ChatGPT Can ‘Flirt’

Zoe Kleinman, BBC:

It [GPT-4o] is faster than earlier models and has been programmed to sound chatty and sometimes even flirtatious in its responses to prompts.

The new version can read and discuss images, translate languages, and identify emotions from visual expressions. There is also memory so it can recall previous prompts.

It can be interrupted and it has an easier conversational rhythm – there was no delay between asking it a question and receiving an answer.

I wrote earlier about how impressed I was with OpenAI’s live demos today. They made the company look confident in its product, and it made me believe nothing fishy was going on. I hope I am not eating those words later.1

But the character of this new ChatGPT voice unsettled me a little. It adjusts its tone depending on how a user speaks to it, and it seems possible to tell it to take on different characters. But it, like virtual assistants before, still presents as having a femme persona by default. Even though I know it is just a robot, it felt uncomfortable watching demos where it giggled, “got too excited”, and said it was going to “blush”. I can see circumstances where this will make conversations more human — in translation, or for people with disabilities. But I can also see how this can be dehumanizing toward people who are already objectified in reality.

  1. Maybe I will a little bit, though. The ostensible “questions from the audience” bit at the end relied on prompts from two Twitter users. The first tweet I could not find; the second was from a user who joined Twitter this month, and two of their three total tweets are directed at OpenAI despite not following the company. ↥︎