OpenAtom 7: Societal Investments in AI
National Investments to Win the AI Race
The emergence of Generative AI is as fundamental as the emergence of Nuclear Fission and it announces the start of a world-transforming race that secular democracies must win. To win that race, the nation must have a purpose that is shaped by both optimism and a clear eyed assessment of risks. Developing the regulatory and oversight infrastructure, brought to life in a National AI and Cybersafety Agency (NACA), is a necessary prerequisite for guiding the nation through the coming onslaught of AI disruption.
Regulation on its own, however, will hamper rather than advance the cause of national AI leadership. It may keep us safe inside a cocoon that applies after-the-fact bandages to growing and foreseeable societal harms, but regulation will not let us win.
Winning — and win we must — requires a societal commitment to moral, technical, defense and economic AI leadership that will endure for generations rather than years.
Creating an Ethical Framework
First and foremost, AI must operate within a consistent, ethical framework that nurtures public confidence and enthusiasm. This framework must be rooted in the fundamental concept of fairness. If we are to have any chance of pursuing a national AI agenda, it must work to the benefit of the whole country. This means ruthless attention to issues of bias in applications and assurance that all schools and universities have access to AI resources. (Biden’s Executive Order addresses bias and related issues at length.)
It also means strong protections for citizens and intellectual property rather than simply rewarding megacap tech companies. I mentioned in OpenAtom 2 that data is the fissile material of AI. We must hold as a first priority the protection of data and fair compensation to people whose data is used. Allowing data to be harvested and used indiscriminately is the equivalent of setting off nuclear tests in Iowa corn fields.
The flip side is that companies must operate against a backdrop of certain and consistent legal protections and responsibilities. If a company is operating within the defined principles of the country, they must be held harmless when there are early failures. The United States is famously forgiving of bankruptcies. We understand that innovation means failure and the bankruptcy process serves to ensure that failures are not lifelong. The FAA has strict testing procedures. Companies operating within those procedures are not sued for a failed plane developments.
Investing in our People
To win in this race, we must invest in our people’s skills in the same way the nation funded nuclear development. Government demand for engineers funded generations of electrical, mechanical, computing, nuclear and aerospace engineering students. Their collective talents kept the country safe and drove uncounted innovations and breakthrough companies. AI can do this again and even more quickly.
Investing in our people cannot be solely defensive. The media, amplified by tech leaders’ breathless and self-serving proclamations of employment armageddon, is devoting lots of attention to job destruction. As a society we must prepare for this displacement, but to focus solely on that aspect is not playing to win. It will lead us to just spend money rather than pursue national purpose.
To win, we must challenge ourselves to revisit the core skills our young people need to succeed. Without the “new math” that flummoxed Baby Boomer parents, Gen X and later generations would have not been prepared to build the STEM industries that shape today’s economy. An economy shaped by AI will require new and different skills yet again. Some skills we value today will become the domain of artisans while other skills will move quickly to the forefront.
Gen X’ers who toiled through writing cursive letters but never received proper computer programming education should take note: this same problem is coming for your grandchildren.
To its credit, Biden’s order directs the creation of university AI institutes, directs NSF money into AI pilots, and funds a handful of researchers … but it is terribly sub-scale. Where is the AI version of Los Alamos labs? Where are the vehicles to lift up millions of people?
Our country is embroiled in difficult conversations about education affordability, inflated credential requirements, and the demise of the trades. This AI moment must be used to revisit how we train, fund, and credential future generations that will likely never use the “new” math nor manually write a dissertation.
Investing in the Military
In a world where the AI risks include losing control of an AI weapon, it may seem odd to suggest heavily investing in military AI. Overly constraining the military, however, would be a strategic mistake.
The military has the budget, talent and discipline to move decisively on innovation. Yes, the programs are almost always shockingly expensive and are often late. There is endless interference from politicians that hamper efficient development and budget management.
Yet a simple truth remains: the US military is in a league of its own in terms of leading-edge weaponry and information systems. They have an excellent safety record when dealing with complex and dangerous things. The country trusts the military with technologies and weapons that can literally destroy the world. By extension we trust the government with digital and cryptographic technologies that can be turned into dangerous domestic surveillance tools. But we also have well developed oversight to guide and manage these fearsome capabilities.
The other reality is that US adversaries will assuredly empower their militaries with any advantage that AI can confer. The asymmetry of AI investments makes this even more dangerous than the risks of nuclear proliferation.
The Cold War was characterized by a race for nuclear supremacy both in terms of weapons and the adjacent surveillance and defensive capabilities. The AI age will be characterized by a race for the most powerful computing, the highest functioning models, and the richest data stores. The military and national security apparatus must have an effectively unlimited budget to win this race for the capacity to conduct both offensive and defensive AI operations.
We must trust our national security apparatus to be a worthy steward of AI.
Investing in our Companies
One may ask where corporate profits fit in this pantheon. Is this just government heavy intrusion into a dynamic space?
Proposing a new national agency and muscular AI regulation is (by definition) government heavy. When backed by a national purpose to be the best, it is also a heavy push that will channel and accelerate our AI investments. The companies sitting downstream of those investments in people, capabilities, and an ethical framework will benefit immensely.
Many companies were born and thrived in the investment environments shaped by the cold war and the race to the moon. National defense provided well funded and motivated anchor tenants for breakthrough technology while economic productivity created broad demand.
Against a robust backdrop of societal AI investment, corporate profits will be strong. Profits will be driven not just by government contracts, but also from the highly productive businesses and applications that will be developed.
Societal investments take many forms but the best are rooted in ethics, shaped by realism, and supported through investment. We must embrace all of these aspects to build the generational mechanisms that will enable western democracies to win the AI race.
