Thursday, September 18, 2008

Thought Networking (from Primal Fusion)

This may be unknown to anyone outside Waterloo, but yesterday a local startup (Primal Fusion) which was in stealth mode for over 3 years talked about what they are doing for the first time. It was interesting to watch the understated local hype since it was run by Yvan Couture, who had succesfully sold his startup to Agfa for a good profit and is also quite active in the community. The only reason I was interested was because I had listened to a talk by Yvan recently at UW. He seemed like a genuine and trustworthy guy which is rather rare in business nowadays.

The core idea behind the company is related to thought networking. They hope to define what it is and be the leader for a new market category. The simple example given was to go from the approach of searching for something to have the required information delivered to you once you have defined your thoughts. Their goal is to capture your thoughts and represent them as semantic networks. After this, these semantic networks representing your thoughts can now be linked to the semantic networks built by the company using information in the web like new sources, wikipedia etc thus enriching your original thoughts. Finally, since your thoughts are captured instead of a static profile as in the case of a social networking site your thoughts can act as a proxy for yourself for frameworks like software agents. Sounds cool!

Finally my personal thoughts with the limited information I had (didn't get invited to private beta yet). It is a cool idea to be able to capture my thoughts online and have them act on my behalf, maybe you will chat to my thoughts next time on Second life. But, there are two key challenges faced by such an effort and I didn't get a good answer for either. It is difficult to capture anything but simple thoughts as there would always be assumptions. Semantic networks doesn't quite solve that problem. The second issue the classic problem with similar AI based systems. The system is a black box which does the magic and links your thoughts to come up with decisions on your behalf. But, if you cannot describe the mechanism of how the system thought about it there would be a trust problem for humans in a world of software agents. The lack of trust might be ok if you just come up with some news articles or travel recommendations for me.
The company is probably betting itself on web 3.0 where semantic networks are expected to replace today's web. If the market does pick up, then they could be selling the servers for businesses to host the internal thought networks, they could also lead the ecommerce market place with software agents conducting business for you and custom ad delivery. But, there is a lot of potential and 'if's at this point and I'd expect the product & offering to morph signficantly based on user feedback in the future before they succeed. Here is wishing success to another Waterloo startup...

5 comments:

Peter Sweeney said...

Thanks for the feedback, Vishnu. You're bang-on with respect to the challenges...and there are many more! We'll unpack some more detail in the coming months on how we address those challenges, so please stay tuned. But briefly:

The simplicity/complexity of the concepts that can be represented varies greatly system by system. Extremely complex knowledge may be represented in many existing systems. The challenge, historically, has been in achieving this level of complexity while simultaneously delivering a simple, elegant experience for the producers and consumers of the knowledge. Simple on the surface, complex under the hood.

Re: black box, I couldn't agree more. This type of criticism has been tabled in the industry over and over, often in the context of search engine algorithms and transparency into their assignment of link "authority". Our perspective on that is not to displace the role of people but rather support their tasks in ways that are uniquely machine-based. Much of that is necessarily black-box (see above re: simplicity). But the scenario you describe doesn't necessarily have to fall under this category. What you described is akin to a type of agent workflow. IMO, that workflow could be provided to agents as instructions on how to conduct your task.

Vishnu Varadaraj said...

Thanks for the clarifications! I guess I was a bit carried away by the 'thought networking' buzzword. The term could indeed lead to many expectations which may not be supported immediately. It is good to know that you are aware of the issues and working on them. The interface did appear simple and elegant in the ppt. Hope there will be a public beta soon!!

Anonymous said...

Vishnu, here is one more comment in the primal Fusion blog
http://www.primalfusion.com/blog/?p=17#comments

Ecommerce developer said...

Great information about networking and its a good resource for networking peoples. thanks for your wonderful information.

titanların öfkesi said...

thank you