‘FTC investigating whether ChatGPT is harmful to consumers’

‘FTC investigating whether ChatGPT is harmful to consumers’

ChatGPT experienced explosive growth when it arrived on the scene late last year. However, it can count on widespread attention from government regulators. The U.S. Federal Trade Commission (FTC) is demanding that OpenAI explain how it curbs dangers regarding data use and image damage to individuals.

The Washington Post has reported this development. This week, the FTC reportedly sent a 20-page document to OpenAI’s San Francisco headquarters, filled with questions about the alleged risks.

As is now known worldwide, chatbots are capable of much. Thanks to generative AI, applications such as ChatGPT and Google Bard are capable of reasonably human communication. Still, there are many drawbacks to them, causing concerns among parties such as the EU and the U.S. government. Consider the vast data sets Big Tech collects to feed its AI model, which raises copyright issues.

Different nature

However, the issues the FTC is concerned about are somewhat different than we have predominantly seen so far. Usually, the pain point is privacy and copyright issues, whereas here the issue is also the so-called hallucinations that such chatbots suffer from. Because a large language model is not on the hunt for truth-telling, it can go entirely wrong on facts. LLMs predict the next word based on many parameters, using a user’s prompt as the ignition flame. In doing so, it takes no account of fact-checking from within itself.

The Washington Post further reports that political legislation to regulate AI will be months away. It is a difficult issue to fathom, even when OpenAI CEO Sam Altman stopped by the U.S. Congress for a hearing in May.

Attempts to push generative AI into a factually correct format are in full swing. In that regard, chatbots are rather an incomplete example of what the underlying innovation is capable of. Potentially, an LLM can dig through a huge data corpus faster and more effectively to arrive at insights than a human alone – with sufficient training, development and the right input.

Also read: Google Bard finally finds its way to Europe