Enceladus — LessWrong

LESSWRONG
LW

Replying toCircumventing interpretability: How to defeat mind-readers

Enceladus4y

Circumventing interpretability: How to defeat mind-readers

". Protein folding representations, be as expressive as natural language"

- I think you're missing a [may] between "representations" and "be".

Replying toWhat an actually pessimistic containment strategy looks like

Enceladus4y

What an actually pessimistic containment strategy looks like

I have thought this exact thing for something like 2 years. I do think there are some potential backfire risks which make this an uncertain strategy.

slowing AI progress in the US and Europe, which is where OpenAI, DeepMind, Meta, etc. are located, does not neccesarily mean that AI progress in China slows as well. To the degree that you think US AI research will be more aligned than China's, this differential slowing of AI progress could be net negative.
slowing AI progress in legible companies and labs may push AI research into more illegible places, which may be less careful on net.
Raising the salience of AI, which this campaign would require could accelerate

Enceladus4y

Long covid: probably worth avoiding—some considerations

>>In the beginning of the pandemic there was a study that did heart scans that suggested that COVID infection generally cause lasting heart inflamation. It seems to me that there wasn't a lot of research to validate to what extend that's true and I see no reason to rule out that thesis.

My understanding is that similar findings occur in other diseases, especially in patients who are hospitalized, so I'm not sure that's novel to Covid.

Replying toLong covid: probably worth avoiding—some considerations

Enceladus4y

Long covid: probably worth avoiding—some considerations

General concerns which push me towards being much less cautious re: Long Covid than you, without getting into the specific details of studies:

We have studied COVID far more than other diseases, so we may be finding stuff that either isn't real or more likely, occurs with other pathogens (we just hadn't bothered to look that closely before.)
1. Upshot: worry less about Long Covid
In general, a strong bias against preclinical/in-vitro work because I think that kind of lit is not very predictive of real-world issues.
1. Upshot: worry less about Long Covid
I'm more pessimistic than you about almost everyone getting Covid. Even with theoretical new vaccines for Omnicron or next-strain that approximate the original vaccine

... (read more)