The Bottleneck in AI Integration
AI is competent enough to replace many tasks of skilled professionals.
But it spikes the coordination effort at the initial integration phase.
Organizational structures in corporate America are notoriously ineffective at coordination relative to lean startups.
This is why AI integration has been inefficient and also why Apollo is skeptical about AI’s potential.
The lack of a payoff in the AI economy is not due to a lack of competence; it is due to a lack of coordination.
Anthropic’s Understanding of the Bottleneck
Anthropic believes that the lack of coordination is in no small part due to workers resisting integration.
(Workers don’t want to contribute to the driving down of their wages—at best—or their displacement—at worst.)
Anthropic’s proposed solution: provide financial incentives for people to integrate and ensure work remains for them.
But if company shares are that financial incentive, then that’s hardly attractive at non-frontier companies with stagnate valuation.
Also two years of mass layoffs has created an atmosphere where the promise of “you’ll still have work” falls on deaf ears.
Anthropic doesn’t have a solution to these problems they gesture at.
A Macro Solution: Economic Absorption and Redistribution
Both OpenAI and Anthropic continue to release domain-specific tools—Anthropic released personal finance capabilities today—that add more competence for corporate American to integrate, spiking the coordination problem even more.
This also means that both OpenAI and Anthropic continue to absorb economic opportunities from startups in the frontier AI orbit—especially when trust is low and AI proliferates many unknown startups.
This leaves three options:
OpenAI and Anthropic continue to absorb economic opportunities, and they reinvest their gains into corporate America to help solve the integration problem.
Anthropic’s creation of a public fund gestures at this potential.
The risk is that the reinvestment is left to the whim of the voluntary giving of these companies, causing capital to land far away from local needs.
OpenAI and Anthropic are prevented from absorbing economic opportunities beyond mere AI models.
OpenAI and Anthropic distribute their economic opportunities evenly across America’s regions, effectively extracting the UK model into America’s private sector.
A Micro Solution: Accelerate Human Coordinators
Macro solutions are complicated and politically-entangled, but there is a clear solution within the immediate control of corporate America: accelerate human coordinators.
While AI can replace tasks of skilled professionals, they are not yet effective at solving the coordination problem at scale, according to a recent paper from Google DeepMind.
As they note, we are in a “minimal AGI” phase, even if American workers are lagging to take notice of this.
In this phase, workers that can accelerate AI integration through effective coordination and trust-building will be increasingly valuable.
Software engineers, mathematicians, and other workers must at risk to be displaced would be very effective in these roles. It’d be both just and effective to grant a fast-track into these roles.
These coordination roles could also be the place to move other displaced workers.
The paper also suggests that displacement may not be as severe as we’d predict. When we distinguish minimal AGI from further stages of AGI, consolidating on technology, pinning a worker to an agent, and moving workloads from the worker to an agent—while staying blind to the temptation of integrating and coordinating agent swarms—can maximally preserve existing structures. It can also make organizational activity, and ultimately cost and value, much more easy to track—a paramount problem according to the latest Wharton AI Adoption report.
These challenges are the challenges in implementing this micro-solution:
It requires shifting AI integration single C-level individuals to a “calvary” of coordinators, and bureaucratic habits might stubbornly resist.
It requires fluidity between roles that may clash into HR-maintained gates.
Most importantly, it could drive down wages for non-coordinating workers:
The Google DeepMind paper writes, “workers whose skills are matched by AI will see their wages decline toward the cost of compute, while specialists and human coordinators can demand even higher wages than before.”
In this event, the problem has only been softened from fears of complete displacement to reduced wages.
However, if corporate America has already committed to AI adoption, then perhaps the coordination efforts will save enough cost to offset the temptation to reduce wages.
Between a Macro and Micro Solution
At some point, America cannot have its cake and eat it to.
We cannot point fingers to critique macro solutions yet not do our apart at the micro level.
But mico solutions would no doubt benefit from the macro solutions.
No matter which layer hits a breakthrough first, the value of human workers needs to be paramount to build out for the common good.
The goal is clear: integrate AI to replace skilled tasks while maintaining skilled workers.
The old adage is still true for a new problem: If there is a will, there is a way.