The Fork in the Road
In Building AI, Which Road Will We Take?
This past March, Elon Musk and more than 1100 people, including Apple cofounder Steve Wozniak, Stability AI CEO Emad Mostaque, researchers at Alphabet's AI lab DeepMind, and notable AI professors, signed an open letter calling for a six-month pause on training AI systems more powerful than GPT-4. The letter, which was issued by the non-profit Future of Life Institute, stated powerful AI systems should only be developed "once we are confident that their effects will be positive and their risks will be manageable." The letter urged AI companies to create and implement a set of shared safety protocols for AI development, which would be overseen by independent experts.
It is clear now that the authors of the open letter have lost the battle as the development of AI proceeds full speed ahead. However, they did succeed in expanding the conversation to include the pitfalls as well as the promises of what is likely to be the most significant communications technology since the printing press. While the caution advocated by these professionals was certainly warranted, we must be careful not to limit the conversation to the usual suspects. That’s because a solution overseen by a group of technology experts may not be the best remedy, especially given the recent performance of experts in handling—or rather mishandling—the recent Covid-19 pandemic. In that case, the world’s supposed best experts from prestigious organizations such as the WHO, NIH, CDC, and FDA got just about everything wrong. They produced a novel vaccine using an unproven technology that failed as its basic job of preventing the acquisition and transmission of the virus. They discarded centuries of experience in managing pandemics by taking the unprecedented step of mandating lockdowns of whole societies and ignoring the well-known benefits of natural immunity. Furthermore, as Elon Musk knows well, public health experts engaged in an unprecedented censorship campaign to silence and demonize any scientific or medical professional who thought differently and challenged the single-minded policies of the public health elite.
Unfortunately, the mounting evidence of the damage of lockdowns, unseen levels of vaccine injuries and deaths, and unprecedented increases in all-cause mortality suggests that these experts were nowhere near as intelligent as either they or we thought. As the evidence continues to pile up, especially now that Musk’s purchase of Twitter has dealt a death blow to the censorship campaign, we are recognizing that there is a very real possibility that, over the next decade, we are likely to learn the cure will take far more people in the prime of their lives than a virus that largely victimized people at the end of their lives. Perhaps the most important lesson we will learn from the mismanagement of the pandemic is that experts are flawed human beings just like the rest of us, and it is an illusion to think trusting authority is a winning strategy for managing a complex crisis. As we approach the development of AI, we need to make sure we do not lose sight of this lesson because the consequences of mismanaging AI will be many times greater than the bungled Covid experience.
Doing Something Extraordinary
If insanity is doing the same thing over and over again and expecting a different result, perhaps it’s time to think differently and do something extraordinary. Instead of trusting authorities and crafting another top-down expert-driven solution, perhaps we should take an original and more innovative approach to developing AI. Perhaps we should follow the lead of an anonymous person or persons who, using the pseudonym Satoshi Nakamoto, published a white paper on October 31, 2008 describing a digital cryptocurrency titled, “Bitcoin: A Peer-to-Peer Electronic Cash System.” In devising this radically different currency system, Nakamoto created the first blockchain database, a game-changing innovation that transforms systems design from centralized hierarchies to distributed networks by using the synergy of all participants rather than expertise of an elite few as the premise for the exercise of power in the system.
The use of a distributed ledger rather than a centralized database is the defining characteristic of this innovative systems architecture. In Bitcoin, there isn’t a single ledger for recording transactions. Instead, everyone in the system has a ledger and all the ledgers are updated simultaneously as blocks of transactions are added to the blockchain. If someone were to hack into an individual ledger and find a way to change its data, the overall system would recognize the anomaly and disregard the infected ledger. This makes hacking extremely difficult, if not impossible, because no single malicious individual agent or coercive government agency can exercise unilateral control within the system. The inability for single agents to exercise control over other agents is a game-changing feature that radically transforms the way the dynamics of the different forms of individual agency work in this unusual systems architecture.
Individual Agency and Social Organization
There are four different forms of agency: autonomy, coercion, authority, and facilitation. Autonomous agency is the capacity for individuals to think and act independently, and to be responsible for the control of their own lives. Coercive agency is the capacity for individuals to control the thinking and the actions of other individuals through the use of force or intimidation to gain compliance. Authoritative agency is the ascribed capacity to guide the thinking of others and to provide processes for directing or permitting the actions of other individuals. Lastly, facilitative agency is the capacity to provide processes for individuals to think and act independently or for groups of individuals to reach agreement on their collective thinking or action.
In addition to the four forms of individual agency, there are four types of social organization: anomie, autocracy, bureaucracy, and self-governance. While all forms of individual agency may be present in each of the organizational types, the basic operating system of an organization is a product of the combination of the two dominant forms of individual agency that define how the system works.
For example, anomie is a paradoxical combination of autonomy and coercion. This combination leads to a prevailing state of chaos because there are no processes, standards or principles to guide the functional coordination of social behavior. Without any fixed norms for social interaction and people doing whatever they want, the chaotic environment usually evolves in one of two ways, depending upon whether autonomy or coercion is the more dominant form of individual agency. When autonomy is more prevalent, the conditions are ripe for creativity. Free from established ways of doing things, individuals can image and explore new ideas. With no one in charge, no single individual can veto their ideas. This form of anomie is typical for start-ups. These are usually small teams where everybody knows each other and everyone contributes. In the absence of prescribed norms, group behavior is generally constrained by a soft version of coercion in the form of situational peer pressure. As long as groups remain small, the creative possibilities of anomie tempered by soft peer pressure is capable of fostering extraordinary innovation.
