Log in

No account? Create an account

Previous Entry | Next Entry


GFGI or Google Is Your Friend

  • 2nd Mar, 2005 at 1:03 AM

As I was walking home tonight, I had the sudden realisation that I understood what Google is actually about.

You see, Google has convinced the world that they are a company with the following attributes:

  1. They will do no evil.
  2. They specialise in search technologies.

This got me thinking that they actually only have an ethics officer because that's a clever way to get people to hand over their private information. When Microsoft asks for people's home addresses, telephone numbers, and their sexual orientation, people raise this huge stink. However, when Google does it, people are happy to fork over all their personal information. After all, Google isn't evil. Heck, why don't I write about my entire life and route all my personal e-mail through them as well?

As for the second point, I have realised that it is also a lie. Not a complete one, otherwise it wouldn't be plausible. After all, they do index the world wide web. But the thing that makes them money is intelligently correlating their search results with advertisments. Or is it vice versa? When it comes down to doing such things as desktop searching, or searching Usenet, or searching news sources, you can come to see that Google is actually the world's first successful artificial intelligence company.

Every AI company before has always made the horrible mistake of calling themselves an AI company. So people actually expect their software to make reasonably intelligent decisions. No scratch that, they expect the AI to make really awesome decisions, since computers are (after all) infallible. However, Google positions itself as a search company, and so if things aren't perfect, who cares? And if there's a little more heuristically good correlation, no-one will particularly notice.

What a clever hack!


( 3 comments — Leave a comment )
2nd Mar, 2005 08:26 (UTC)
Not evil — I think a restatement of what you said is that, by appearing to be not evil, Google is more likely to succeed in the marketplace. There's actually a strong culture among Googlers (including the founders) that the best way to appear non-evil is to be non-evil. On the other hand, there's a definite concern that if you hand over your life to Google, you want them to not just be non-evil, but stay non-evil, which is a lot to expect out of a corporation whose not-so-secret objective is world domination.

As far as AI, in a sense, everyone is doing some form of AI these days, and no one is stupid enough to actually call it AI.
3rd Mar, 2005 16:42 (UTC)
Agreed, on both counts.

Which is why, even though I read and manage most of my mail via gmail and actually prefer using it to Outlook 2003(!), I still primarily use my uwaterloo address (which is set to forward copies to gmail) because I have a greater degree of control over how I manage messages arriving at it.

And in terms of AI: yeah, look at things like CRM and data mining and any sort of current DBMS R&D. The trick seems to be getting people to view the software not as making intelligent decisions for you, but instead helping bring things to your attention or optimize things in the background without affecting the actual results.

To put it into filtering terms, when the cost of a false negative is very high (i.e., an ai makes a wrong decision where no one is watching it but nonetheless depend on its correct performance) people are not going to trust the system unless it is proven to be very, very, very reliable and effective.

But when a system is instead there to enable some previously intractable task -- well, hey, if it misses something, no problem; it's not like you would have been able to notice it on your own anyway. Similarly, for something like database optimizations, the worst case is that the DB performs about the same or slightly slower than it would without the optimizer, with the best (and average) cases offering power(s) of magnitude speed boosts.

The market for AI technologies is not in taking over roles previously filled by humans; it's in making previously-intractable jobs possible by doing what a human reasonably cannot.

I saw something, somewhere (likely some other entry on my friends list) that suggested viewing Google's business as that of being the world leader in clustering... it's an interesting contrast to this.
22nd Dec, 2006 22:38 (UTC)
RTFM (read the f'n manual) has now been replaced by GFGI (go f'n google it).
( 3 comments — Leave a comment )