User guide

What is the EAI database？

The Environmental Artificial Intelligence (EAI) Database is an open and public resource collating the various costs of different AI and ML firms, products, and services like OpenAI’s now widely-used ChatGPT. These environmental costs are typically spread along the so-called AI ‘pipeline’ from the collection of user data to ‘model inference’ – the technical term for using a ML-driven service. 

The project takes inspiration from various tech database initiatives, including the AIAAIC Repository, an ‘independent, open, public interest resource detailing incidents and controversies driven by, and relating to AI, algorithms, and automation’ (AIAAIC 2024). 

How do I access the database?

The EAI Database is an open-access database that runs through Google Sheets. It can be accessed through the link under the section “link to database” from the EAI Database Google Sites.

How do I use the database?

The EAI database can be used in a variety of ways, for example, as a source for research by journalists or simply used by adults curious about the environmental impact of AI. Further information about possible users of the database can be found under the section “User Profiles”. The database is divided into two sheets:

Database, this sheet includes the database, which is divided into categories

Glossary, this sheet includes the glossary, which explains terms related to the sources

Classification of sources

The database is divided under different categories which simplifies research. The categories have been listed below:

Headline

This category includes the headline of the source, which has been linked to the link of the source

Author(s)

This category includes the author(s) of the source. Some sources are anonymous and do not disclose the author, so the name of the company who published the source has been used instead. For instance “The Economist”. This affects the reliability of sources, so please keep in mind this detail.

Date

This category includes the year in which the source has been published

Country

This category includes the country or countries mentioned in the sources. This category can have multiple options. The classification “Worldwide” is a general classification, as the source may not specify the country or countries, or most countries have been mentioned, this does not exclude that some countries may not be included.

Developer(s)

This category includes the developer(s) mentioned in the sources. For instance “OpenAI”. 

Technology type

This category includes the technology type that has been analyzed in the source. This category can have multiple options. For instance “Machine Learning” and “Large Language Model”. The classification “All AI Technologies” is a general classification that includes all possible AI technologies. Please keep in mind that some AI technologies might not be included as this is a generalized classification.

Environmental impact

This category includes the environmental impact(s) that AI has caused. The current categories are:

Carbon Emission

Energy Consumption

Water Usage

Land Use

Rare Earth Minerals & Material Use

E-Waste & Hardware Lifespan

Air Pollution

Other Pollution

Non renewable Energy Use

GHG Emissions

Other Raw Materials

In the future the categories might expand as more sources will be added to the database.

Industry funding 

This category includes any industry funding mentioned in the sources

Cost of impact 

This category includes the cost of impact of AI, this category will vary depending on the category “Environmental Impact”

Metaphors of Environmental Cost

This category includes metaphors of environmental cost. This category facilitates the understanding of the environmental impact. For instance “Exchanging 20 messages with ChatGPT equals 500ml of water”.

Media types

This category includes what kind of media type the resource is. The current categories are:

Academic Paper

Social Media Post

Blog 

News Report

Government Report

Institutional Report

Conference Paper

Preprint(discretion advised)

Policy Paper

Literature Review

Source

This category includes the link to the source

DOI

This category includes the DOI of the source

Added by (initials)

This category includes the contributor of the source. To maintain anonymity only the initials have been disclosed.

Additional notes

This category includes any additional notes to the source. This will usually be useful information like additional findings related to the environmental impact of AI.  


Glossary

The glossary can be found on the second sheet. It includes the following categories:

Terms

Full title

This category includes the full title as often the terms are abbreviations

Definition

This category includes a general definition to the term

Contributors of the database

All entries to the database are carefully chosen and manually added by its contributors.