It’s essentially the joke I put in this blog post, only spelled out for why it’s so silly.
There is an irony here.
Of all the things that text-based LLM “AI” could really kick ass at it should be email. I have a massive corpus of email, decades worth, that a model could be trained on. Hot should I respond to an email? Offer me a response that is what I am probably going to say based on all that prior data. Do I tend not to respond to certain types of emails… well that’s interesting too! Help me be myself better, don’t just help me elaborate.
Another twist here is that AI/ML is already being used at a large scale by email providers, and for much longer than any of this stuff has been a buzzword. Spam detection! There is already tons of AI deployed to be monitoring and learning from email. It seems like an obvious pivot to tap into some of that power to help me do email rather than just save me from bad email.