Dario Amodei, CEO of Anthropic, stated on Wednesday that AI could potentially wipe out half of all entry-level white-collar positions within the next five years. A senior executive at LinkedIn mentioned last week that AI has already begun to displace new graduates from their jobs. In April, individuals voiced concerns that AI is poised to take jobs, including their own. Even experts are cautioning about the significant potential of AI to transform the economy.
What is the basis for these concerns?
The primary goal of the leading AI companies is to develop artificial general intelligence (AGI), defined as “a highly autonomous system that outperforms humans at most economically valuable work.”
This is not merely a hypothetical discussion; companies are investing over a substantial amount to progress towards AGI. Additionally, global entities are supporting these efforts to realize this goal.
They are making considerable progress. Current AI models perform comparably to humans on various tests. They demonstrate superior code generation abilities compared to most programming professionals and excel in answering science-related questions.
Consequently, AI industry leaders believe they could achieve AGI soon.
It is a widely held belief among insiders at top AI companies that widespread technological unemployment, where AI replaces human jobs, is imminent. AGI is expected to impact every sector of the labor market, starting with white-collar jobs and eventually extending to blue-collar workplaces as robots improve.
In a post-AGI world, AI is likely to perform tasks better and more affordably than humans. While developing a cutting-edge AI model is costly, replicating it is inexpensive, and these costs are rapidly decreasing.
A frequently suggested solution to potential technological unemployment is government-provided universal basic income (UBI). However, this could drastically alter societal participation since work is the primary leverage for most individuals. The modern economy relies on a straightforward exchange: labor for financial compensation, based on the worker’s time or skills.
The economy depends on workers’ skills, judgment, and spending habits. Therefore, workers have historically negotiated for better pay and shorter work weeks because of their importance to the economy.
AGI is poised to change or even eliminate this dynamic. For the first time in history, human labor might become obsolete for wealth creation, as machines can perform tasks more efficiently and cheaply. Consequently, companies may no longer need human labor to increase profits, and governments may not depend on it for tax revenue.
This could lead to what some call “resource curse for people,” where powerful entities like governments and corporations develop AGI and subsequently lose the incentive to invest in human capital.
Similar to oil-rich nations affected by the “resource curse,” governments may not need to invest in their citizens to maintain power. In the worst-case scenario, they might disregard human welfare altogether.
However, the technological path is not predetermined. Solutions can be developed to address this issue.
Many who address the significant risks of AGI, such as it becoming uncontrollable or assisting terrorists, focus on centralized and regulatory solutions, like tracking AI chips and requiring permits for training AI models. They aim to prevent malicious actors from accessing powerful AI and to avoid the accidental creation of AI that could threaten global existence.
However, AGI will not only be a tool for mass destruction but also a means of production. Centralizing the means of production is not merely a security concern but a fundamental decision regarding the distribution of power.
Instead, security threats from AI should be averted by developing defensive technologies. AI itself could help secure the code that runs critical infrastructure. Investments in biosecurity could prevent engineered pandemics. A focused effort, like Operation Warp Speed, for AI alignment could ensure AGI remains under control.
By protecting against the extreme threats posed by AGI, the technology can be broadly distributed, ensuring power remains in the hands of the people.
Human-enhancing AI should be prioritized over human-automating AI. Steve Jobs famously described computers as “bicycles for the mind,” because they enhance efficiency and speed. With AI, the goal should be to create a “motorcycle for the mind” rather than a complete replacement.
The market for technologies that preserve and expand human power will be substantial. Currently, the fastest-growing AI startups focus on augmenting human capabilities rather than automating them, such as code editors. As AI becomes increasingly powerful and autonomous, developing human-enhancing tools today could pave the way for human-owned tools in the future. AI tools could capture knowledge and transform it into personal advantages.
The labor of the masses could be replaced either by the AI and capital of a select few or by the AI and capital of everyone. Technologies should be developed that allow regular people to participate, run them on affordable hardware, and maintain control over their data, instead of relying on a few large corporations. Individuals could own businesses, using AI under their control to solve complex problems.
The role in the economy could shift from direct labor to managing AI systems, similar to how a CEO manages employees, or to guiding AI systems working for them, like a company board setting long-term strategy.
The economy could operate automatically and at superhuman speed. Even if AI performs tasks better, owning and controlling a piece of it would empower individuals, rather than depending on UBI that may not materialize.
Paraphrasing , the problem with AI capitalism arises when there are too few capitalists. If everyone owns a share of the AI future, everyone can benefit.
AGI will make effective institutions and governance more critical than ever. Democracy must be strengthened against corruption and economic pressures before AGI arrives, ensuring that regular people can succeed even if governments and large corporations no longer depend on them.
Currently, an AGI race is underway, even if most of the world is unaware. AI labs have an advantage in AI, but to automate tasks, they need to train their AIs in the skills and knowledge necessary to run the economy and then outperform the people currently providing those goods and services.
Can AI be used to empower people before AI labs train AIs to replace them? Can autonomy be maintained even as AI becomes superintelligent? Can a future be achieved where power still comes from the people?
Answering these questions is everyone’s responsibility.
