US challenges hackers to break ChatGPT and other AI models
The Joe Biden adminstration is challenging thousands of hackers to see if they are able to break into artificially intelligent (AI) systems such as ChatGPT. This is part of a series of new actions designed to promote “responsible” artificial intelligence (AI) innovation that protects Americans’ rights and safety.
The White House announced last week that leading AI developers in the US, including Google, Microsoft, OpenAI, Nvidia and Anthropic, would in August participate in a public test of their AI systems at Defcon 31. This is one of the world’s largest cybersecurity conferences held annually in Las Vegas.
At the same time, US Vice President Kamala Harris met the bosses of tech firms OpenAI, Microsoft, Alphabet and Anthropic to discuss critical AI risks and opportunities, particularly those pertaining to cybersecurity, biosecurity, and safety.
“This independent exercise will provide critical information to researchers and the public about the impacts of these models, and will enable AI companies and developers to take steps to fix issues found in those models,” the White House wrote in its briefing.
“Testing of AI models independent of government or the companies that have developed them is an important component in their effective evaluation.”
OpenAI’s large language model (LLM) service ChatGPT gained popularity last year as an educational tool that can speed up writing tasks and even teach people how to code. However, there are many concerns that the technology is biased and makes answers up, as it looks at internet content from the late 1990s until only 2021.
While President Biden’s administration recognises that “AI is one of the most powerful technologies of our time”, it also wants to make sure that any possible risks are mitigated.
The US government is therefore placing a “fundamental responsibility” on tech firms to make sure that their products are safe before they are deployed.
In February, the US president signed an executive order directing federal agencies to ensure that they “root out bias” in the design of any systems used. The order was to ensure that citizens were protected from computer algorithmic discrimination.
This came following multiple examples showing that the use of AI in courts, academia and commerce could lead computers to making discriminatory decisions. These were due to historical data fed into them that showed a clear bias against certain classes of society.
But the announcement also shows clear recognition from the US government that AI companies do offer great potential value.
Alongside the announcements about risks, the White House also outlined plans to create seven new national AI research institutes, backed by $140 million (£111m) in government funding.
The idea is to advance breakthroughs using AI in areas including climate, agriculture, energy, public health, education, and cyber security.
“The momentum of AI is enormous right now and is past the point where it can be banned. The smart approach is to learn as much as possible about AI technologies so that they can be governed,” analyst firm Gartner’s global head of research Chris Howard told The Standard.
“Education leads to better regulation. Regulation leads to better standards. Standards make innovations practical and useful.”
He says that the US government’s approach to dealing with AI is a result of a prescedent set by Microsoft in 2018, when the tech giant made the unusual move of requesting that lawmakers come up with legislations to govern facial recognition.
“My experience with these tech companies is that they welcome government engagement during the early stages of implementation,” Howard added.
“An open dialogue is better than finding something down the road and you haven’t built that engagement. That’s when bad things happen. Understanding a technology like AI from multiple angles will mitigate knee-jerk reactions.”