AI and Library Resources
It's important to know the co-relation between Generative AI & Library Resources by understanding a little bit about how Open Source AI works, as opposed to Closed Source AI.
Open Source AI
Open Source AI refers to models and tools that are publicly available and accessible for anyone to use or modify. These kinds of models allow access to the provenance of their training data and/or source code. They will also, sometimes, re-train based on data or prompts that users put in.
Closed Source AI
Closed Source AI, by contrast, has a proprietary approach to software development and does not make its source code or training data publicly available. It will also not retrain based on user input; any modifications are made by the developer organization or a limited group of authorized user
Why does this matter?
This matters to us, as users, because it informs what kind of information we upload or enter into Open Source AI systems. Once something has been entered into an open source tool, the creator or license holder can't control if and how it is reused by the model.
This puts certain information at risk. For example:
- Private or privileged information (e.g. trade secrets, identifying information, etc.)
- Copyrighted or licensed information (e.g. articles or reports downloaded from databases, books, etc
Adapted from Bryant University Libraries
How does this impact Library Resources?
When the library subscribes to a database on behalf of the college's faculty, staff, and students, we need to stick to that resource's licensing agreements. Those licenses protect the intellectual property of that resource.
Most database vendors prevent sharing their resources with those who do not have authorized access. This is why if you're logging into the databases while off-campus, you'll be asked to input your user name and password.
Because users have very little control over how their input will or could be used by an Open Source AI platform, this also applies to inputting database content into most AI platforms.