Artificial intelligence may end up being more impactful than the Industrial Revolution. The private sector already has invested vast sums into developing and deploying AI. To accelerate AI innovation and adoption in the United States, the Trump administration recently announced a sweeping “AI Action Plan.”
AI already is being used in fighting disease, doing sophisticated math, transporting people and products, supporting regulatory compliance, and designing physical structures. Last year, two Nobel Prizes had ties to AI—the Nobel Prize in physics “for foundational discoveries and inventions that enable machine learning with artificial neural networks” and the Nobel Prize in chemistry “for computational protein design” and “protein structure prediction.” In announcing the 2024 chemistry Prize, the Nobel Committee proclaimed, “Life could not exist without proteins. That we can now predict protein structures and design our own proteins confers the greatest benefit to humankind.”
An AI-enabled future is inspiring. AI’s considerable capacity to enhance a person’s daily life includes augmenting human skills, raising a person’s productivity, and performing tasks, creating opportunities for people to spend time on more pressing priorities. At a macro level, AI should spur economic growth just as other technological breakthroughs have for centuries. Even more so, imagine if AI-human collaboration identifies a cure for cancer, leads to new food sources that redress hunger, and allows the building of buildings that better withstand earthquakes, tornadoes, hurricanes, and fires—none of which is far-fetched given the pace of AI research and AI’s expanding frontier.
We appreciate that as remarkable as AI’s benefits are, AI has stirred concern for many, if not fear. Advanced AI systems are rapidly getting better at what they do. This roots some individuals’ primary worry that humans won’t be able to keep up with AI and that AI eventually will assume control. Attention also has centered on, among other things, the misuse of AI, deepfakes, hallucinations, AI’s effect on jobs, and safeguarding values that certain AI applications might affront.
Trust is paramount
How we achieve the tremendous benefits and opportunities AI offers while mitigating and managing associated risks will continue as a focus, as a matter of both policy and practice. Trust in AI is needed for AI to flourish over time, delivering gains for humanity. If guardrails and oversight are too lax, risks may turn into actual harms that we aren’t willing to accept as a society. On the other hand, AI guardrails and oversight that is too heavy-handed and restrictive may come at the expense of AI breakthroughs that spur solutions to vexing problems, lead to astounding discoveries, and raise standards of living.
With AI innovation and adoption surging ahead, the real-world value of AI is clear. A key unlock to realizing AI’s full promise will be to use AI itself to help mitigate and manage concerns with AI—something the private sector can advance.
People modulate their behavior for different situations, self-correct when held to account, and internalize virtues that inform their decisions. Advanced AI may be able to do the same, staying within defined parameters and constraints to meet society’s expectations. As AI advances, it may be possible to architect and train AI systems not only to self-identify if the system is deviating from accepted performance, but also to self-improve to comport with a society’s goals and values. Technological efforts to accomplish this are under way.
Even if realized, this type of AI alignment would not entirely address core concerns.
AI that supervises AI
Another layer needs to emerge: AI that supervises AI. People will need to collaborate with AI agents to monitor that other AI systems do what they are supposed to do. Advanced AI agents can be developed and deployed to identify if other AI agents or deployments of AI act aberrantly or nefariously, do not behave in accordance with performance benchmarks, thwart objectives they were to promote, or pose other harms. For example, with human oversight, a company’s compliance AI agents can help ensure that the AI systems the company uses in its business adhere to legal and regulatory requirements—which is impossible for humans alone to do because of the speed and scale at which a business’s advanced AI systems can operate.
The volume and extent of AI uses is enormous and expanding. This presents a difficulty. For example, when using AI to market products, provide interactive customer support, draft agreements, or create other content, businesses need to confirm that what’s generated by their AI systems is compliant. But human bandwidth, on its own, will not be enough to review and assess all the uses for compliance. This capacity gap will widen as AI advances and is adopted more widely and acts more autonomously. Companies can’t fall short of their compliance obligations; nor can they unduly hesitate, let alone pass, on using AI agents in the marketplace when their competitors are actively doing so to get a strategic and operational edge. A solution is for companies to deploy AI agents to bolster compliance, augmenting human effort and judgment. AI agents’ unprecedented ability to evaluate in seconds or minutes whether AI outputs and behaviors comply with extensive legal and regulatory requirements enables the compliant adoption and implementation of AI at scale.
AI should be embraced, even as we prepare for the change it brings. Risks need to be accounted for. But they should not dampen the commitment to fostering AI development and adoption. That commitment includes investment in basic and applied AI research. And it includes appropriately calibrating risk management and governance that builds trust without curtailing ingenuity and innovation.
Leveraging AI to help ensure that other AI systems operate beneficially and reliably—in accordance with their intended purpose—would contribute to AI’s long-term promise.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
