What we’re about
How can we survive in world of superintelligent AI? This sc-fi predicament is rapidly becoming a real one, and it makes sense to have a plan.
This group is for people interested in this problem. We will explore the machine learning literature that tries to solve it, discussing a text at each meeting.
If this sounds like your kind of thing, please come along!
Upcoming events (1)
See all- How can we understand what an AI is thinking?The Prince George, Brighton BN1 4EQ
Neural networks are notoriously inscrutable black boxes. The field of mechanistic interpretability looks to change this by reverse-engineering them into human-understandable components.
We will be introducing ourselves to these techniques by discussing the blog God Help Us, Let's Try To Understand AI Monosemanticity by Scott Alexander:
‘You’ve probably heard AI is a “black box”… This is bad for safety. For safety, it would be nice to look inside the AI and see whether it’s executing an algorithm like “do the thing” or more like “trick the humans into thinking I’m doing the thing”. But we can’t. Because we can’t look inside an AI at all. Until now!... [Big AI company/research lab Anthropic] claims to have gazed inside an AI and seen its soul… How did they do it? What is inside of an AI? And what the heck is “monosemanticity”?’
If this looks interesting to you, whether you have ideas to share or just want to listen, please come along!
