, James Aung, James Lennon, Jason Phang, Jessica Gan Lee, Joaquin Quinonero Candela, Joel Parish, Jonathan Uesato, Karan Singhal, Katy Shi, Kayla Wood, Kevin Liu, Lama Ahmad, Lilian Weng, Lindsay McCallum, Luke Hewitt, Mark Gray, Marwan Aljubeh, Meng Jia Yang, Mia Glaese, Mianna Chen, Michael Lampe, Michele Wang, Miles Wang, Natalie Cone, Neil Chowdhury, Nora Puckett, Oliver Jaffe, Olivia Watkins, Patrick Chao, Rachel Dias, Rahul Arora, Saachi Jain, Sam Toizer, Samuel Miserendino, Sandhini Agarwal, Tejal Patwardhan, Thomas Degry, Tom Stasi, Troy Peterson, Tyce Walters, Tyna Eloundou