Slashdot Mirror
FBI Conducts Feasibility Study on Project Sentinel
leave-no-trace writes "CNN reports that "FBI officials hope to award a contract by the year's end for a complex new software program (dubbed Sentinel) to replace a failed project that was canceled this year at a cost of more than $100 million to taxpayers." The system is supposed to include search capabilities, protocols for processing and handling FBI reports, security issues and a new system for records management. FBI Director Robert Mueller told lawmakers he is unable yet to place a price tag on the Sentinel project."
Hmmm...I don't think federal executive positions pay as much as you think they do. http://www.opm.gov/oca/05tables/html/ex.asp
If you don't know where you are going, you will wind up somewhere else.
I think SAIC's failure to execute is in small part due to the underpinning technology, in large part due to an FBI leadership that was not on the same page, but mostly due to the fact that the management of this project was mishandled.
SAIC is certainly not blameless, but I think this is a pretty good summary of what happened. The biggest problem was the FBI trying to add major new requirements to an existing project as a response to 9/11. Agreed, SAIC should have said no, but what defense contractor does that when talking to the government?
how do you define all of the costs associated with something that has never been done before and has extremely difficult engineering problems that have to be solved at various interim steps in your process. Add to that the lack of basic science supporting your engineering decisions.
You don't define those costs. Those costs should already have been accounted for in your R&D budget. The R&D is there to ensure that you do actually know how to do something before you try and do it. If you are midway through a project and still doing R&D, then somebody has fucked up somewhere.
For the hardware setup (scale) and general search solution, Google is very good. However, it is not for every problem.
Google does not have near the contextual capabilities of some (custom-fitted) search engines. At some point, you need automation and a level of reliability. You can't have a person looking at everything. And repeated searching, which we take for granted, is often necessary on the same dataset to garner sufficient results. Who says when we have found the right information?
Google does not provide complex taxonomy or a feedback loop mechanism (which can be very complicated - often patented or proprietary).
In the original PageRank thesis, it was made clear that context was entirely up to the user. When dealing with records (i.e., highly redundant data that must be cross-referenced extensively), Google falls flat.
Let me greatly over-simplify. Consider, "Joe Smith civilian" and "Joe Smith terrorist". Google will not distinguish the two Smith's. It will only distinguish the phrase in relation to the index. So - even if we have a link between Smith the terrorist and smith the civilian, we can still mix them up (unless we mark everything explicity). We need context (not just words in the same document, sentence, etc.), and as our search pattern hones in on matches (repeated, refined searching), we need better classification or we go in circles.
Really, the cost for this isn't as much as you might think it is for the work involved (or compared to other government projects).
I work for a police station and run people through FBI checks periodically. Don't get out your tin-foil hats though, most FBI checks are for employee background checks, criminal records needing to be checked directly through the FBI don't occur very frequently. But for you people out there who think what they have now is sufficient, you're wrong.
Currently the only system in place is a terminal where you have to know the cryptic commands to make it work (think unix CLI x10). Many law enforcement agencies are in the process of making the switch the "new" ncurses-like interface where knowing the cryptic commands isn't as necessary. In fact, I believe their funding would be well spent on a new system because it would probably save the goverment in general that much each -year- in training costs savings.
I know it's not a popular belief, but not everything the government does is bad.