At Inky (https://inky.com) we're using a combination of computer vision, anomaly detection, and domain-specific hacks to identify zero-day phishing emails "from first principles" (as I like to say). And it works! But the pushback from the security establishment is impressive. I like to say that there are two widely-held but false beliefs about phishing: 1) phishing is solved; 2) phishing is unsolvable.
The truth is that we can already see clearly that within 3-5 years machines will be good enough at identifying phishing emails that attackers will move to another vector... but you'd never know it listening to "Security Thought Leaders."