Why AI needs HI to see the whole picture
In a recent Solera Fleet Solutions Fireside Chat, Solera fleet safety experts Jeff Griswold and Sean Ritchie explored the very real benefits that AI can bring to video safety programs and the indispensable role that HI — human intelligence — must continue to play. Jeff is Solera’s Vice President of Product Management, and Sean is Vice President of New Business and Solutions Engineering.
AI is not new to video safety
Jeff and Sean’s comments have been edited and condensed for this article. For anyone interested in improving fleet safety, we recommend viewing Fleet Safety: The Human + AI Synergy in its entirety.
“AI has been part of video safety for quite some time,” Jeff explained. “It’s used in ADAS (advanced driver-assist systems) to warn of forward collisions, lane departures, speeding, and so forth.” Jeff also mentioned that DMS or driver monitoring systems use driver-facing cameras and AI to alert drivers to drowsiness, cell phone use, and other risky habits.
Why AI needs well-trained human intelligence
Artificial intelligence is an extremely valuable addition to fleet safety that can supplement human observation, knowledge, and context but cannot simply replace it. It can see what’s in the video but can’t necessarily say why a driver took an action.
“We use human intelligence to identify over 80 different situations that we call observations. Human intelligence can determine if another vehicle cut the driver off or if there was a near or actual collision,” Jeff said. Looking within the cab, human observation can also reveal if a driver is reading paperwork while driving, using a cell phone, smoking, etc.
The point is safety, not cool technology
When Sean asks customers why they’re interested in video safety, the answer is rarely about technology. “Typically, they want to improve fleet safety as quickly as possible, or they want to do it better than they currently are,” he said. The challenge? “Time and focus.”
According to Sean, “Roughly 80% of the risk within a fleet is represented by only about 20% of the drivers. So if a fleet can focus on improving that 20%, they mitigate most of the risk. If they are leveraging an AI-only solution, a few things become problematic.”
“Fleets can’t afford to waste time looking at videos that don’t have risk, and that’s what they’ll do with an AI-only solution,”
he said. Then, there are false positives, “an extraordinarily dangerous waste of time,” he added. “The value that a human Intelligence Center brings to AI is that it gives you a true path to be able to prioritize your time.”
How Solera’s Human Intelligence Center works
Every new Solera review analyst spends six to eight weeks in a classroom being trained on the 80+ observations before being assigned to a team, where seasoned analysts will oversee them. As videos come into the center the analyst”
- Watches the full video completely.
- Looks at the sensor input during the event.
- Annotates the video with the observed risk(s).
- Gives video access to the customer for the appropriate driver training.
The analyst’s work goes through multiple layers of quality review. Customers also do not need to wait for the review process to be completed before they see the video themselves. As Jeff said, “Once the criteria to generate the video are met and the data is recorded, it’s immediately available to the customer.”
To learn more about Solera’s fleet safety solutions, visit SmartDrive.
