Will Superalignment succeed in their core goal according to me, a complete layperson?
22
5.5k
resolved May 20
Resolved
NO

I have absolutely no expertise in Alignment but I am concerned about misaligned AI and I try to pay attention to the issue in the news.

It would be very surprising if Superalignment succeeded at their goal in 4 years, given that they have admitted that the goal is very ambitious.

There are currently several markets about if Superalignment will succeed in their goal in their own estimation, but none about if they will succeed in anyone else's estimation.

I am very skeptical that Superalignment will succeed, but would be delighted to be proven wrong. I have sold all my stakes in these markets, and will not make any more bets in this market or any other Superalignment markets so as to try to keep myself unbiased.

I therefore offer my completely non-expert, common sense opinion as a judge.


For this market, success is defined by these quotes from the introduction post:

"Our goal is to solve the core technical challenges of superintelligence alignment in four years."

"Solving the problem includes providing evidence and arguments that convince the machine learning and safety community that it has been solved. If we fail to have a very high level of confidence in our solutions, we hope our findings let us and the community plan appropriately."


I am hopeful that I will not have to judge any of the technical aspects of solving alignment because I can simply look at the safety community and judge if they have been convinced the problem is solved.

I'm putting the resolution date as the end of 2027 to allow time for the safety community to come to a consensus.

This market resolves YES if I judge they have achieved their goal and convinced the community they have done so, and NO otherwise.

Get Ṁ600 play money

🏅 Top traders

#NameTotal profit
1Ṁ131
2Ṁ70
3Ṁ62
4Ṁ59
5Ṁ55
Sort by:

Well the team has been "effectively dissolved" and the original market has resolved no, so I think I should do the same here.

@Joshua I am aware that some of the key members have left OpenAI. However, is it not possible OpenAI will promote or arrange new team members to the project? That would be the default for a normal organisation.

This market was inspired by this comment from @EliezerYudkowsky. I agree that an external panel of judges would be better, but unfortunately I don't know of any that are willing to resolve my manifold markets.

More related questions