Notice how irreplaceable you are.
As a human being, no one can replace you. To be specifically you. That is just irreplaceable.
So the actionable would be don't become the model of yourself. That is a machine.
Connect more with your human and that will make you more irreplaceable and might bring forth deeper potentiation rather than some AI generated ideas that don't have deep potency because they're disconnected from reality.
And that might lead to something inspiring and that will scale because it's inspiring, not because it's super machine-like.
Do the Jamais vu thing, which is like look at something like you've never seen it before.
The opposite of Déjà vu.
And the idea is that as you're doing that, you will notice the things that are contingent.
This is especially important to do for at the infrastructure level, the things that you take most for granted.
Those are the things that are gonna change. So that's something that you have to be prepared for.
Soloware is a solo activity, until software agents get really good we cannot expect an individual to maintain the software, keep upgrading unless they invest some attention.
The promise of community is that sense of things that they will integrate, cohere and bring together for you that you don't have to select.
What remains valuable in the future?
Do reach out if you are interested in fine grained credit assignment in a world where it is very easy to "steal" your stuff.
@michael_nielsen@littmath Sahil at https://t.co/kn0i32PBdh is doing research inspired by ideas similar to your last sentence.
This is probably the best intro: https://t.co/uDaGpcQAWN
Groundless Alignment is excited to announce applications are open to the three tracks we are accepting applications under.
This is in collaboration with AI Safety Camp!
Track Three - MoSSAIC. This stands for Management of Substrate-Specific AI Capabilities. Both mechanistic and abstract approaches might be insufficiently cynical to meet the evasiveness of AI risk as it gets increasingly self modifying.
We need an approach that takes into account the highly specific implementation details while being sufficiently general.
Participants will work on threat modeling, clarifying limitations and designing experiments for investigation of AI systems in the near future.
This is led by Matt Farr.
Many thanks to @richarchngo for his generous donation of 15,000 USD towards our Groundless Alignment Residency that will happen in Goa, India. If you are interested to visit then reach out to DMs.