Pioneers
2. Artificial intelligence
2.2. Timnit Gebru
Figure 1: Timnit Gebru.
Source: TechCrunch (2021)
Downloadable teaching resource
Overview
Timnit Gebru (ትምኒት ገብሩ) is a leading figure in AI ethics.
Background
Timnit Gebru was born in Addis Ababa, Ethiopia in 1983 of Eritrean descent, and went to the US as a political refugee in 1999. She went on to achieve bachelor’s and master’s degrees in electrical engineering at Stanford, and a Ph.D. from the Stanford Artificial Intelligence Laboratory (Dataiku, 2021).
Contributions
Gebru initially worked on designing circuits and algorithms, including work on facial recognition software, for Apple as an intern. She went on to a successful career primarily as a researcher in AI, with a focus on AI ethics. Gebru worked at Microsoft Research in the FATE (Fairness Accountability Transparency and Ethics in AI) group, studying algorithmic bias and ethical concerns in data use (Dataiku, 2021). After noting the lack of black people at the 2016 Neural Information Systems Conference, Gebru was instrumental in launching 'Black in AI' to improve representation in the AI sector (Black in AI, 2024).
While at Microsoft, she co-authored a research paper titled 'Gender Shades', looking at bias in facial recognition software. This was linked to an MIT project promoting intersectional and inclusive product testing in AI (MIT Media Lab, no date). The Gender Shades project revealed differences in error rates, particularly worse for darker females. The project is presented at gendershades.org and the paper available via MIT Media Lab (Buolamwini and Gebru 2018; Buolamwini et al., 2018).
She joined Google as a research scientist on their ethical AI team, concerned with the implications of AI and technology for "social good" (Dataiku, 2021). In 2020 she worked on a paper covering ethical risks of AI language models with other authors including Google staff. The paper proved controversial, and she was forced out of Google after refusing to withdraw the paper or remove Google employees as authors (Dataiku, 2021). This paper is discussed in the Feature section below. She was later named one of 'The 100 Most Influential People of 2022' by Time (2022).
Feature: Stochastic Parrots?
A version of the paper that led to Gebru's departure from Google was presented at the ACM Conference on Fairness, Accountability, and Transparency (FAccT '21) and is available in the conference proceedings (Bender, et al., 2021). There is a summary of the paper and the surrounding events from the MIT Technology Review (Hao, 2020).
The paper raises questions of the possible risks around AI and Large Language Models (LLM) such as the technology behind ChatGPT, and considers how these risks might be handled. This may include not undertaking specific developments.
The highlighted risks include:
- Environmental costs in energy use and emissions of AI training and use.
- Deep-seated problems with large data sets, including bias, prejudice and lack of diversity.
In conclusion, the authors contend:
"In this paper, we have invited readers to take a step back and ask: Are ever larger LMs inevitable or necessary? What costs are associated with this research direction and what should we consider before pursuing it? Do the field of NLP or the public that it serves in fact need larger LMs? If so, how can we pursue this research direction while mitigating its associated risks? If not, what do we need instead?" (Bender, et al., 2021, p. 619).
See also
After leaving Google, Gebru founded the Distributed AI Research Institute (DAIR Institute, 2024a). The Institute supports community-focused AI research independently from Big Tech, arguing for a critical and appropriate development of AI. You can view their projects and publications at www.dair-institute.org (DAIR Institute, 2024b).
References and further reading
Bender, E. M., Gebru, T., McMillan-Major, A. and Shmitchell, S. (2021) 'On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?🦜', Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. New York, NY, USA: ACM, pp. 610–623. Available at: https://doi.org/10.1145/3442188.3445922 (Accessed: 31 January 2025)
Black in AI (2024) Home ┃ Black In AI. Available at: https://www.blackinai.org/ (Accessed: 02 February 2025)
Buolamwini, J. and Gebru, T. (2018) 'Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification', Proceedings of Machine Learning Research, 81, pp. 1–15, Conference on Fairness, Accountability, and Transparency, New York University, NYC, February 23 -24, 2018. Available at: https://www.media.mit.edu/publications/gender-shades-intersectional-accuracy-disparities-in-commercial-gender-classification/ (Accessed: 02 February 2025)
Buolamwini, J., Gebru, T. Raji, D., Raynham, H., and Zuckerman, E. (2018) Gender Shades. Available at: http://gendershades.org/overview.html (Accessed: 02 February 2025)
Dataiku (2021) Timnit Gebru: The Computer Scientist Fighting for a Fairer World. Available at: https://www.historyofdatascience.com/timnit-gebru-the-computer-scientist-fighting-for-a-fairer-world/ (Accessed: 02 February 2025)
Distributed AI Research Institute (DAIR Institute) (2024a) Timnit Gebru Launches Independent AI Research Institute On Anniversary of Ouster from Google. Available at: https://www.dair-institute.org/press-release/ (Accessed: 31 January 2025)
Distributed AI Research Institute (DAIR Institute) (2024b) Distributed AI Research Institute. Available at: https://www.dair-institute.org (Accessed: 31 January 2025)
Gebru, T. and Torres, É. (2024) The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence. Available at: https://doi.org/10.5210/fm.v29i4.13636 (Accessed: 17 March 2025)
Hao, K. (2020) We read the paper that forced Timnit Gebru out of Google. Here’s what it says. Available at: https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/ (Accessed: 31 January 2025)
Massachusetts Institute of Technology Media Lab (MIT Media Lab) (no date) Gender Shades. Available at: https://www.media.mit.edu/projects/gender-shades/overview/ (Accessed: 02 February 2025)
Perrigo, B. (2022) Why Timnit Gebru Isn’t Waiting for Big Tech to Fix AI’s Problems. Available at: https://time.com/6132399/timnit-gebru-ai-google/ (Accessed: 17 March 2025)
TechCrunch (2021) File:Timnit Gebru crop.jpg. Available at: https://commons.wikimedia.org/wiki/File:Timnit_Gebru_crop.jpg (Accessed: 31 January 2025)
Time (2022) Timnit Gebru Is on the 2022 TIME 100 List | TIME. Available at: https://time.com/collection/100-most-influential-people-2022/6177822/timnit-gebru/ (Accessed: 02 February 2025)