For example, it could hypothetically do this via inventing a new, more powerful technology which might be more useful than transformers for AGI, and refocusing AI alignment research on this technology, saving AI alignment research months or years or decades of misdirected effort.
Or it could hypothetically do this by providing technology which significantly speeds up AI alignment research, and which did not already exist.
The capabilities research does not have to be net-positive, or actually used in a net-positive way, or even invented, in 2023. It only has to be seen as net-positive eventually.
Mar 5, 1:06pm: Will AI capabilities research done in 2023 or after be net-positive for AI alignment research? → Will some piece of AI capabilities research done in 2023 or after be net-positive for AI alignment research?
1,000How and when would this resolve no?
In whose judgment are you measuring net positive?
Will you trade on this market?
Will the positive impact on capabilities increase be included as a negative?
@MartinRandall I guess I'm assuming that this question will be discussed by experts between now and the closing date, and it would resolve no by the closing date if no-one has come up with a convincing argument for how any piece of capabilities research was in fact met-positive for alignment research.
And yes, capabilities increase will count as a negative in the overall assessment of a piece of technology.
But I think this market might have more value as a prediction market in the present than as a resolved prediction market at the close date.
What is this?
What is Manifold?
Are our predictions accurate?
Why use play money?
Why use play money?Related questions
What is this?
What is Manifold?
Are our predictions accurate?
Why use play money?
Why use play money?