AI Research Organization Files Complaint Against GPT-4

The Center for Artificial Intelligence and Digital Policy (CAIDP) describes AI product GPT’s latest iteration as “biased, deceptive, and a risk to privacy and public safety.” It filed a complaint with the US Federal Trade Commission (FTC) to this end on March 30.

The CAIDP aims to stop the development, release, and marketing of even more powerful AI systems.

Case alleges unfair and deceptive acts

According to the CAIDP, an independent NPO, GPT-4’s commercial release violates the FTC Act, which outlaws deceptive or unfair practices or acts impacting commerce. As proof, the NPO drew attention to contents in the GPT-4 System Card:

We found that the model has the potential to reinforce and reproduce specific biases and worldviews, including harmful stereotypical and demeaning associations for certain marginalized groups. AI systems will have even greater potential to reinforce entire ideologies, worldviews, truths and untruths, and to cement them or lock them in, foreclosing future contestation, reflection, and improvement.

No independent assessment preceded GPT release

In addition, OpenAI released the product for commercial and public use knowing full well that these risks existed. No one undertook an independent assessment of GPT-4 before it was released.

In light of all this, the CAIDP wants the FTC to investigate OpenAI products as well as products being developed by other AI system operators.

Latest version is 10x more intelligent than ChatGPT-3

GPT-4 is believed to be ten times more intelligent than ChatGPT-3, which was released in November. It passed some of the most difficult exams with results in the top 90th percentile and can detect smart contract issues on the Ethereum Mainnet.

Recently, a host of AI experts including Elon Musk and Steve Wozniak signed a petition to suspend development of AI systems that are even more powerful than GPT-4. CAIDP president Marc Rotenberg was also among the signers.