2021 – AI Impacts

Do neural networks learn human concepts?

This page is a stub. It does not necessarily represent much of what is known on the topic. Our understanding is that the degree to which neural networks learn concepts that are potentially understandable to

AI Vignettes Project

An ongoing effort to write concrete plausible future histories of AI development and its social impacts.

Argument for AI x-risk from large impacts

An argument that since AI will likely have a large impact on society, it is especially likely to be an existential risk to humanity.

Beyond fire alarms: freeing the groupstruck

Katja Grace Sept 2021

Fire alarms are the wrong way to think about the public AGI conversation.

Vignettes workshop

Daniel Kokotajlo June 2021

Write down how AI will go down!

Fiction relevant to AI futurism

A list of stories potentially relevant to thinking about the development of advanced AI, including both those intended as futurism and those intended as entertainment.

What do coherence arguments imply about the behavior of advanced AI?

There is probably a pressure to become more coherent, if you have preferences, and thus to become more goal-directed.

April files

Katja Grace April 2021

Internal drafts for feedback

Coherence arguments imply a force for goal-directed behavior

Katja Grace Mar 2021

Behavior that is permitted by the ‘coherence arguments’ may still be discouraged by them.