"2) Funding by NSF and similar public money grant program."

Based on both what I heard and what I experienced, it's private foundations that would have the lower standards, because they are agenda-driven and the people who work there have the mission to find scientists doing research on whatever the topic-of-the-year is.

I guess the best way is just to post on a bunch of platforms and have those web pages back-up by various archiving services (notably the Wayback Machine).

Note that you can also set up your writing to be automatically archived via archive.is (example).

OriginStamp is a Bitcoin-based service that seems to be solid.

It’s used by, for example, namespace’s Project Registration Database. (I built a custom PmWiki plug-in for this purpose, that allows any wiki page (or any wiki page you want) to automatically be OriginStamp’d upon creation. I haven’t released the code yet, but if anyone’s interested, I will.)

There are many dedicated Timestamping Services for this Use-Case, one that has been around since '95 and with a corrrespondingly solid reputation IMHO is http://www.itconsult.co.uk/stamper/stampinf.htm. But there are other services as well, just search for "Timestamping Service"... I'd steer clear of the Bitcoin-based ones, they tend to be... enthusiastic but not too proffessional.

Besides the subsystems making their own predictions, there might also be a meta-learning system keeping track of which other subsystems tend to make the most accurate predictions in each situation, giving extra weight to the bids of the subsystem which has tended to perform the best in that situation.

This is why I eat junk food, sans guilt. I don't want my central planning subagent to lose influence over unimportant details. Spend your weirdness points wisely.

I guess the best way is just to post on a bunch of platforms and have those web pages back-up by various archiving services (notably the Wayback Machine).

Publishing a book is probably too much work for most used cases.

I've also heard one could publish text on the Bitcoin blockchain for example, but I'm not sure how well that works.

Sweet! I added it to the list of podcasts by people with a LessWrong profile: https://wiki.lesswrong.com/wiki/List_of_Podcasts#By_people_with_Less_Wrong_profiles

As a Chinese I want to contribute some thought into this topic.

One thing I want to mention is the difference in language. Classical Chinese is a language extremely difficult to master. It literally take decades of effort to be able to write a decent piece. It is hard not because of complicated grammar or complex sentence structure. But because it focus on poetic expressions and scholarly idioms. This language is very enjoyable to read and relatable when used in expressing emotions and ideas. However it is quite cumbersome in expressing precise logic and definitions. Yet at least before the new cultural movement in 1916 it is generally regarded that anything worth put into writing should be done in Classical Chinese. This severely limits the participation of the general populace. Even if someone is trained enough to put down scientific related topics in Classical Chinese it is unlikely to be regarded as a masterful piece and gather much audience. Just like if a poorly written piece is posted in lesswrong we are more likely to skip it regardless of the content it is expressing.

I didn't take the time to check whether it did or didn't. If you would walk me through how it does, I would appreciate it.

I am claiming that you can't make a human seriously superhuman with a good education.

Is the claim that $\delta o/\delta r$ for humans goes down over time so that $o$ eventually hits an asymptote? If so, why won't that apply to AI?

Serious genetic modification is another story, but at that point, your building an AI out of protien.

But it seems quite relevant that we haven't successfully done that yet.

You couldn't get much better results just by throwing more compute at it.

Okay, so my new story for this argument is:

• For every task T, there are bottlenecks that limit its performance, which could be compute, data, algorithms, etc.
• For the task of "AI research", compute will not be the bottleneck.
• So, once we get human-level performance on "AI research", we can apply more compute to get exponential recursive self-improvement.

Is that your argument? If so, I think my question would be "why didn't the bottleneck in point 2 vanish in point 3?" I think the only way this would be true would be if the bottleneck was algorithms, and there was a discontinuous jump in the capability of algorithms. I agree that in that world you would see a hard/fast/discontinuous takeoff, but I don't see why we should expect that (again, the arguments in the linked posts argue against that premise).