In an op-ed published in Breaking Defense, CSET Space Force Fellow, Michael O’Connor, delves into the increasing concerns about AI safety and the need for rigorous safety testing before AI systems are deployed.
Drawing parallels from the history of military flight testing, O’Connor emphasizes three crucial lessons that can be applied to AI safety:
- Establish Design Standards Considering the Human Operator
- Test Systems in Realistic Tasks, Not Just Ideal Conditions
- Systematically Expand the “Envelope” of Safe Performance
Read the full op-ed in Breaking Defense.