Smart Intelligence Analysis Software: What Operators Need Versus What Operators Get
February 24, 2020
DarkCyber noted “The ABCs of AI Enabled Intelligence Analysis.” The major problem with today’s intelware solutions is stated clearly:
The inability to adjust analysis tools to the operational environment is a prodigious problem.
Vendors want operators (licensees) to adapt to their environment. The idea is that the vendor’s environment is the only way to get the most out of an intelware system. What if the customer does not like this approach? Yeah.
Now the marketers, developers, and field engineers will insist that this statement is incorrect.
Here’s a passage from the write up which explains the operator’s point of view:
There are two key concepts to any data-centric system: First, analysis tools and applications should change with the data, and second, data should be easily accessible. Analysts must be able to configure the tools and algorithms of the systems to meet the realities of the battlefield, and data access should be as seamless as possible.
So what’s wrong? Here’s the explanation:
Within a data-centric context, the use of machine learning algorithms has led to breakthroughs in nearly every analysis endeavor, from fraud detection to image identification. To take advantage of these advances, intelligence analysts need systems that allow them to use computational tools and to constantly adjust, or retrain, their algorithms to a changing battlefield. Unfortunately, nearly all analysis software products in use today — including advanced systems like Palantir or Analyst Notebook — are closed systems that do not allow analysts to code custom algorithms, use the latest machine-learning algorithms, use the latest research in “explainable AI,” or even allow analysts to provide feedback to the software’s algorithms.
DarkCyber recommends taking a look at this write up.
Several observations:
- Marketers, vendors, and field engineers are busy with their own agendas. As a result, paying customers are usually ignored. Their requests are not on the road map, too difficult to make, or of no interest.
- Existing intelware solutions are purpose built to require training, support, and tradition. In one demo, the marketer could not understand that his actions were obscured by the control panel of the video conferencing system used to show off features. The person did not listen; the eager beaver was on auto pilot.
- Legacy systems like Analyst Notebook are often rarely used. The license is simply paid because, as one top dog law enforcement professional told me, “We don’t want to be without it. But no one has been to training recently. It is just here.”
These three problems are not part of the “AI baloney party.” I think these dot points underscore how deep the disconnect and how severe a problem today’s intelware helps foster.
For those who want to point out that certain tools developed in other countries are “better, faster, and cheaper.” Based on DarkCyber’s exposure to these systems, the newest tools are repeating the errors of the past 20 years.
A goldfish knows only water. The real world is different. But intelware fish don’t die. They force the customer to learn how to exist within their watery world.
Stephen E Arnold, February 24, 2020
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
