Return to site




 

Human in the Loop is Mission Critical

The Example: NASA Mathematician Katherine Johnson...

About Katherine, NASA astronaut John Glenn said, "Get the girl to check the numbers." With those words, Katherine Johnson, a trailblazer and brilliant mathematician changed the course of history. She was an African American, a 44-year-old "girl" and a "human computer" assigned to NASA's Flight Research Division at Langley Research Center in Hampton, Virginia.

First, Johnson manually calculated the flight trajectory for Mercury astronaut Alan Shepard, who became the first American to fly into space on May 5, 1961. Meanwhile, the agency had been adopting electronic computers, then a room-size IBM 7090. It was put to the test by John Glenn prior to the launch of Mercury capsule Friendship 7 On Feb 2, 1962. A Marine officer and engineering test pilot, Glenn orders his team to...

"...get the girl to check the numbers. If she says the numbers are good, I am ready to go." 

Johnson verified the machine calculated flight trajectory as correct. Glenn accepted the result and became first American to orbit the Earth. The mission went on to become one of NASA's great achievements of the 20th Century. Until the the book

 "Hidden Figures" was published in 2016, Johnson lived in relative obscurity as a highly respected NASA mathematician, retiring in 1988.

Katherine Johnson, Trailblazer and NASA Mathematician

Human in the Loop is Required for Pioneering Systems

Pioneering complex systems require a "human in the loop" to check results before they are used. Many use cases for AI are still at this stage. AI is inherently weak because the data used for building its data science models is inherently limited. Model performance is only as good as its training data and that data is always limited. Until AI has "inference," the ability to learn without historical data, it will always be severely limited in capability when operating outside its training parameters (also known as edge cases). Humans and AI must operate in tandem to overcome these limitations. AI's great power can enable humans and their machines to scale their capability. But for AI to succeed it must follow the approach used by NASA in the Mercury space program. Here, NASA relied on the skills of used Katherine Johnson to manually double check mission critical results of their IBM 7090 mainframe.

We Use Human in the Loop for Successful AI in Decision Making

We use human inference to enhance decision making with AI. This overcomes the problem of limited data that does not fully describe the problem, that result in a high incidence of false positives and false negatives. It also prevents alerts on anomalies that not related to the use case, eg, transients in operation, startups and shut downs, etc.

The DecisionIQ platform and data science technology for improved decision making using AI. 

All Posts
×

Almost done…

We just sent you an email. Please click the link in the email to confirm your subscription!

OK