Chanel’s CEO tested ChatGPT and saw a flaw 🤦
OpenAI ChatGPT is clearly not perfect, they're no where near what they envision as advanced AI, although their getting good at it everyday.
There was more of an experiment done on ChatGPT and it proved a limitation with the model. If you're a fan of fashion, you might probably be familiar with the brand called Chanel, they seek fragrances, bags and all sorts of luxury items.
The CEO of Chanel, Leena Nair recently went to Microsoft and requested for the AI prompt team to use ChatGPT to generate her an image of the Leadership team of Chanel. This was to test how accurate Chatgpt was. Obviously Google would give exactly so.
But with Chatgpt, she got an image of a lineup of men in suits. There was no women, no diversity in race, it was just a male dominated scene.
This really surprised her and the main reason is that Chanel’s actual workforce is 76% women. Right there is your proof of how inaccurate Chatgpt is. I'm wonderin of I should still trust it for image generation because I like generating AI images a lot.
It's also a proof of how outdated Chatgpt is in terms of the data it has been trained with. To bring such results means it assumes the outdated stereotypes of men usually dominating a big company and it also proofs that it doesn't have accurate information ll the time. Besides it says so in the User interface that Chatgpt will not be accurate all the time.
This is also making me realize that AI still sees the world through biases baked in by human history and data.
If you've been using chatgpt, Claude and other platforms a lot for creating emails and basically general questions you'd notice that it's not just image problems, the AI language models often describe men as experts and women as delightful.
I'm not going to just claim Chatgpt is discriminating and you can't blame the tool because it's still being trained. But I would be lying if I said I wasn't disappointed in this. The way GPT 4 was hyped before they released it, I expected more from it
Nair just like myself isn’t against AI at all but she hopes they integrate it thoughtfully and we should at least get an AI Model that respects both ethics and accuracy.