Mass Hacking Event in Las Vegas to Test AI Technology Limits

OpenAI, Google and Microsoft are joining hands with the US government in Las Vegas for a mass hacking event to test the limits of their AI technology.

The initiative is aimed at finding how chatbots can be manipulated to cause harm, if they share private information to other users and why they assume a doctor is a man and a nurse is a woman.

The mass hacking event is planned for this summer’s DEF CON hacker convention in Las Vegas and is expected to draw several thousand people. The event has been planned in accordance with the White House’s Blueprint for an AI Bill of Rights, which is a set of principles to limit the impacts of algorithmic bias, give users control over their data and ensure that automated systems are used safely and transparently.

Currently, there is a community of users who are trying their best to trick chatbots and highlight their flaws. Many of them are hobbyists who show off humorous or disturbing outputs on social media until they get banned for violating the product’s terms of service.

The mass hacking event will not just find flaws, but will also figure out ways to fix them. Companies that have agreed to provide their models for testing include OpenAI, Google, chipmaker Nvidia, and startups Anthropic, Hugging Face, and Stability AI. The event will hopefully be the start of a deeper commitment from AI developers to measure and evaluate the safety of the systems they are building.

The event will tackle the large language models that have attracted a surge of public interest and commercial investment since the release of ChatGPT late last year. This new technology is built on what’s known as large language models that emulate the cultural biases they’ve learned from being trained upon huge troves of what people have written online.

For example, a search for a female doctor in a commercial chatbot would have a higher chance of returning results that describe them as a nurse, rather than a doctor. Similarly, if asked to draw a picture of a doctor, it would mostly portray a male doctor.

As these foundation models become more and more widespread, it’s really critical to ensure their safety. There have been instances where users were able to get chatbots to tell them how to make a bomb or fabricate information and present it as fact.

Companies are providing their models for testing and the event will be the first to tackle large language models. The goal is not only to find flaws but to figure out ways to fix them. This initiative can be considered as a direct pipeline to give feedback to companies. The mass hacking event will provide valuable insights to the AI developers, which will eventually help in building safer AI systems.