> You think a good way to tell whether I'd be good at 
   > writing code is my opinion on philosophy outside my
   > domain of expertise?

It is a way (I'm not sure how to quantify it) to understand if someone can develop something that nobody has ever imagined before. My criteria for evaluating the response was whether or not the candidate could engage with the question. It is something that happened often in my experience at Google, you were presented with an 'impossible' requirements and it was up to you to figure out what you could do about it. If someone was going to say "Well I don't know anything about that, ask someone else." Then the company knew they weren't going to be as successful as we would want them to be.

In the case of the physics student it was just their passion, they had gone way beyond the course work because they were interested in interesting questions.

> If someone was going to say "Well I don't know anything about that, ask someone else." Then the company knew they weren't going to be as successful as we would want them to be.

Except that is EXACTLY the correct response -- "I don't know, but I understand that physicists have a good idea of what space-time is, and I'll review the literature and see what the scientific consensus is and we can work from there."

The exact WRONG answer, in real life if not the interview, is "Hrm, I'm not sure but how about I make something up on the spot!"

I completely agree with you.

When I said (paraphrased) "I don't know, ask someone else" I meant literally that. People who are asked a question about something they don't know anything about can either, do as you suggest propose a way to research the question (win), or do as I've observed refuse to go any further (lose). People who make things up on the spot generally are weeded out before we got to that point.

Not if the question is “how should we say this in Greek?“ and the guy in the next office knows Greek.

I dislike the one where they ask you how would you handle a situation that is highly dependent on the organisation and the organisation goals and procedures. The only answer really I give is not to answer such silly questions.

google interviews try to gauge how smart you are, not how well you can program

I fail to see how someone who is not a domain expert on the nature of space-time asking someone else who is not a domain expert can result in a reliable determination of how smart either person is.

To be clear, I don't think the correct answer of "I don't know, but I'll do a review of the literature and see what the scientific consensus is" is what the interviewer was hoping for, which is my contention.

You can have it both ways. I'd probably say something like "That's not my field of expertise, though if you'd like me to speculate I'd be glad to."

The point is I don't see how speculation about things you know nothing about is supposed to be an accurate gauge of a persons inate intelligence.

How you speculate about things you know nothing about is a far better gauge of problem solving skills than asking questions about an area where you can rely on learned facts, in my opinion.

It's not going to be accurate, but that is not the point. The point is whether you return a blank stare or start trying to reason and analyse the problem.

Sure, you might end up with any number of "I have no idea what the right value for this is - if it wasn't an interview, I'd look it up, but for the sake of argument let's say the value is X". That's fine. Your ability to make that leap and continue is more important than the right result (so is your ability to recognise and admit that you don't know and are substituting guesses)

What if I put it this way.

Do you think speculation on things that are an inferential step[1](or a couple of steps) away from things you currently know could be a gauge of a person's innate intelligence?

Intelligence is required to make inferences. Assuming a common starting ground, a person who could make the most logical inferences from it to explain a result would suggest that he's smarter. I do not claim to say a person who makes 10 good points is less smart than someone who makes 11 good points. But he surely you can agree that a person who is able to make 0 inferences is likelier to be less intelligent than someone who makes a hundred.

[1] Where my idea of inferential step is taken from - http://lesswrong.com/lw/kg/expecting_short_inferential_dista...

> Do you think speculation on things that are an inferential step[1](or a couple of steps) away from things you currently know could be a gauge of a person's innate intelligence?

No, I do not. It introduces a large capacity for things to go wrong, and it is very difficult for the interviewer to separate themselves from process. Speculative development is a road marked by dozens or hundreds of failed efforts before you hit on success. Either the interviewee happens to hit upon one of the few "correct" answers in the time available -- password guessing -- or some tolerance of wrong answers is to be accepted. But it is quite difficult to differentiate bad speculation from fruitless speculation, in a way that usefully differentiates candidates based on capability.

that is correct, however that's not exactly what the person in question was looking for while interviewing (not for the other person know have knowledge about the subject) but the ability to be able to understand a concept completely once explained, and work within the rules of the concept. This of course is arbitrary, but not a completely useless question.

This is not true. (I interview candidates at Google)

If you don't test a candidate how good a programmer they are, you shouldn't be too surprised when you end up hiring a lot of bad programmers.

Does Google tend to get a lot of bad programmers?

If you are hiring tens of thousands of employees, its impossible to expect all will be top of the line.

My guess is Google has plenty of bad programmers, or worse Imposters.