The article didn't give the direct response but quoted another speaker at the conference who said "Over the past 50 years, we've lowered the bar for what we consider AI," -Stephan Nusser, senior manager of the User-Focused Systems research group.
The essential defining component of AI to me is an explicit attempt to model or study human cognition/neural processes.
A lot of the learning/search stuff being published is a mix of traditional engineering tools (like Kalman filters, used for decades for state estimation) with some kind of search/learning metaheuristic and does not give a flying hoot about being "intelligent".
Decades ago we used Kalman filtering in naval navigation systems - it was well-established engineering then (Rudolf E. Kalman developed his filter in 1960). The original article's title is somewhat misleading in that Kalman filtering isn't usually considered to be AI.
But it's very nice to see this article anyway, if only to know that those old boat filters are creating Yahoo's home pages. Hey, it might even be the same code!8-)) [Tip to Yahoo: Speed 'er up to 23 knots and, if you start getting significant outliers, you've got a rounding problem in your matrix routines.]
Tom may like to read about Paris Hilton and Harry may like reading about what President Bush had for dinner. I am not interested in both, but I do bother if some company gets funded, if some website gets bought or if someone writes a 'hello world' app in a hundred languages. I don't have the same interests as Tom or Harry. And featuring their likes on my headlines just because they clicked doesn't make an uber-cool feature or a handy tool. To me, it just sounds like a bad automated-Digg clone.
This is totally a commercial idea. They seem to be interested in the clicks and henceforth the dollars. The idea would have been a runaway hit if they tried to personalize news as per personal interests.
Well, this system cannot personalize news until they ask me to specify my interests and tweak the news personally for me.
It can be done with a Yahoo Pipes app ofcourse. But its still like using a calculator for 3+4. How about a direct feature - "personalize ur news".
One way to read such news is to use the Google News Search RSS feeds. It works for me.
I mentioned just a few days back, that I was working on a news engine - for news analysis and news search (it's for a contest and I'm working on it since I am a news buff). The idea is currently half complete since I have my exams on(will be over by 15th of this month). I found a simple way to personalize news :)
I just gave the idea a try and it does seem to be crappy, but its an easier way because that's the simplest data I have access to as a developer. Will post it here soon. I have to convert the app from php to python (going to run it on Google App Engine)
John McCarthy; that name reminds me of something...
"No one has yet buit a monument so high that a bird can't fly over it and shit on it" --John McCarthy
That's my favorite quote.
A bit of info if you don't know who John McCarthy is: the man who coined the term "Artificial Intelligence"
| This is totally a commercial idea. They seem to be interested in the clicks and henceforth the dollars. The idea would have been a runaway hit if they tried to personalize news as per personal interests.
There was a great magazine article sometime back -- that I can't find now -- that featured a guy who'd been in the radio business for a long time complaining that the ratings and rankings system that he'd developed effectively killed radio, causing so many stations to all play the same homogeneous crap. Yahoo's system, and news in general reflect the same problem: following the money leads to this steaming pile of crap that a lot of people seem to hate.
But there's another edge to this too. People who claim to be interested in specific subjects, or in meatier or niche news, are still drawn in to some of the vacuous material. So, nobody actually really wants a completely personalized news service; I'm convinced that that's one reason why jaanix hasn't made it big, while Reddit continues to steam along. People will tune Jaanix for the news feeds that they think they ought to like, and then they quit following it after a while because they'd actually rather click on all of the [pic]s on Reddit's front page.
Hey! I had never heard of Jaanix. Thanks for mentioning that. It's interesting. Its like some twitter for news, minus the phone syncing(I summed it up like this when I saw the follow button for news keywords).
But I couldn't figure out what the "follow [keyword]" button does. I clicked it. AndChecked the FAQs but found nothing. Even their FAQ page has voting system. A bit weird. They should have had a normal FAQ page.
The sliding bars on the homepage to customise the news you want is good. But I am able to notice a bug in it. I am using Firefox-3 on windowsXP and when I click on a sliding bar, the marker moves to the position, but the green shade seems to be somewhere else. I had to do around 10 clicks to get the green shade and the marker to coincide (surely had some fun with it).
And whenever I changed the slider the news gets optimised by Ajax. That's cool again.
I noticed they allow public voting on their news items. They are easily exposed to 'click' spam. Not requiring users to signup to vote actually seems like wanting to be spammed.
And I also found out that they don't requires users to signup to comment. Seems like some sort of Open Digg or Open Reddit.
But one thing I found nice on Jaanix was that their youtube channel(not sure what they call it) at http://www.jaanix.com/youtube they have this 'quick review' kind of feature for the videos. Just place your cursor on the video thumbnail and you will see more thumbnails of the video.
Well, Jaanix or Digg or Reddit, they are all cool for user submitted news. Maybe all the other mainstream news might also be subbmitted. But you don't see all mainstream news there. You might surely find out the Release date for iPhone 10G on Digg, but there are surely some news items that don't reach the top stories even in the section news.
Those are covered by mainstream news fellas. In simple words, you can make a pizza at home. It'll surely be tasty. But still you will like McDonald's Pizza. It's like... two different good tastes, so you like both :)
And still I haven't found any source except Google News Search feeds for personalized mainstream news.
