Navigation software shows how different programmers can produce different decisions from similar data. The same problem applies to air combat: an AI trained on known aircraft may struggle to interpret an unfamiliar adversary or a genuinely new capability. Human judgment is fallible, but people can reconsider earlier decisions and infer possibilities beyond a known database. The imagined encounter with an unknown sixth-generation fighter shows why the first moments of surprise may matter more than an AI fleet's ability to learn from losses afterward.
Notable Point · 11:26-17:55
Training data, blind spots, and novel threats
Machine learning can optimize within its experience, while its training choices and unfamiliar situations can create dangerous blind spots.