LessWrong 2.0 Reader

Hmm, I have sympathy for this tag, but also I do feel like the tagging system probably shouldn't implicitly carry judgement. Seems valuable to keep your map separate from your incentives and all that.

Happy to discuss here what to do. I do think allowing people to somehow tag stuff that seems like it increases capabilities in some dangerous way seems good, but I do think it should come with less judgement in the site's voice (judgement in a user's voice is totally fine, but the tagging system speaks more with the voice of the site than any individual user).

Oh, yeah, admins currently have access to a purely recommended view, and I prefer it. I would be in favor of making that accessible to users (maybe behind a beta flag, or maybe not, depending on uptake).

I don't know the answer to your actual question, but I'll note there are slightly fewer mech interp mentors than mentors listed in the "AI interpretability" area (though all of them are at least doing "model internals"). I'd say Stephen Casper and I aren't focused on interpretability in any narrow sense, and Nandi Schoots' projects also sound closer to science of deep learning than mech interp. Assuming we count everyone else, that leaves 11 out of 39 mentors, which is slightly less than ~8 out of 23 from the previous cohort (though maybe not by much).

New concept for my "qualia-first calibration" app idea that I just crystallized. The following are all the same "type":

1. "this feels 10% likely"

2. "this feels 90% likely"

3. "this feels exciting!"

4. "this feels confusing :("

5. "this is coding related"

6. "this is gaming related"

All of them are a thing you can track: "when I observe this, my predictions turn out to come true N% of the time".

Numerical-probabilities are merely a special case (tho it still gets additional tooling, since they're easier to visualize graphs and calculate brier scores for)

And then a major goal of the app is to come up with good UI to help you visualize and compare results for the "non-numeric-qualia".

Depending on circumstances, it might seem way more important to your prior "this feels confusing" than "this feels 90% likely". (I'm guessing there is some actual conceptual/mathy work that would need doing to build the mature version of this)

Yeah, I should use that. I'd need to remember to unbookmark after reading it I suppose.

we are outside in front of the entrance

Further observation about that second sentence.

It might very well be cultural. I am American and I literally cannot imagine doing either of those things with a woman I am not in a relationship with. I've never seen anybody else do them either.

In addition to that from my perspective, I think that if every day of the year you consume the same amount of potassium you (as a typical office worker) likely consume either too much or too little on some days. 

I think you are assuming optical transistors and photonic computing are the same thing, but they are not. Optical transistors are a component that could be used for photonic computing, but they are not necessary, and companies may have a better shot at getting photonic computing to work at scale without them.

Optical transistors try to function similarly to electronic transistors but use photons instead of electrons for signal processing. You are correct that optical transistors are not currently not great and it's an active area of research to get it to work.

However, photonic computing is a broader concept that may or may not involve optical transistors as some of its components. Given the limitations of current optical transistors (as you point out), I understand that companies working on this typically use alternative photonic techniques to make it more feasible and practical for deep learning matrix multiplication.

Optical transistors are just not as technologically mature (and may never be) as photonic components like modulators and waveguides. For example, the paper I linked in the post is titled "Experimentally realized in situ backpropagation for deep learning in photonic neural networks", they do not use optical transistors. Instead, they use some of the following components: Mach-Zehnder interferometers, thermo-optic phase shifters, Photonic integrated circuits, and Silicon photonic waveguides.

The final setup allows for matrix operations for backpropagation.