Inaugural post

[injygo]

This is a post intended to lessen the activation energy necessary for me to post. Note that I'm still figuring out tags, comments, notifications, etc.

To help me figure out the comment system, please leave a brief comment about exactly one of the following:

• your favorite animal
• the last book you fell in love with and why
• your opinion of eliezer yudkowsky
• whether you love the color of the sky

Comments

[thirqual]

#3: Holder of the most original way of showing he did not grok the practical side of science.

[injygo]

Can you elaborate on this?

[thirqual]

Sure.

See That Alien Message, specifically that part:

" A Bayesian superintelligence, hooked up to a webcam, would invent General Relativity as a hypothesis—perhaps not the dominant hypothesis, compared to Newtonian mechanics, but still a hypothesis under direct consideration—by the time it had seen the third frame of a falling apple. It might guess it from the first frame, if it saw the statics of a bent blade of grass."

Ignoring the part where the superAI has managed to make sense of what if was seeing through the webcam, it will absolutely not be able to do what is suggested with the 3 frames of the apple falling (or with the one of the blade of grass). The instrument simply does not have the resolution to allow such a feat. This is a key misunderstanding. This whole sequence is pretty bad tbh.

(the common way to not grok the practical side of science is to say "Singularity!". Every time you say "Singularity", an experimentalist has a stroke. Please think of the experimentalists)

[injygo]

I still don't understand what the misunderstanding is. How does science really work, such that the scenario in "that alien message" is not plausible?

The model of learning-about-physics I'm getting from "This Alien Message" is like this:
1. The intelligence receives some kind of data.
2. The intelligence wonders what it means, and constructs several theories, which may be far-fetched or complicated.
3. The intelligence receives further data, which supports or rules out various parts of theories.
4. Goto 2
Is your objection that this is unlike the way science works in our civilization? Or that there would not be enough data for the intelligence to consider relativity as a theory?

[thirqual]

The misunderstanding is in both how much and what information is contained in the data, how far you can make progress with it, and finally how one comes up/discards theories. The assumption is that the intelligence will come up with a specific model (General Relativity) as an hypothesis from the data. This is incorrect, the 3 snapshots of the fall of the apple (or the single one of the blade of grass) are grossly insufficent for this: there is no perceptible difference between the results of Newtonian mechanics and General Relativity on those pictures, and, very importantly, there is no basis in that data to build a relativistic model (specifically, which did not emerge from observations of falling bodies -no measurements precise enough to do so at the time-, but from issues in understanding the interactions between moving bodies and electromagnetism - there is *nothing* about the second in those data).

This is very important. It's not just a misunderstanding by Eliezer about the resolution of the data, but also about how you build theories from experimental data.

(I would have little objection* to a statement such as "given knowledge of the state of physics in the 1880s, the intelligence will come up with, at the very least a set of experiments to perform to resolve issues with Newtonian mechanics and, possibly, the first bricks of relativity)

*on those grounds

[youzicha]

Are you saying that there is not enough resolution to deduce either Newtonian gravity or general relativity, or just not enough resolution to distinguish them? If it's the latter, then I don't think this is in contradiction with the quoted passage.

[thirqual]

See the comment above for a bit more depth in the explanation. There are several other problems in the way for deducing Newtonan gravity from 3 stills of a falling object (e.g., you really really need non-linear motions to build that model with any degree of satisfaction, contrary to the popular story about retreating from plague and head trauma by fruit).

The argument one could oppose that is not really in contradiction with the quoted passage is the one from sophistication/complexity, as Eliezer allowed for "perhaps not the dominant".