Intro to AI for GLAM: Key Points

Welcome

Intro to AI for GLAM is for staff working in the GLAM (Galleries, Libraries, Archives, and Museums) sector.
The lesson is a high-level conceptual introduction to AI and machine learning that will empower GLAM staff to apply those technologies within their own institutions and collections.
This lesson will not cover coding, statistics or maths.

Machine Learning is a subfield of AI which identifies patterns in data
Supervised learning algorithms learn by example
Unsupervised learning algorithms put data into groups of similar objects or records

Conceptual models describe a general relationship between inputs and outputs
Trained models are a numerical realisation of a conceptual model learned from data
Explainable AI is an essential tool for understanding complex models

Bias occurs when a dataset is not representative of the population, it is incomplete or skewed.
The presence of bias in the classifications and predictions of machine learning may have far reaching consequences for society, amplifying inequality and unfairness.
There are abundant opportunities for bias to enter ML systems at all stages of the pipeline including when datasets are constructed, when a models learning is refined and reinforced, and when predictions made by a model are interpreted by humans and applied to real world scenarios
There are a range of strategies available today to help mitigate bias.

Machine learning projects involve many considerations beyond training a model.
The predictions made by the same machine learning model can be ‘translated’ into actions in different ways. The extent to which you ‘automate’ decisions versus keeping a ‘human-in-the-loop’ will depend on the problem you are tackling, your organization and your model’s performance.
The use of Machine learning by GLAMs is relatively new. Sharing results and lessons learned will likely help GLAMS realize the potential benefits of machine learning.