Artificial Intelligence > The Forever Web App Project

The Forever Web App Project

Published: November 15, 2009

I recently came up with an idea for a little AI project. The premise is that I want to put a finite amount of time into preparing a web-based program/system where it is then left to it's own devices to see how long it can survive for and anyone associated with the project can't assist once the whole thing kicks off.

Survival is a good goal for an AI project. And if the project "survived" for 20 years, for instance, that would be pretty cool but it would actually be pretty trivial to do from what I've figured out.

See, I original thought of setting up something that could make money (through advertising or donations or something) and even pay it's own bills (hosting, dns, etc) this would be done by hacking together some rough API's and keeping fingers crossed that shit wouldn't break over time. I followed this idea through a bit further and thought the system could detect when it's in trouble financially or technically and be able to alert any random visitors to the site that it happened to be in trouble. Then it would need to allow for itself to be helped by a stranger - and this is where it gets tricky.

After running with the idea for a while I realise I could set up a super basic program on Google's AppSpot (AppEngine) and it wouldn't need to worry about paying bills and stupid API's. The big thing it would have to watch out for is updates to the software behind the AppEngine because Google says they do their updates whenever they feel like it and if that happens to break your code, you have to fix it.

I doubt any update is going to break a "hello, world!" app and if my app was up there saying "hello, world!" for 20 years that doesn't seem like the most amazing AI project getting around. So I'm wondering if there is a way to jazz it up. Any ideas?

So, going back to the idea of the program looking out for itself (and it's hosting, blah, blah, blah). It would be fairly easy to get it to do self diagnostics all the time to ensure everything was running smoothly. If things are looking a bit ill I was wondering how to develop the trust system. Say, a stranger visits the site and sees the alert that survival app is sick. How can the app know it can trust the person and then open itself up in someway (like passing on login details or some better and secure option) so it can be healed by the nice stranger? Is there a way to measure the trust of a stranger and make sure they have a reason to remain trustful while helping?

I feel this is the key to the survival of the app - communicating and asking for help. Even the "hello, world!" app could benefit from this approach. I'm thinking this concept is quite similar to the cute robot set loose in Washington Square Park, NYC.

Any thoughts or suggestions would be totally appreciated.

Any Comments?


» Metric Love Rocks the Orpheum

Sudoku & Artificial Intelligence

Published: March 16, 2009

Using AI on Sudoku could be considered overkill. Either way, any algorithm written to solve a Sudoku puzzle could be considered intelligent by understanding what it accomplishes.

Internet Lifeforms

Published: March 5, 2009

The Internet may give us strange clues as to what the definition of life really is. Seeing Alex Dragulescu's visualisations you really begin to wonder.