Search This Blog

Saturday, January 27, 2007

Why Artificial Intelligence if you can have Peer-Peer Intelligence ?




It is interesting to note how competitive the search world is getting. The eternal quest for ‘finding the right answer’ is leading this technically advanced generation back to the basics – why not utilize the intelligence of the actual human brain that we are all trying to simulate ?

Solutions like the recently discontinued ‘Google Answers’ and the currently blossoming Yahoo Answers are examples where you can ask targetted questions to ‘experts’, who can provide you with educated answers. However, in this model, the ‘experts’ are usually more qualified than the average Joe. The power of involving real humans is when you can expand in scale, as well as make it attractive for those contributors, in terms of personal profit. In other words, a model that can utilize the ‘spare time’ of millions of people to offer collective intelligence. Another example of this is the recently launched Linked In’s Question tab. For those who don’t know, LinkedIn is a business networking tool that allows you to connect to several industry professionals. I use it a lot as well. Now that LinkedIn has a great network of Friends, Friends of Friends and so on, it makes sense then, to traverse this network to get targetted information. So they launched Linkedin Answers, where a person can ask a “Question” and anyone in the network can answer it. Questions range from “What is the one killer-app you would like in your cell phone” to “How can we make the law more user friendly?”. The answers come from your network, which may include engineers, VPs, CTOs, CEOs of companies who you get to tap for information. The answers are detailed, and often fabulous. I bet you could’nt get the same quality from a plain google search.

Similarly, Google went the human route for Image Search. While you can apply complex word algorithms to textual articles, how do you provide relevant image searches ? How do you know a particular image represents a search query like “long tail theory”. Simple, tap into real humans with the Google Image Labeler. Make it a game, get information in return. Beats the heck out of AI engines.

And then, yesterday, I used ChaCha, the guided search engine. The logic here is again simple, utilize the free time of millions of peers – and have them search for you. They have a huge network of guides, and one guide can invite another. When you launch a ChaCha search, and choose the guided option, a real user responds and chats with you to refine the search. I spent the day talking to several of them – they are real people, who get paid for the hours they log into ChaCha. Depending on which ‘level’ they are graded at (a function of user feedback on their search report, and other factors), they can earn anything between $5 to $10 an hour. I talked to one Guide, a mother of three, who says she loved this ‘part-time job’ – when she needs some extra cash, she logs in, helps people and earns money. When they get ‘promoted’ to a higher grade, their pay can double, she says. I asked if they are trained, and I was told they do go through some level of keyword search training. Personally, I did not find their level of expertise to be any better than the average user (I think I can search far better with advanced search parameters). The challenge, I think, with ChaCha would be to train their people so that more people see a benefit of using their guides. I tested them with search queries for which I knew the answer and waited to see how long they took to get me the right answer but it was an interesting model – I am sure there are many who would find it interesting. It also seems that their payout is heavily based on how I rate their help, so not a single one tried to hurry me. Infact, we even chatted about general things while the Guide searched. I wonder then, if ChaCha like systems could effectively merge search as a service into a larger social networking tool!
See their blog for other ways to increase your payout using this service.

All in all, it is very interesting to see how the human collective is being utilized to provide further personalization to the world of search. While one may have thought that 'Simulated Intelligence' would eventually take jobs away from people, I guess the Internet is balancing this well by reaching out to people and offering them new ways to 'mesh into this new world intelligence' and supplement their living. Infact, I was at Internet Telephony last week, where I spoke on 'Effect of Web 2.0 on Telco World' and in one of the keynotes, Rich mentioned that this particular event was full of Military people, because they wanted to learn how they can use VoIP technology to help those disabled in war to earn a living !
Vive-la-Internet.

3 comments:

  1. Yes - absolutely right. However you need to do two things, both of which are classed as AI.

    You need better indexing. You don't want stations to be elastic. (The season of spring was translated by Google as La estacion de ressorte Clearly the comment must be appropriate. Unless you can get a good Spanish translation appropriatness is not guaranteed. I have arued that bueno espagnol = Turing. That is to say if you have a petabyte database and you always fetch out an appropriate part of it you have passed the Turing test. A chatterbox, of course, maintains a database of its utterances and your replies to help it further to find that appropiate response.

    ReplyDelete
  2. Ian, not sure if this is a test of yours, but I don't think it passed the Turing test for me ;-)

    ReplyDelete
  3. I have gone through the above article and it is interesting and I appreciate to the author and I to have a link related to refer with more information. and I feel it will be helpful to you.voip guide

    ReplyDelete