Hacker News Comments on
Past, Present, Future Vision of AI - Google and AAAI 2011
GoogleTechTalks
·
Youtube
·
2
HN points
·
1
HN comments
- This course is unranked · view top recommended courses
Hacker News Stories and Comments
All the comments and stories posted to Hacker News that reference this video.> Google now uses tons of data to make decisions about what to show me when I search, but neither you nor I have any idea what they are.To be fair, a lot of people at Google probably don't know either. Peter Norvig had a great talk [1] in which he discussed searching Google's code base for "naive Bayes" and found some code in 2006 with a funny comment:
> “And it was fun looking at the comments, because you’d see things like ‘well, I’m throwing in this naive Bayes now, but I’m gonna come back and fix it it up and come up with something better later.’ And the comment would be from 2006. And I think what that says is, when you have enough data, sometimes, you don’t have to be too clever about coming up with the best algorithm.”
[1] https://youtu.be/ql623nyCdKE?t=5m25s
I don't want to sound like an anti-algorithm-think-of-the-humans advocate...I love algorithms...but I think your statement:
> Google used to use PageRank, which was a relatively "objective" criteria in that at least it applied equally to all pages.
...besides being too oversimplyfing for even a simplification -- I don't think the evolution of Google could remotely be reduced as: first there was purity of math, then came the money -- but I also think that you overlook the bigger issue...No one argues that algorithms are statements based in mathematical truths. It's the decision to use an algorithm -- including the weighting of its tradeoffs -- that is decidedly biased and opinionated.
How is "a page should be evaluated by the number of pages (plus the authority of those pages) link to it" not an opinionated statement of the way information should be organized, versus the pre-Google algorithms of "If a page has a lot of mentions of a word, it must be particularly relevant to that word"?
The fact that PageRank had to be changed and modified as soon as people figured out how to create networks of backlinks should in itself be evidence that an algorithm -- and the decision to use it -- is not just objective truth in the way that 1 + 1 = 2 is.
By Peter Norving