I recently interviewed at Google's Pittsburgh office and they definitely do development there. I'm not sure how you're defining "core" though. The teams in the office work on three products that are kind of secondary to the Google experience. I think all of the core libraries, search, etc... is in the Mountain View office.

This "hit the ground running" mentality was pretty amazing to me in my recent job search. I was looking at quite a few companies. Most of my professional work has been in Python manipulating data in various forms. I decided that I'd be interested in doing something different and started looking at web companies. All of them have this idea that you need to be able to hit the ground running so I was rejected for quite a few jobs. I've looked into web programming in my free time and there's really not much to it. I could easily be up to speed within a month but as you said, this is completely unacceptable to many companies for some reason. I laughed when I saw many of the companies still trying to find someone to fill the role I was rejected for 6 months later.

When I hire people (developers) I focus on the following things:
1) Do you understand the basics of programming? The exact definition of "basics" will vary by the job. Writing basic scripts probably just requires an understanding of how to use classes, functions, etc... Some more advanced roles might require more complex data structure and algorithmic knowledge but this really needs to be targeted to the role being hired for.
2) Can you think? More specifically, give the person a hard problem and see if they're able to adjust their thinking when it becomes apparent that the initial approach is flawed
3) Does you ask questions? I've found that people that won't ask questions in an interview also won't ask questions on the job leading to them taking many wrong approaches. I intentionally under specify things to encourage questions.

That's about it on the technical side in my opinion. If they have those things they're probably able to learn anything else you require. I have some additional things I look for while hiring into my current company because I know the personality characteristics required to survive here but that's going to vary from company to company.

I've interviewed at Google several times since 2007 including a couple weeks ago. I can confirm that their hiring practices haven't changed. They released a statement that basically said "Nobody knows how to hire" but that was the extent of it.

I used to say I had no particular brand loyalty. Then I bought a couple of AMD cards in Crossfire and now I can wholeheartedly say that I side with Nvidia. The number of graphical issues, blue screens, etc... I had to put up with was absolutely ridiculous. When I ran a single card it behaved much better but still had a lot of issues. In short: the drivers were absolutely horrendous and this was on Windows which is presumably where most of their effort goes.

My problem is getting people to listen to my estimates instead of revolting when they hear something they don't like. I typically don't need to estimate long projects so my estimates are typically on the order of days or weeks. As a result I can usually spit ball my estimates pretty accurately off the top of my head. A recent example was a project that I said would take two weeks. My boss balked and said "Why will it take so long? I'll put down one week". Sure enough, the project took 2 weeks.

That's definitely possible and I'd even say likely. Valve does have a preference for PC and anything from them since HL2 on PC has required Steam if my memory serves correctly.

I was hoping for Half Life 3 but since it was launched under a "living room" moniker I was positive that it wasn't going to be. I also highly doubt they'll make Half-Life 3 SteamOS exclusive. Valve doesn't make any extra money off of SteamOS as opposed to someone having Steam on Windows. Also, it's just not Valve's style.

This is an interesting problem that I like to talk about fairly regularly. Society is going to go through a very painful transition as more and more jobs becomes obsolete due to technology. This is really something that a lot of people should be thinking about. How will society function when nobody can get a job because machines can do it faster, better, and cheaper? Profit based growth no longer makes any sense as we approach that point. What's the alternative and more importantly, how do you handle the transition?

To what degree? CMU had driven a bus cross country with very limited manual intervention when I was an undergrad nearly 10 years ago. The bus only worked on the highway and the "driver" told it when to change lanes and took over to exit the highway but it made most of the trip by itself. I heard this story from Takeo Kanade but unfortunately my Google-Fu is failing to turn anything up so it's possible I'm mis-remembering.

Google's self-driving cars have driven over 300,000 miles without causing accidents while in automatic mode. There are two accidents I'm aware of. One, the car was rear-ended while stopped at a red light. The other was when the Google car rear ended another Prius. In that incident the car was in manual mode.

The assumption you seem to be making is that driving AI needs to be perfect. I'll save you the trouble: it won't be. It just needs to be better than humans. That's something that any auto manufacturer should be able to prove quite easily and would likely become part of standard safety ratings.

A far larger concern should be users that decide hacking together their own driving AI would be fun or those that maliciously hack into other cars. Even these are likely to be relatively small concerns.

I hate when people do that. If you're passing just speed up a lot then slow back down once you're back in the slow lane.

Seriously? 100,000 miles? You realize that's along the lines of 6-7 years of driving for people 19+ right? I'm hoping I just missed a really obvious joke

From arguments I have with my wife on this topic, it's all about not being comfortable with the loss of control. She doesn't even drive yet she insists that manual overrides need to be available in case the algorithm messes up so the human can take over and salvage the situation. I've tried explaining that this is unrealistic anyway because in a vast majority of cases the human is just going to mess things up worse if they take over. The human will take over in cases when they shouldn't since they'll think the machine lost control when it's actually fine. The rare instance when a person taking over actually helps things will be greatly overshadowed by the instances when they screw things up.

People are just uncomfortable about not having control and are also disturbed by looking at things through a statistical lens.

Last I saw Google announced that they had over 300,000 miles with their autonomous cars. That was in August 2012. Still impressive but nowhere near millions.

I find DUI laws to be ridiculous. Just because you CAN take over the vehicle doesn't mean you will. The law should be reserved for those that actually have taken control of the vehicle. Of course, reality means that you are probably correct in your assessment so that needs to be considered.