“Human-in-the-loop” is an increasingly popular way of describing the application of human judgement to AI training. And of course, for some types of AI training humans remain an essential ingredient.
"But who are the humans in the loop?" people often ask, and it's a reasonable question. Lots of machine learning projects involve proprietary or otherwise sensitive data. And some projects involve data that can only be viewed by people with particular credentials or levels of security clearance.
Here at Alegion we routinely prepare ML training data for financial institutions, healthcare organizations and government agencies. These types of organizations can have sensitive projects, with high requirements for data security. And because not every client has the same data security concerns, we typically begin conversations by describing our four-level security framework.
4 levels of security
The bare minimum: Known Workers
Many projects use open or public data that aren't subject to particular threat or security concerns. But there is a certain level of security needed on all projects. We respect that. That’s why every one of our data specialists has passed a stringent background check, and can sign client- and data-specific non-disclosure agreements and attestations. And regardless of their location, these workers are work on data over 256-bit SSL encrypted connections.
Give me more: Virtual Private Crowd
At this security level data specialists work in a secure isolated environment on Amazon's web services platform. In this environment data specialists work in a custom browser, and they are denied the ability to save data or to cut and paste it or otherwise move it from their browser.
That’s not enough: Isolated Private Crowd
Sometimes a client's project requires that data specialists work in a particular physical environment, on-site at a gated and secure location. Access to these work centers is by badge, and phones, cameras and other electronics are not permitted on the premises. We have work centers both in the US and offshore.
I can't tell you what I'm working on, but those won’t do: +I - ITAR/EAR Provisions
We prepare training data for projects that impact national security. For these clients, we can go beyond the physical environment to ensure that only US citizens, on US soil, are involved in the project.
We've got all the crowds
We’ve got this all covered -- whatever your security needs. Take a look at this white paper for more detail. And drop us a note if you’d like to talk about how we do human-in-the-loop.