[moultano]

He's describing something that was the biggest factor for me not deciding to go back to school. I came to Google and started working on search engine spam. I also periodically read papers on "Adversarial Information Retrieval" as it is called in the literature. Only a handful of the papers I read had useful ideas, and I don't think there's anything that can really be done to improve the situation. There are certain things you can't really research without access to the real thing. Information Retrieval is one of them, and I'd be equally skeptical of parallel computing research unless the university has a 1000 node cluster to work with.

This sucks for the world, because the research in these things that Google does doesn't really escape to advance the state of the art elsewhere.

[Lewisham]

Sadly, most academics realize that they're data-starved, but industry isn't able to/doesn't want to work with academia to help gain access to that data. Even if you do get access, you're NDA'd so hard that publishing useful papers is a challenge.

It's a shame, because universities are essentially free labor, just on much longer timeframes and lower probability of immediate success. I understand why things are the way they are, but I'd love to find some way of reconciling this issue. It's already killed a research project of mine before it started.

[moultano]

Google realizes this too, and there are a lot of people here that would like to help. There are a few huge issues that always come up.

1. Paranoia about giving competitors the edge they need.

2. Paranoia about giving spammers the edge they need.

3. Privacy. It's nigh impossible to anonymize user data enough to release and still have it in a useful form. The AOL logs debacle hurt everyone.

One of my friends had a really interesting take on this. His belief essentially was that the industry advances when people switch companies and (legal or not) take their institutional knowledge with them. Unfortunately, academia is hard to transfer into.

[price]

<quote>His belief essentially was that the industry advances when people switch companies and (legal or not) take their institutional knowledge with them.</quote>

This is why it's so important that in California the law protects employees leaving companies to work for other companies -- this is legal there. You can't take code, data, hardware designs, or other concrete intellectual property -- that's forbidden by the agreements every tech company requires employees to sign. But if you couldn't take institutional knowledge with you, you couldn't work at a new company at all. (How could an ex-Googler unlearn how web-scale systems are built of many disposable pieces?) And if an agreement purports to restrict you from switching jobs, then California law repudiates that restriction.

Unfortunately some other states have no such law, and cheerfully enforce non-competes against engineers trying to switch companies. And guess what? The industry doesn't move as fast in Massachusetts as in California.

[kevinpet]

If I could point to one thing other states could do to improve their economy, it would be adopt the exact language of California's anti-non-compete and the "you own your own ideas, unless developed as part of your work" law.

Not that this is going to single handedly grow a silicon valley in Montana overnight, but it has so little downside. The reason for the exact language is because if they use the exact language consciously, they effectively import the case law surrounding those laws.

[hga]

Actually, I know of no other state that has such a law (although the Michigan legislature accidentally zapped their's for a fairly long period and research has been done on this).

I've never lived/worked in California/SV, but based on what I've read about it (e.g. Shockley -> Fairchild -> many companies is the golden example) etc. and how I've experienced on the East Coast (Boston and D.C. areas) a number of failures to even get started-up due to non-completes, I'm convinced that this one and only unique to California feature is the #1 secret behind Silicon Valley's long term success, its crushing of Route 128, etc.

[_delirium]

Georgia was one until two weeks ago, at least de facto. The language wasn't quite as clear as California's, but it was essentially impossible to enforce non-competes. A bunch of big companies managed to get an initiative on the Nov. 3 ballot to overturn that, though, and it passed, partly due to misleading language that's currently being litigated. The ballot question read, Shall the Constitution of Georgia be amended so as to make Georgia more economically competitive by authorizing legislation to uphold reasonable competitive agreements?, which is a bit, uh, indirect about the fact that its main impact will be to legalize noncompetes in GA.

It caused a bit of ruckus within libertarian-leaning circles in Georgia, with some people lining up on the pro side, arguing freedom of contract, and other people lining up on the anti side, arguing right-to-work. It was probably the only major issue where progressive groups and tea-party groups actually had the same position (against), but they were both greatly outspent by the Georgia Chamber of Commerce and business-Republican groups.

[hga]

Are you sure about that? I didn't study the issue at all, just read one or two items, but what I understood was that it closed a severability "loophole" that sounded like a very thin reed to depend on.

The very possibility of a lawsuit, even if you eventually won due to the strictness imposed on enforcement, is quite enough to quash a lot of startups. Sure, you can win, after N months and M thousand dollars spent not paying full attention to your startup (assuming you don't get TROed in the beginning). I'm also told that being the subject of a lawsuit is incredibility draining.

In California as I understand it your opponent won't even get to first base.

So, let me put it this way: in all the discussions of California's no non-compete regime, if Georgia was so good why did no ever say anything about it, as compared to e.g. Michigan? People have been pushing Atlanta as a potentially good startup area, why not advertise this advantage?

[fizx]

> His belief essentially was that the industry advances when people switch companies and (legal or not) take their institutional knowledge with them.

Ah, so pay $3.5MM to retain those people ;)

[edge17]

While I do agree, it's important to note that the state of collaboration between American universities and industry is far greater than any other country and has helped both grow. It's one of the problems in places like India, where there isn't much collaboration between private institutions and industry.

Of course things could always be better, but it could also be worse.

[kermit_de_fro]

Minor picky point, but universities have access to such computing power via time-sharing at national labs if it is not available from their own department.

[moultano]

And I do think that a lot of good parallel computing research gets done. It's just a lot harder.