logseq/pages/Omnivore___21-02-2024___Token 1.21- Dark side of LLMs.md

full-title:: Token 1.21: Dark side of LLMs site:: omnivore.app labels:: Newsletter date-saved:: 21-02-2024 date-archived:: 04-03-2024 is-archived:: 10 source:: Omnivore state:: archived

- ### Highlights
	- > Hugging Face has over 500k models available to fine-tune or use out of the box. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#bf6c1675-e536-4104-8a24-a4415c6206bb)
	- > arXiv has over 19,000 papers on LLM. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#e67a536b-501c-4dfb-a013-77d855f888e4)
	- > 2,200 papers were published this year alone i.e. in the past 50 days. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#14df1213-b60c-46c5-b48b-4629faeb67aa)
	- > vulnerabilities in large language models are talked about very little. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#4178deda-6ded-4ecb-9c38-04168e124d4e)
	- > vulnerabilities in large language models are talked [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#85625771-26bc-40e5-8a12-a406681b2aab)
	- > when a single model is serving millions of users, a lot of things can go wrong when the LLM’s vulnerabilities are taken advantage of. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#c1d757c8-5bbb-4068-ad1d-f78c146c3355)
	- > LLMs, if misused, could [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#8f10d4ae-1a75-49c6-8ec5-eae682ec2f0a)
	- > The model can be tricked into revealing personal information from its training corpus [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#e28faa00-a0dd-42b4-84ff-ca98c228daad)
	- > There's a risk that LLMs could inadvertently generate content that **infringes on copyrighted materials**, leading to complex legal challenges for creators and users alike. [⤴️](https://omnivore.app/me/token-1-21-dark-side-of-ll-ms-18dcd9ea14a#149245ee-dc17-4604-9dac-9e2111faeabd)