What I mean by mainstream news is - news by news agencies like Associated Press etc (I don't know what these agencies are called).
But your point is right. People will surely find that they are interested is more stuff when they follow something randomly. Again: all that's going to be featured on the front page of Y! is celeb gossip for sure. Not everyone likes celeb gossip. So its going to be forcing people to read celeb gossip. I would rather read news on any other source like Digg or Google News or Jaanix (as you said random news is good).
I have a Q and also and A:
Why did Yahoo choose celeb news over any other topic?
Because they found that they would get more news news in the celeb news category than any other. Which translates to more links, which in turn translates to more clicks and then more money.
On one hand they have a really cool thing like BOSS and the other they have a plain stupid idea of dollars from celeb gossip.
I've been following (and rooting for) jaanix for a long time now -- pretty much since their debut -- and it's strange: I really really like their idea, and I really really like that they've put a ton of work into improving it. Yet, I have a really hard time sticking with it myself; I find myself gravitating instead towards Reddit, which frankly I have come to loathe. I don't understand it, and I wish I did.
| Why did Yahoo choose celeb news over any other topic? Because they found that they would get more news news in the celeb news category than any other.
That's an interesting angle. I have to chew on that for a while. I always just figured that it went that way for the same reason that all the cheap glittery celebrity gossip mags line the supermarket aisles: it's everyone's guilty pleasure. They sell more copies of "People" in a week than they'd sell of "Discover" in a month.
But, it could still come back to the news always being new, that people are drawn to junk like Britney's latest embarrassment because it's always new, even when it's really still the same old thing.
Or, maybe they're just barely literate gossip hounds that get some kind of psychological positive feedback from finding out that there are other people even more screwed up than they are. ...
I've had a yahoo mail account since before the internet. It's my spam and bacon account. And my habit is to go through yahoo.com to get to mail. So I see all the stories and I've never clicked on a celebrity gossip story. Once in a blue moon there will be a science story and I've clicked on a few of those. So what do I keep seeing over 99% of time?
Decades ago we used Kalman filtering in naval navigation systems - it was well-established engineering then (Rudolf E. Kalman developed his filter in 1960). The original article's title is somewhat misleading in that Kalman filtering isn't usually considered to be AI.
But it's very nice to see this article anyway, if only to know that those old boat filters are creating Yahoo's home pages. Hey, it might even be the same code!8-)) [Tip to Yahoo: Speed 'er up to 23 knots and, if you start getting significant outliers, you've got a rounding problem in your matrix routines.]
One of my favorite stories from my grandfather was how he used Fourier transforms to calculate the weight of cotton on the fly as it was being processed in the machines he developed. This was high technology in the 60s!
I am not linkbaiting. If you check the article's title, you will see that it is unchanged. The title comes from this part:
"Agarwal calls it Coke for short, and his fellow researchers are affectionately known as Cokeheads."
Also, I don't like editorializing and injecting my opinion and pg doesn't as well (I'm guessing he's the one who reverts headlines to their original state when they're editorialized). I've changed few headlines in the past but only when they were grossly misrepresentative of the article or just plain crappy and didn't convey the content of the article.
So, if you have a problem with the headline, complain to The Register please :).
I don't go to the Register for news, so their linkbait headlines aren't a problem for me until someone submits them here, even if it's unchanged. So I will complain here.
Changing a "curiosity" headline to make it clear what the article is about is not editorializing (except in the most literal sense, so let me preempt any pedants on that point) or injecting opinion.
Example: "Yahoo! Researchers develop algorithms to select stories for front page"
Is that less intriguing? Yes. Will it get less karma? Yes. Will the people who are genuinely interested in the topic covered still click on it? Yes.
pg has reduced the length of headlines few times now. It now stands at 90 chars. Also, pg has changed few of my headlines that I amended with some more info from the article and has reverted them back to the original. What do I think? I think pg is right on this issue. Rewriting headlines is a slippery slope since we all have our biases. I'd rather read the original headline and make up my own mind on the issue than to read what nickb or jraines thinks this article is about.
What you're asking for would be better served with a blog and not a link aggregation site.
Finally, I'm not getting paid for this work and I don't have much free time. Personally, I don't wanna waste my time trying to think up headlines since that's not my job.
Just a couple of days back I read about SocialBrowse here. The idea was cool(it did sound to me like a live and interactive version of Hacker News). It would be interesting if Yahoo tries to deliver headlines that way... like a popup on my taskbar whenever a 'friend' of mine likes a news headline. I mentioned friend in quotes because I mean in terms of the 'followers', like in twitter. I can then follow the person if I like the news he/she reads.
Great jraines! You've got a point there. I would like to see headlines like the one you mentioned in the example. Straight-to-the point headlines. It makes reading easier. I think this point should be mentioned in the submission guidelines for Hacker News(and would better if its a rule). Simple headlines save time too. Example: I spend most my of my time with no access to a computer, away from home. I am browsing on a mobile phone with 8kBps internet speed. And simple headlines saves me page loading time. I usually avoid clicking on headlines that are confusing. (and thats why I use mobile lingo in most of my comments) :)
would like to see others' opinions on this matter(its a pretty serious one).
I love you, John!