Current machine learning models are unsuited for risky applications like autonomous driving.
What’s at stake:
Most machine learning models developed for corporate applications never make it to production. Greater engineering discipline and standards are needed to catch failures earlier in the development process.