How do you validate AI? Is validate even the right word?

AI is already moving into GxP, sometimes through purpose-built AI applications and sometimes as new AI capabilities within software organizations already use. But how do we know when an AI application is trustworthy enough for a regulated use? And is traditional computer system validation even the right starting point? 

Val talks with Andy O’Connor of Phanero about why “AI” is not one single technology and why the first question should perhaps not be “How do we validate it?” but “What are we asking it to do?” 

SHOW NOTES: