The GPT Store isn’t doing AI any favours

An app store that was meant to get more developers and users interested in AI might be doing just the opposite.

Driving the news: OpenAI’s GPT Store is full of spam bots that violate the company’s policies. Some impersonate people or infringe on the copyright of media franchises like Star Wars and Avatar: The Last Airbender. Others claim to thwart plagiarism detectors and ChatGPT’s own content safeguards.

• OpenAI isn’t liable for its users’ copyright infringement under U.S. laws, as long as it takes down offending content when requested.

Catch-up: The GPT Store was launched to let developers share and sell custom GPT-4 chatbots. It was thought it could do for AI what Apple’s App Store did for mobile development, attracting developers with compelling ideas while creating a new revenue stream for OpenAI.

• Developers, however, have been frustrated by a lack of users, sparse analytics, and a slow rollout of monetization options.

Why it's happening: OpenAI says bots don’t get onto the GPT Store until they pass a mix of human and automated reviews. If bots breaking the rules are slipping through, it means OpenAI needs to beef up human content moderation, revamp its screening tools, or review its policies (or a combination of all three).

Why it matters: This doesn’t look great for OpenAI, which already faces regular criticism over copyright infringement and other dishonest uses of AI. But it also doesn’t look good for AI as a whole, which seems to constantly be grappling with a reputation crisis.

• Election-related deepfakes are still spreading, despite controls major AI operators have created to curb misinformation.
   
• Click farms are embracing AI-created images to ramp up social media spam. The latest trend is pictures of Jesus made out of things like shrimp and plastic bottles.
   
• AI-generated articles are making their way into scientific journals, and we don’t just mean the penis rat.

Bottom line: OpenAI can’t be blamed for every AI-generated problem on the internet. But if it’s giving a platform to such blatantly spammy chatbots, there are certainly some things it could be improving.