OpenAI might must halt ChatGPT releases following FTC criticism

A public problem might put a short lived cease to the deployment of ChatGPT and related AI techniques. The nonprofit analysis group Heart for AI and Digital Coverage (CAIDP) has filed a criticism with the Federal Commerce Fee (FTC) alleging that OpenAI is violating the FTC Act by its releases of enormous language AI fashions like GPT-4. That mannequin is “biased, misleading” and threatens each privateness and public security, CAIDP claims. Likewise, it supposedly fails to fulfill Fee pointers calling for AI to be clear, truthful and straightforward to elucidate.
The Heart desires the FTC to research OpenAI and droop future releases of enormous language fashions till they meet the company’s pointers. The researchers need OpenAI to require unbiased evaluations of GPT services earlier than they launch. CAIDP additionally hopes the FTC will create an incident reporting system and formal requirements for AI turbines.
We have requested OpenAI for remark. The FTC has declined to remark. CAIDP president Marc Rotenberg was amongst those that signed an open letter demanding that OpenAI and different AI researchers pause work for six months to provide time for ethics discussions. OpenAI founder Elon Musk additionally signed the letter.
Critics of ChatGPT, Google Bard and related fashions have warned of problematic output, together with inaccurate statements, hate speech and bias. Customers can also’t repeat outcomes, CAIDP says. The Heart factors out that OpenAI itself warns AI can “reinforce” concepts whether or not or not they’re true. Whereas upgrades like GPT-4 are extra dependable, there is a concern folks might depend on the AI with out double-checking its content material.
There isn’t any assure the FTC will act on the criticism. If it does set necessities, although, the transfer would have an effect on growth throughout the AI trade. Firms must watch for assessments, and would possibly face extra repercussions if their fashions fail to fulfill the Fee’s requirements. Whereas this would possibly enhance accountability, it might additionally gradual the at the moment fast tempo of AI growth.