Jason Calacanis has started a new pet project which he is determined to drive to the scale of the Yahoo, Google or Microsoft search in the area of web information categorization. It’s the Mahalo Project.
The case apparently seems to be simple: There are lots of algorithm based search engines that, thanks to all the SEOs, you get a lot of junk which isn’t meant to be there. We are starting to slowly see a reliving moment of how the web used to be till Google came along and put things in some sort of order. It might all be unravelling back to chaos yet again. [Needless to say, Jason calls all these SEO optimizers by very interesting and radical names – none of which are pleasant.]
Mahalo aims to sort through information for 10,000 of the top keywords, using people. Their task is to look up for specific and accurate information using google or any search engine, create the page and then manage it. It’s going to be a full-strength manual work involved here. Apparently with the venture already funded with Sequoia, the concept has gained steam and is already available in some format – though nowhere close to the envisioned experience.
Jason is also looking at crowdsourcing the entire deal, so that individuals can take up work, to create such pages with accurate information. They get paid for doing that.
Now, crowds of people contributing.. hmm… where does that sound familiar. Yep, Wikipedia. It is a well known fact that Jimmy Wales is working on a search engine project named Wikia. The correlation between Wikipedia and Wikia is that they’d both be contributed by the general public of users and consumers. Given the success of Wikipedia and the learnings in managing and stabilizing the quality of the content, the same can be put to very good use within Wikia.
So the question remains, Is Jason Calacanis going after Jimmy’s Project? And if this is the sort of future we are looking at, ventures which to leverage the “human” aspect and touch, what is the possibility of opportunities that countries such as china and India bear, with that context?
- Launching an Experiment to Spot Problems Worth Solving - November 2, 2015
- Introducing In50hrs Healthcare Edition - May 14, 2014
- Ideas Pitched / Prototyped at In50hrs Pune 5 - September 16, 2013
While we are at it, check this out too. Big names behind this.
http://www.chacha.com
No. I was just veering aroung to agree that AI would be better off trying to build an “unconscious intelligence” that may be finite in utility than to aim at building a “simulated conscious mind” with its boundless complexities.
“That leaves the algo to try, err, record, measure results, fine tune and reload. The author (with all prejudices of his own subjectivity) figures out its limitations, adds in additional context criteria and reloads it.”
are you suggesting something like Recommendation engine like Amazon . [ People who bought this book ,also bought these item ] or something like a learning algorithm based approach . do you think its a scalable way ?
i think if someone wants to capture that “intent” there will be too many parameter of profiling ,and way to many exception too handel .and order of magnitude increase in accuracy/ relevance of result will not be proportionate.
a IIT Delhi based company named Tensor Tech [http://www.t6india.com] is doing a search engine by applying tensor algebra on user profile and search history and they are using Google Serach technology in the back end . this seems to be a better approach to me .
Prashant,
“In case of algorithmic approach at least there is a way to know it and account for it . my point is variance is not that easy to ensure and aggregate in human operator scenario too.”
Well argued.
But what is our primary need? To get an objective search result that syncs close to the context what the querist (a human mind) has as possible. Discovering why it varies or how to account for it is secondary requirement – it just helps to fine tune the algo, an information that is of use to its author, another “human operator” as you say and not the querist, whose problem it seeks to solve.
That leaves the algo to try, err, record, measure results, fine tune and reload. The author (with all prejudices of his own subjectivity) figures out its limitations, adds in additional context criteria and reloads it. By then remember, it’s an impatient human mind at the other end – that’s constantly shuttling between analytical thought, common sense, analogical thought, free association, creativity and even hallucination – that keeps extending the realms of the context itself.
But yes, there’s a widespread agreement on the fact that software has a high chance of modeling an “unconscious intelligence” though limited in utility, yet useful (think of algo trading strategies used by hedge funds) – than going after a “simulated conscious mind” which even if gets built someday, that could be pretty useless (for context search etc.) since a conscious human mind has no technical limits of variance.