i love TED talks

[rodrigo]

i was just watching this one, it's pretty interesting.
B8ofWFx525s

[CarlaRant]

Hmmm...interesting tidbits. I wonder which search engines are avoiding these personalized alogorithms.

[rodrigo]

Hmmm...interesting tidbits. I wonder which search engines are avoiding these personalized alogorithms.

there may be some, but in any case i dont think this things are a bad thing in general. for people who are interested in a bunch of stuff, it wont necesarily hide important things, it may hide some stuff, but you're able to find em. problem is when it happens with people who havent shown any interest in important issues, cause then it'll just keep replicating the tendency of being not in the know... i guess.

[xsecx]

there may be some, but in any case i dont think this things are a bad thing in general. for people who are interested in a bunch of stuff, it wont necesarily hide important things, it may hide some stuff, but you're able to find em. problem is when it happens with people who havent shown any interest in important issues, cause then it'll just keep replicating the tendency of being not in the know... i guess.

it's not just that, it's having technology try and second guess what you may be interested in. It's in interesting problem. What happens when you have too much data? Do you take the extra time to just sift through everything or do you hope that based on past searches and behaviors that there's an algorithm that can try and guess that because you like fishing when you search for hook, they return fishing information and not stuff on hookers or prosthetic limbs.

[xCrucialDudex]

Hmmm...interesting tidbits. I wonder which search engines are avoiding these personalized alogorithms.

I believe you can just turn off the personalized search and get a "standard" Google back in your browser window.

http://www.google.com/support/accounts/bin/answer.py?hl=en&answer=54048

Creating a second account and turning off personalized search for it, or simply using Google without any associated account and turning off personalized search, then accessing google.com in a different browser (say Opera if you usually use Google Chrome where you sing in to Google with your main account) will give one standard Google search results :)

[xCrucialDudex]

it's not just that, it's having technology try and second guess what you may be interested in. It's in interesting problem. What happens when you have too much data? Do you take the extra time to just sift through everything or do you hope that based on past searches and behaviors that there's an algorithm that can try and guess that because you like fishing when you search for hook, they return fishing information and not stuff on hookers or prosthetic limbs.

As I see it, ultimately the question is can we know what we will consider as relevant in any point of time in the future?

In my personal experience my interests sometimes shifted radically, I'd become interested in something I'd hated, or in something completely new and utterly unexpected, something I didn't know or couldn't even think of before it happened.

Is it just me or are we really asking the question "can we learn to predict the future accurately?"

[xCrucialDudex]

Btw love TEDTalks too!

[xsecx]

As I see it, ultimately the question is can we know what we will consider as relevant in any point of time in the future?

In my personal experience my interests sometimes shifted radically, I'd become interested in something I'd hated, or in something completely new and utterly unexpected, something I didn't know or couldn't even think of before it happened.

Is it just me or are we really asking the question "can we learn to predict the future accurately?"

that's the exception, and as such would just be a victim of optimization. Most people don't radically change things, so it's possible with enough data to go through that you can predict based on past behavior what someone would likely be interested them and put them first. Otherwise you're eventually never going to find what you're looking for because there will be no way to prioritize it and determine which should rank higher.

[xCrucialDudex]

that's the exception, and as such would just be a victim of optimization. Most people don't radically change things, so it's possible with enough data to go through that you can predict based on past behavior what someone would likely be interested them and put them first. Otherwise you're eventually never going to find what you're looking for because there will be no way to prioritize it and determine which should rank higher.

No, when I think of my friends I can clearly see that they would pick up some new hobbies or develop new interests that were totally unexpected. One of my friends, who used to be really into software development, eventually started taking up Hapkido classes and bought himself a camera, he suddenly went crazy about martial arts and photography. He never really thought of himself as a person who loved martial arts or art in particular before that. He never really talked about it much except stating on occasion he was clueless when it came to art. He thought of himself as a software developer mostly.

I mean you can watch one film and that may determine a new direction in life for you. Almost momentarily an interest in a previously unknown area to "recommendation algorithm" (let's call it RA for now) emerges. Few people go to IMDB and click Facebook Like button or divulge their interests in other similar ways online. How exactly a RA would figure this out?

With most Internet users being younger people some 15 to 30 years old, approximately whopping 50% of that crowd is undergoing constant personal development with swiftly changing values and interests on an almost weekly basis, that also are unlikely to bother to report those changes back to the Net... I can't really see how pulling all the scarce bits of information users willingly or unwillingly divulge can provide enough of data to reliably predict anything.

I have quite a few accounts on social networking web-sites and I generate a fairly bulky list of activities that includes information about what music I listen to, what films I watch, what books I read, who inspires me, etc. and yet rarely Facebook's recommendations present anything relevant to my interest, or Google for that matter. It looks more like guessing at this current stage, even with all the open data I willingly choose to divulge. In fact, turning off personalized search on Google for a while gave me more interesting new results than before.

That kinda makes sense, because people are usually drawn to novel, what RA currently try to do, or it only looks like they're doing so, is reinforce whatever interests one already has. Basically, this is what the guy in the TEDTalk video said if I recall correctly.

I've been paying some attention to Last.fm from its early days when it still was just an Audioscrobbler service, and MusicBrainz library that kinda both attempted to provide good music recommendations. Through all these years little have changed, they still work rather poorly, no matter how much they know about my tastes in music (and I've provided more information than average users do).

Fundamentally, I think we just can't reliably predict what will be relevant to our values and interests in the long run. Anything extra, outside of the current scope of interests becomes very hard to predict, yet very often it is exactly what sparks a genuine interest, the novel and previously unknown.