We’ve been talking a lot about newsworthy incidents involving AI. While not exactly headline making, there’s been constant action by advocates and academics to push AI practitioners to build ethical and safe systems using AI technology. After years of AI professionals, mostly women, sounding the alarm on the potential dangers of AI, and witnessing a rise in deepfakes President Biden issued an Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
The EO comes as the world struggles to deal with the emotional, political, and economic fallout wrought by social media companies. AI can exacerbate many of the same ills social media has propelled. Generative AIs seemingly mimic human behaviors and can lead folks to emotionally connect with AI products. These products can harm folks by harming their self-esteem, stealing personal details, and affecting behavior through misinformation.
What’s In the Executive Order?
The executive order isn’t a short read, and it’s beefy.
The executive order calls out 8 guiding principles for implementing artificial intelligence and machine learning systems. It states any policies coming out of the order take into account the views and inputs from academics, allies, industry, society-at-large, and labor unions.
Given the enormous reach of AI, taking input from all sides is necessary, let’s just hope it doesn’t turn into flawed policy.
AI Should Be Safe and Secure
Biden called out current issues with safety and security in the first guiding principle. We know there are serious issues with deepfakes and adversarial threats against AI. The principle goes beyond simple models, however. Any policies must implement safeguards affecting biotechnology, cybersecurity, and critical infrastructure. All are areas impacting national security.
Set the US as the AI Leader
The second principle establishes an order to position the United States as the leader in AI and ML systems. The order calls for increased funding for the education of artificial intelligence and collaboration with industry and academic leaders. There’s also a small snippet about intellectual property rights and protections for both model engineers and the holders of data those engineers need.
Support American Workers
The third principle orders policymakers and implementers to provide protection for American workers. AI as a job killing machine may be over hyped, but many workers will be impacted and displaced by advances in this new technology. The order provides support for job training and education to help workers transition into new domains. The order also provides support for labor conditions to not worsen or impose new safety risks due to the use of AI technology.
Protect Civil Rights
Keen listeners and readers know we’ve reported on the mistakes AI have made when identifying people. AI and other ML techniques are frequently being used for evaluating prisoner recidivism and credit scoring. Incorrect training, biased model engineering, and errors can create harmful impacts through our society. The executive order requires policymakers to identify techniques and ways of building models that protect our civil rights.
Protect Consumers
The fifth principle focuses on protecting consumers. The order pushes policymakers to ensure current consumer protection laws aren’t circumvented by AI/ML policies. This principle also makes note about using AI to increase quality, while lowering prices for the consumer. These are also good things.
Protecting Privacy
The sixth principle is all about privacy. We’re fully aware laws and policies have dropped the ball on consumer privacy regarding social networks. The EO seeks to reenforce existing laws in privacy protections and records retention. It also pays lip service to AI being a vector for leaking private details on citizens. Industry is taking a look into this by employing new techniques like differential privacy to minimize data exfiltration.
Watching the Watchmen
Principle seven outlines the need for the federal government to use AI and ML technologies appropriately. This principle states that the Fed will employ people all across society from multiple industries, experts in social domains, and from marginalized communities to manage risks associated with the government’s own use of AI. We’ve seen law enforcement abuse the use of AI and other technologies in the guise of protection. This principle seeks to change that. Hopefully states follow suit.
America Leading the Way
The eighth principle isn’t specifically attuned to AI, but is more of a visionary principle. The last principle calls for the US to be the global leader in spearheading innovation, and also in an ethical and equitable way. The Fed will promote responsible AI in conjunction with other nations and global corporations.
The government’s previous policy of having a laissez faire attitude towards social media companies has not worked. Algorithms feeding our basest behaviors and misinformation has promulgated our worst social ills. Here’s hoping we get ahead of the harms and dangers AI can create by being more proactive.