| Revision as of 13:04, 13 December 2024 editBelbury (talk | contribs)Extended confirmed users, Rollbackers75,019 edits Adding local short description: "Field of study in artificial intelligence", overriding Wikidata description "field of study in artificial intelligence that aims to give machines the ability to "forget" learned information"Tag: Shortdesc helper← Previous edit | Revision as of 21:11, 24 December 2024 edit undoVillaida (talk | contribs)Extended confirmed users8,064 edits added "needs more sources" template because article had one source.Tag: Visual editNext edit → | ||
| Line 1: | Line 1: | ||
| {{Short description|Field of study in artificial intelligence}} | {{Short description|Field of study in artificial intelligence}}{{More sources|date=December 2024}} | ||
| '''Machine unlearning''' is a branch of ] focused on removing specific undesired element, such as private data, outdated information, copyrighted material, harmful content, dangerous abilities, or misinformation, without needing to rebuild models from the ground up. Large language models, like the ones powering ChatGPT, may be asked not just to remove specific elements but also to unlearn a "concept," "fact," or "knowledge," which aren't easily linked to specific examples. New terms such as "model editing," "concept editing," and "knowledge unlearning" have emerged to describe this process.<ref name="Liu_2024"> |
'''Machine unlearning''' is a branch of ] focused on removing specific undesired element, such as private data, outdated information, copyrighted material, harmful content, dangerous abilities, or misinformation, without needing to rebuild models from the ground up. Large language models, like the ones powering ], may be asked not just to remove specific elements but also to unlearn a "concept," "fact," or "knowledge," which aren't easily linked to specific examples. New terms such as "model editing," "concept editing," and "knowledge unlearning" have emerged to describe this process.<ref name="Liu_2024">{{Cite web |title=Machine Unlearning in 2024 |url=https://ai.stanford.edu/~kzliu/blog/unlearning |archive-url=http://web.archive.org/web/20241213234527/https://ai.stanford.edu/~kzliu/blog/unlearning |archive-date=2024-12-13 |access-date=2024-12-24 |website=Ken Ziyu Liu - Stanford Computer Science |language=en-US}}</ref> | ||
| == History == | == History == | ||
| {{Unsourced|section|date=December 2024}} | |||
| Early research efforts were largely motivated by Article 17 of the ], the European Union's privacy regulation commonly known as the "right to be forgotten" (RTBF), introduced in 2014. | Early research efforts were largely motivated by Article 17 of the ], the European Union's privacy regulation commonly known as the "right to be forgotten" (RTBF), introduced in 2014. | ||
| ==Present== |
==Present== | ||
| {{Unsourced|section|date=December 2024}} | |||
| The GDPR did not anticipate that the development of ]s would make data erasure a complex task. This issue has since led to research on "machine unlearning," with a growing focus on removing copyrighted material, harmful content, dangerous capabilities, and misinformation. Just as early experiences in humans shape later ones, some concepts are more fundamental and harder to unlearn. A piece of knowledge may be so deeply embedded in the model’s knowledge graph that unlearning it could cause internal contradictions, requiring adjustments to other parts of the graph to resolve them. | The GDPR did not anticipate that the development of ]s would make data erasure a complex task. This issue has since led to research on "machine unlearning," with a growing focus on removing copyrighted material, harmful content, dangerous capabilities, and misinformation. Just as early experiences in humans shape later ones, some concepts are more fundamental and harder to unlearn. A piece of knowledge may be so deeply embedded in the model’s knowledge graph that unlearning it could cause internal contradictions, requiring adjustments to other parts of the graph to resolve them. | ||
Revision as of 21:11, 24 December 2024
Field of study in artificial intelligence | This article needs additional citations for verification. Please help improve this article by adding citations to reliable sources. Unsourced material may be challenged and removed. Find sources: "Machine unlearning" – news · newspapers · books · scholar · JSTOR (December 2024) (Learn how and when to remove this message) |
Machine unlearning is a branch of machine learning focused on removing specific undesired element, such as private data, outdated information, copyrighted material, harmful content, dangerous abilities, or misinformation, without needing to rebuild models from the ground up. Large language models, like the ones powering ChatGPT, may be asked not just to remove specific elements but also to unlearn a "concept," "fact," or "knowledge," which aren't easily linked to specific examples. New terms such as "model editing," "concept editing," and "knowledge unlearning" have emerged to describe this process.
History
| This section does not cite any sources. Please help improve this section by adding citations to reliable sources. Unsourced material may be challenged and removed. Find sources: "Machine unlearning" – news · newspapers · books · scholar · JSTOR (December 2024) (Learn how and when to remove this message) |
Early research efforts were largely motivated by Article 17 of the GDPR, the European Union's privacy regulation commonly known as the "right to be forgotten" (RTBF), introduced in 2014.
Present
| This section does not cite any sources. Please help improve this section by adding citations to reliable sources. Unsourced material may be challenged and removed. Find sources: "Machine unlearning" – news · newspapers · books · scholar · JSTOR (December 2024) (Learn how and when to remove this message) |
The GDPR did not anticipate that the development of large language models would make data erasure a complex task. This issue has since led to research on "machine unlearning," with a growing focus on removing copyrighted material, harmful content, dangerous capabilities, and misinformation. Just as early experiences in humans shape later ones, some concepts are more fundamental and harder to unlearn. A piece of knowledge may be so deeply embedded in the model’s knowledge graph that unlearning it could cause internal contradictions, requiring adjustments to other parts of the graph to resolve them.
References
- "Machine Unlearning in 2024". Ken Ziyu Liu - Stanford Computer Science. Archived from the original on 2024-12-13. Retrieved 2024-12-24.
 | This article needs additional or more specific categories. Please help out by adding categories to it so that it can be listed with similar articles. (December 2024) |