[extropy-chat] Year 2014: Googlezon.com

[Adrian Tymes]

--- Jose Cordeiro <jose_cordeiro at yahoo.com> wrote:
> http://epic.chalksidewalk.com/

The projection makes one IMO fatally flawed
assumption: that automated reccomendations work, and
work really well.  I, personally, am usually far more
disserved than served by Amazon's reccomendation
engine, and I see the links on Friendster reflecting
more false friends rather than true communications.

Perhaps if Friendster were linked to Gmail, with an
option to promote/score higher those friends with whom
a user exchanges email a lot - the real intent of
which would be to weed out the vast array of false
links, which were fed into the system but do not
represent a true social tie.

But I, and those I know the shopping habits of, treat
Amazon like any other store and almost never use their
reccomendations - except occasionally as shortcut
links, when we were going to Amazon to buy that item
in the first place as well as the item that caused the
reccomendation, yet Amazon thinks its reccomendation
had something to do with it because that's the link we
followed.

There's also the problem of taking snippets of
different news stories out of context and creating a
news article from them.  It might make a good demo,
but actually being able to understand the context
enough to pull that off intelligibly in most real
world cases would require far more AI progress than
we're likely to see in the next 5.5 years (give or
take half a year).  See how slowly Babelfish has come
along, for example.