Turns Out, Talking to a Flirty Robot Isn’t Actually That Useful
The failure of OpenAI’s much-hyped voice mode
When OpenAI announced its new voice mode for ChatGPT in a dramatic demo earlier this year, it felt like the movie Her had come to life.
In a helpful and somewhat flirty voice, the system coached an OpenAI employee through breathing exercises and helped him translate languages.
OpenAI has now rolled it out to many users, including me. I tried it out, and I’m not impressed.
Bad Modalities
I love ChatGPT. I’m one of those power users who turns to the bot for nearly everything: ideas for YouTube videos, recipes for chicken croquettes, feedback on obscure photography topics, Python coding advice, and much else.
I especially love ChatGPT’s multi-modal capabilities. I can type out random notes, drop in an article, or even snap a photo and upload it to the app.
For example, I was recently cleaning up my backyard for a party, and my outdoor furniture was grimy. I snapped a photo, sent it to ChatGPT, and asked for a custom cleaning solution that would tackle the accumulated dirt.
ChatGPT sent me a recipe involving vinegar and other natural ingredients. I made the cleaner, and it worked…
