We don't have a lot of time, but if we go back many years to when we were doing statistics, normally we were designing experiments to make sure that our samples were representative so that at the end we would get statistically significant results.
Now we're in a world where there is just a lot of data, and you take whatever you have, and it gives what it gives.
The issue of designing systems based on the representativeness of data is a key issue. Very often, when we say that systems are biased, it's just that the initial data samples are not equal or are not representative in an equal way. This is the challenge generally. It's not the technology per se; it's the data that has been provided to the system.
Dr. LaPlante mentioned all the issues with AI. It points to something like our AI system becoming a critical system that should be regulated. It's like when you design cars or airplanes; you have to demonstrate all these issues of reliability, reproducibility and all these elements. A lot of these questions point to this, in fact.
AI is still the new generation—