OpenAI restricts face recognition function due to concerns about GPT-4 face recognition performance
It is clear that OpenAI is taking measures to limit the face recognition function due to concerns that the ability to identify a specific individual provided in the image recognition function of GPT-4 may violate the privacy law. became.
A ChatGPT That Recognizes Faces? OpenAI Worries World Isn't Ready. - The New York Times
Report: OpenAI holding back GPT-4 image features on fears of privacy issues | Ars Technica
https://arstechnica.com/information-technology/2023/07/report-openai-holding-back-gpt-4-image-features-on-fears-of-privacy-issues/
It was revealed that GPT-4, announced in March 2023, was being developed as a ``multimodal AI'' that accepts not only text but also image input.
Following this announcement, OpenAI is working with a startup called Be My Eyes to develop an app to explain images to visually impaired users and help them better interpret their surroundings. is working on
However, according to the story of a blind woman living in New Zealand, The New York Times reported that at some point, an app developed by OpenAI stopped providing facial information and 'privacy protection.' The face is hidden because of this,' he said. The New York Times wrote, 'Women expressed disappointment.'
Sandini Agarwal, a policy researcher at OpenAI who was interviewed by the New York Times on this point, acknowledged that the reason for the above is that OpenAI has reduced the facial recognition ability of GPT-4.
According to Agarwal, OpenAI's system can identify public figures whose pages are created on Wikipedia, but Illinois states that ``use of biometric information requires explicit consent from citizens'' and In regions like Europe, these features may violate privacy laws.
In addition, the app may misinterpret information such as 'gender' and 'emotion', and displaying such information may lead to inappropriate or harmful results. OpenAI is aiming to eliminate safety concerns before the image analysis function of GPT-4 becomes widely used.
The challenge of image recognition facing generative AI models like GPT-4 is something that not only OpenAI but also many companies have to overcome. Ars Technica, a technology media, cited the case that Google's AI 'Bard' and Microsoft's Bing AI based on GPT-4 solved the 'CAPTCHA' of the authentication system, saying, 'Computer vision equipped with AI may be on everyone's devices sooner or later, but companies need to solve complex problems before they can be widely released with minimal ethical repercussions. is also clear,' he said.
Related Posts:
in Software, Posted by log1p_kr