Biden admin’s AI Safety Institute not ‘sufficient’ to deal with risks, must check user ‘procedures’: expert


Experts tell Fox News Digital that the Biden administration’s plan to establish an artificial intelligence (AI) safety commission may prove “necessary” but not “sufficient” to address potential risks for the burgeoning technology.

“The odds are [the algorithm] is not where the majority of the risk lies,” said Phil Siegel, founder of the Center for Advanced Preparedness and Threat Response Simulation (CAPTRS). “It is more likely the risk lies in the users either using it for bad or just plain misusing it.”

President Biden on Monday signed an executive order that the White House said included the “most sweeping actions ever taken to protect Americans from the potential risks of AI systems” – the requirement for companies to notify the government when training new models and sharing results of “read-team safety tests.”

“These measures will ensure AI systems are safe, secure and trustworthy before companies take them public,” the White House said of the executive order.

EXPERTS DETAIL HOW AMERICA CAN WIN THE RACE AGAINST CHINA FOR MILITARY TECH SUPREMACY

The administration also announced the establishment of the AI Safety Institute – under the oversight of the National Institute of Standards and Technology – which will “set the rigorous standards for extensive red-team testing to ensure safety before public release.”

Secretary of Commerce Gina Raimondo said the Biden administration would use its AI Safety Institute to evaluate known and emerging risks of “frontier” AI models and that the private sector “must step up.” (AP Photo / Andrew Harnik / File)

Speaking at the Bletchley Park summit in the United Kingdom, U.S. Secretary of Commerce Gina Raimondo said Wednesday that the Biden administration would use its new AI Safety Institute to evaluate known and emerging risks of “frontier” AI models and that the private sector “must step up.”

Siegel compared the White House approach to that of an airline checking a plan for “safety” but not checking maintenance procedures, the pilots’ training or crews.

WHAT IS ARTIFICIAL INTELLIGENCE (AI)?

“All are necessary,” he said. “Similarly, a safety board can’t just check the algorithms. It needs to check procedures for the users.”

FTC chair hands Biden a pen

Experts tell Fox News Digital that the Biden administration’s plan to establish an artificial intelligence safety commission may prove “necessary” but not “sufficient” to address potential risks for the burgeoning technology. (Alex Wong / Getty Images / File)

“We can make tech providers help,” he continued. “Like we have the banks provide KYC (know your customer) procedures to prevent money laundering, we can require the tech providers provide KYC for user application safety,” Siegel added.

The Center for Advanced Preparedness and Threat Response Simulation tackles these kinds of problems regularly, looking at decision-making and intuition among users in public health, engineering, public policy and other industries and training them in games to improve those skills. As such, user behavior remains a central concern – much like it will with AI.

NOT OUR NATION’S JOB TO KEEP ALLIES ON ‘CUTTING EDGE’ OF AI DEVELOPMENT, FORMER CIA CHIEF SAYS

Many critics of AI since earlier this year have highlighted the myriad pitfalls the technology presents, from deepfake technology disrupting elections and generating child abuse material to using AI-generated algorithms to break through even the most complex digital security systems and access sensitive information.

Bletchley Park, birthplace of computing machines

The Bletchley Park Estate in the U.K. is shown on the second day of the AI Safety Summit on Nov. 2, 2023. (Chris J. Ratcliffe / Bloomberg via Getty Images)

Christopher Alexander, chief analytics officer of Pioneer Development Group, acknowledged that while it is a good idea to force companies to share their information rather than hide it away – in what one expert previously described to Fox News Digital as a “black box” of content – the current system appears to have “no transparent appeals process.”

CLICK HERE TO GET THE FOX NEWS APP 

Alexander told Fox News Digital that he also worried that “political agendas could bias the safety approval process” because the agency, established by executive order, puts its management at the behest of the sitting president.

Some critics have already raised political bias concerns, such as with China requiring any new AI technology to conform to the ruling party’s socialist values

Fox News Digital’s Greg Norman and Reuters contributed to this report.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *