Who will fund Marvin the Personal Assistent?
The OSGi Alliance will require less of my services for the coming year because the R4 specification work is mostly complete and activity for R5 has not yet ramped up. The good news this opens up lots of possibilities for other activities. I already have a number of small gigs planned, but I also have these ideas that have been simmering in my head for a long time. Ideas that I would love to work on but which are too big to do on my own. I intend to write down one idea per week. This will be fun, and maybe a company will get excited and provides me with the funding to develop a project around the idea. If not, it will be nice to have the ideas outside my head and I am bound to learn something from the feedback. Enjoy!
Marvin Vision
The biggest idea in my head for the past 9 years has been Marvin. In 1997 I was hired by Ericsson Research in Kista/Stockholm. Initially I developed a 3D user interface for smart telephone switch but early on I was asked to develop a presentation about something for a yearly planning meeting. This presentation turned into a project called Marvin, a personal assistant.
The idea was that Marvin was context aware. Through electronic means it knew where you were and where you planned to be next. This in the physical sense of knowing the location you are as well as your itinerary as well as the not so physical sense of what status your projects are and what your plans are in different projects. Using your context, Marvin would create prioritized advices. Typical use cases were:
- When you get to airport, your PDA shows information how to get a taxi. When you get in the taxi, it gives you the address. Once you arrive in the hotel, it tells you your reservation number.
- When somebody tries to reach you by chat, mail, or phone, it will first contact Marvin. Marvin will decide the importance and route through the message.
- If you are in a meeting and have to catch a flight after the meeting, Marvin will warn you in time to go to the airport, taking into account the traffic situation.
- At any moment in time, you are one click a way of a list of people that you are likely to get in touch with.
Key for Marvin is that he should not be intrusive. The advice is there, but you?re free to ignore it. It learns from your communications and actions, but it does not require you to put in a lot of configuration information.
The main driver behind Marvin is active information. Search engines have been the best thing to hit the internet since SMTP but I do not believe they scale in the long run. Searching requires me to take the initiative, which is costing me time and energy. To search for all the possible things I am interested in, I just do not have enough time. I want Marvin to search for me and have the right information when I need it. Glimpses of the use of active information can be seen in RSS feeds, Google alerts, and Mylar.
Will people pay for this? I think so. Though we resist paying for online content, we do pay for our telephone because it provides an important personalized service. Marvin would be a great service for a telco operator to host. The telco can fully integrate this offering with other services. With my traveling life, I?d be willing to pay $50-$80 a month because it would save me lots of time and aggravation. Additionally, the use of bandwidth and telecom devices could generate additional revenues. I actually expect that in the end providers are willing to pay the operator to host their services. The American Express travel agency could generate additional revenue if it could integrate with the personal assistant. This is a win-win situation because their expertise would make Marvin more useful in the travel arena.
Obviously, Marvin touches, if not overlaps, with Artificial Intelligence (AI). We all know how hard this area is and how many failures are on its path. So what are the chances of success?
I think the time is ripe for. Not that there are any really major advances in AI, but I think many use cases can be implemented as services. If the architecture provides a communications and security infra-structure, we could leave many scenarios up to providers without requiring lots of overly clever code.
For Marvin, I developed a so called Info Base. This was a database highly optimized for creating semantic associations between objects. Pluggable code could create these associations and required behavior.
The Marvin ideas were too far ahead of their time in 97, the state of the art needed to do this just was not there. However, the predicted explosion of the Internet, the ubiquitous presence of connectivity, limitless storage, and advances in software technology definitely look like the time is right to develop a Marvin server as roughly sketched here.
I am therefore looking for a company that wants to fund a feasibility study for Marvin. The business model of telcos is very aligned with the Marvin model, but other companies that provide personal services are of course also applicable.
posted by Peter @ Tuesday, June 06, 2006
