1. No killing (universally agreed across legal systems and religions) 2. Health (including mental health, longevity, happiness, wellbeing) 3. Biosphere, environment, other living creatures 4. AI safety 5. Mars: backup civilization is fully aligned with the virtue of life preservation
These principles were applicable to the Network State and I think first three of them can be repurposed towards AGI alignment.
(another core belief is GRAVITY—I believe in GRAVITY—GRAVITY brought us together)
2. Pause, defection, incentives
New proof-of-X algorithm to ensure compliance with AI moratorium. Ensuring supercomputers are not used for training more powerful models.
Proof-of-work consumes loads of energy.
It could be a mixture of algorithms, that is more energy friendly:
peak power for a short amount of time (solve something complex quickly)
proof of storage
A mixture of different algorithms to ensure various elements of the data centre are yielded unsuitable for other means. I know too little about the challenges of operating a data center, I know too little about training AI, ultimately I do not know.
I’m just aware of the incentive to defect and no obvious way to enforce the rules.
AGI—alignment—paperclip maximizer—pause—defection—incentives
I would like to expose myself to critique.
I hope this is a place where I can receive some feedback + share some of the insights that came to me.
https://​​en.wikipedia.org/​​wiki/​​Dunning%E2%80%93Kruger_effect—“people with low ability, expertise, or experience regarding a certain type of task or area of knowledge tend to overestimate their ability or knowledge”
I’m somewhere on the spectrum 🤡
1. AGI alignment metrics
To avoid paperclip maximizer and solving climate change by eliminating humans I suggest the following value: LIFE
I’ve embedded this principle into Network State Genesis and described it in the founding document in the following way:
1. No killing (universally agreed across legal systems and religions)
2. Health (including mental health, longevity, happiness, wellbeing)
3. Biosphere, environment, other living creatures
4. AI safety
5. Mars: backup civilization is fully aligned with the virtue of life preservation
These principles were applicable to the Network State and I think first three of them can be repurposed towards AGI alignment.
(another core belief is GRAVITY—I believe in GRAVITY—GRAVITY brought us together)
2. Pause, defection, incentives
New proof-of-X algorithm to ensure compliance with AI moratorium. Ensuring supercomputers are not used for training more powerful models.
Proof-of-work consumes loads of energy.
It could be a mixture of algorithms, that is more energy friendly:
peak power for a short amount of time (solve something complex quickly)
proof of storage
A mixture of different algorithms to ensure various elements of the data centre are yielded unsuitable for other means. I know too little about the challenges of operating a data center, I know too little about training AI, ultimately I do not know.
I’m just aware of the incentive to defect and no obvious way to enforce the rules.
So much easier to prove the existence of aliens.
So much more difficult to disprove.
So much easier to prove you did the thing.
So much more difficult to disprove.