On the other hand, when coercion is more dominant, anomie is likely to take the shape of a mob as a brutal few leverage their autonomy to maneuver a group to engage in senseless behavior. This subtle use of coercive agency foments a ruthless hierarchy where self-appointed rogue leaders manipulate and galvanize the autonomy of the group members into a powerful form of groupthink that that is often psychologically or physically violent. Because there are no functioning rules or norms, unlike the other forms of organization, there is no basic operating system for anomie. This explains why anomie is rarely a permanent state. Successful start-ups will eventually morph into either bureaucracies or self-governed enterprises, while mobs either dissipate or transform into autocracies if they can effectively seize coercive power.
Autocracy is the combination of authority and coercion. The organizational design principle for totalitarian regimes is obey authority. Autocracies assure order and eliminate the chaos prevalent in anomie by extinguishing autonomy. In autocracies, individuals have few, if any, individual rights. They are not free to speak their minds and can make very few decisions for themselves. The autocracy’s operating system is the surveillance state where the elite few exercise vast and invasive coercive control of its subjects by closely monitoring all their activities. This is the approach that the Chinese Communist Party is attempting to perfect with its use of social credit scores to micromanage behavior and reinforce maximum obedience among its citizens. This is also the approach used by many governmental and corporate entities in enforcing mandates to either take insufficiently tested vaccines with unknown long-term effects or risk being terminated.
Bureaucracy is the combination of authority and autonomy. Up to this point, it has been the most prevalent form of social organization in both corporations and government agencies in the Western world. The organizational design principle for bureaucracies is trust authority, and as long as authority can be trusted, these centralized structures, while not optimal, are often workable. The operating system is the regulatory state. Accordingly, maximum autonomy is ascribed to an elite few who have the legitimate authority to issue regulations and direct or permit the actions of others in their organization. Thus, for example, in a corporation, bosses provide directions to subordinates, who are free to exercise autonomy within the bounds of those directions. Similarly, a licensing board has the authority to certify professionals such as doctors, accountants, or pilots, and only those with licenses can exercise autonomous agency within their fields of specialty, as long as they comply with generally accepted professional norms.
Self-governance is the combination of autonomy and facilitation. The organizational design principle for these decentralized structures is nobody is smarter than everybody. Accordingly, unlike autocracies and bureaucracies that amplify the intelligence of the elite few at the top, self-governing organizations use facilitative processes to aggregate the collective intelligence of everyone and leverage the wisdom of the crowd. This is the form of social organization used by Bitcoin. The coin holders have a high degree of autonomy, and the distributed ledger system acts as a facilitative mechanism that allows individuals to conduct transactions without the need for elite intermediaries.
James Surowiecki, in his seminal book The Wisdom of Crowds, identifies the four conditions that are necessary to access the self-governed organization’s collective intelligence. If any of the four are missing, crowds are prone to chaos, but with all four conditions, groups are highly likely to be far more intelligent than even the smartest individual in the group. The four conditions are diversity of opinion, independent thinking, local knowledge, and an aggregation mechanism.
The prerequisite for both diversity of opinion and independent thinking is free speech. In a self-governing structure no one’s point of view—no matter how eccentric—is censored or cancelled. Everyone’s voice matters. And while we may strongly disagree with another’s opinion, the ability to freely express that opinion is a foundational principle. The condition of local knowledge assures that people who have practical information are not ignored by a self-proclaimed intelligentsia who believe, often mistakenly, that they are smarter than everyone else. Lastly, the primary aggregation mechanisms in self-governing structures are tools or processes for turning private judgments into collective decisions, such as voting, algorithms, or, in the case of Bitcoin, proof of work consensus mechanisms. The integration of these four conditions provides the circumstances for the effective functioning of a peer-to-peer network, which is the basic operating system for self-governance.
The Fork in the Road
Throughout the twentieth century, the two dominant forms of government have been communism and democracy. Communist regimes are clearly autocracies with long and well-documented histories of exercising coercive power to assure citizen compliance with the directives of authorities. In these totalitarian states, individual rights are nonexistent or greatly diminished. On the other hand, although most democracies profess to be self-governing structures, the intention is more aspiration than practice. While democratic citizens may have the right to elect their leaders, these leaders are typically temporary stewards for administrative agencies where the decisions that impact people the most are made by unelected and oftentimes unaccountable bureaucrats. In democracies most people experience their governments as regulatory states rather than self-governing enterprises. While the leaders of democracies may believe they are facilitating self-governance, the reality is that they are leading bureaucracies. And, unfortunately, these bureaucracies can easily become autocracies. There are numerous examples of administrative states morphing bureaucracies into autocracies through the use of emergency powers, first with Covid and, on the horizon, with climate change. As we contemplate the future of AI, the practical prevalence of autocracies in both forms of government is a troubling development.
The essential challenge, which is reflected in the concerns of the 1100 people who unsuccessfully called for a pause in the development of AI, is to choose which road to take when we approach the technological fork in the road. Will we take the usual road and build AI as a top-down hierarchical structure that has the potential to be the most pervasive authoritarian social system the world has ever known, controlling every aspect of our day-to-day lives? Or will we take the road less travelled and follow the example of Satoshi Nakamoto by building a distributed peer-to-peer AI system that rapidly aggregates the collective intelligence of all participants to produce knowledge and solve problems that none of us—most especially the elites—could ever accomplish on our own? If we choose the former road, we are likely to be ensnared in the pitfalls of a technological nightmare. If we choose the latter, we are likely to experience the promise of an incredible leap in human intelligence. We need to choose wisely and we need to choose soon because the fork in the road is right in front of us.
And read also:
The Future of Management in a Radically Different World
Follow me on X/Twitter. Check out my website or some of my other work here.