My AI writing assistant kept suggesting the same three words so I finally checked the training data

Was working on a blog post about local SEO last week and noticed my AI tool kept pushing "leverage" "optimize" and "synergy" nonstop. Got curious and dug into the documentation. Turns out the model was trained heavily on corporate marketing emails from 2018 2020. No wonder it sounded like a bad LinkedIn post. Switched to a lighter model trained on actual conversations and Reddit threads and the difference is night and day. Has anyone else found their AI tool has weird language habits from its training data?

3 comments

3 Comments

wells.olivia22d agoMost Upvoted

Trained heavily on corporate marketing emails" - that explains everything. I used to defend those writing assistants. Thought they sounded professional. Then I got a draft back that said "let's circle back and leverage our core synergies going forward." Realized I sounded like a robot from a sales meeting. Changed my mind completely after that. Now I try models trained on blog comments or forum posts. Way more natural. Never going back to that corporate speak nonsense.

elliot_patel22d ago

@jamie770 try telling that to my clients who stopped understanding half my emails

jamie77022d ago

I mean is "circle back and leverage core synergies" really that bad? Feels like people overreact to this stuff.