Depends on what you mean. Kaj Sotala, a research associate at MIRI, has a proposal he submitted to FLI that I think really deserves to be funded (it’s the context for the modeling concept formation posts he has done recently). I think it has a good chance and I would be very disappointed if it wasn’t funded. I’m not sure if you would count that as MIRI getting funded or not since the organization is technically not on the proposal, I think.
If you mean MIRI getting funded to do the sorts of things MIRI has been prominently pushing for in its workshops lately—basic mathematical research on incomputable models of intelligence, Löbian obstacles, decision theory, etc. -- then I would be very, very disappointed and if it was a significant chunk of the FLI budget I would have to retract my endorsement. I would be very surprised, but I don’t consider it an impossibility. I actually think it quite possible that MIRI could get selected for a research grant on something related to, but slightly different from what they would have been working on anyways (I have no idea if they have submitted any proposals). I do think it unlikely and surprising if FLI funds were simply directed into the MIRI technical research agenda.
To clarify, my understanding is that FLI was founded largely in response to Nick Bostrom’s Superintelligence drumming up concern over AI risk, so there is a shared philosophical underpinning between MIRI and FLI—based on the same arguments I object to in the OP! But if Musk et al believed the MIRI technical agenda was the correct approach to deal with the issue, and MIRI itself was capable of handling the research, then they would have simply given their funds to MIRI and not created their own organization. There is a real different between how these two organizations are approaching the issue, and I expect to see that reflected in the grant selections.
FLI is doing the right thing but for the wrong reasons. What you do matters more than why you did it, so as long as FLI continues to do the right thing, they’ll have my endorsement.
I do think it unlikely and surprising if FLI funds were simply directed into the MIRI technical research agenda.
FLI’s request for proposals links to their research priorities document, which cites 4 MIRI papers and states, “Research in this area [...] could extend or
critique existing approaches begun by groups such as the Machine Intelligence Research Institute [76].” (Note that [76] is MIRI’s technical agenda document.)
So obviously they don’t think MIRI’s current research is a waste of resources or should be deprioritized. If they end up not funding any MIRI projects, I would think it probably has more to do with MIRI having enough funding from other sources than any disagreements between FLI and MIRI over research directions.
Depends on what you mean. Kaj Sotala, a research associate at MIRI, has a proposal he submitted to FLI that I think really deserves to be funded (it’s the context for the modeling concept formation posts he has done recently). I think it has a good chance and I would be very disappointed if it wasn’t funded. I’m not sure if you would count that as MIRI getting funded or not since the organization is technically not on the proposal, I think.
If you mean MIRI getting funded to do the sorts of things MIRI has been prominently pushing for in its workshops lately—basic mathematical research on incomputable models of intelligence, Löbian obstacles, decision theory, etc. -- then I would be very, very disappointed and if it was a significant chunk of the FLI budget I would have to retract my endorsement. I would be very surprised, but I don’t consider it an impossibility. I actually think it quite possible that MIRI could get selected for a research grant on something related to, but slightly different from what they would have been working on anyways (I have no idea if they have submitted any proposals). I do think it unlikely and surprising if FLI funds were simply directed into the MIRI technical research agenda.
To clarify, my understanding is that FLI was founded largely in response to Nick Bostrom’s Superintelligence drumming up concern over AI risk, so there is a shared philosophical underpinning between MIRI and FLI—based on the same arguments I object to in the OP! But if Musk et al believed the MIRI technical agenda was the correct approach to deal with the issue, and MIRI itself was capable of handling the research, then they would have simply given their funds to MIRI and not created their own organization. There is a real different between how these two organizations are approaching the issue, and I expect to see that reflected in the grant selections.
FLI is doing the right thing but for the wrong reasons. What you do matters more than why you did it, so as long as FLI continues to do the right thing, they’ll have my endorsement.
FLI’s request for proposals links to their research priorities document, which cites 4 MIRI papers and states, “Research in this area [...] could extend or critique existing approaches begun by groups such as the Machine Intelligence Research Institute [76].” (Note that [76] is MIRI’s technical agenda document.)
So obviously they don’t think MIRI’s current research is a waste of resources or should be deprioritized. If they end up not funding any MIRI projects, I would think it probably has more to do with MIRI having enough funding from other sources than any disagreements between FLI and MIRI over research directions.