Can AGI become a tide that lifts all boats, not just the yachts?

I still remember that one random summer day when the YouTube algorithm queued me to this video and Ray Kurzweil's work. He discussed the S-curve law of evolution, present in both biology and technological progress. This curve seems just as relevant in AI.

The Evolution of Life and Technology

Life on Earth began with single-celled organisms that dominated for billions of years. For a long time, evolution appeared slow. But then came the Cambrian Explosion—around 541 million years ago—when complex, multicellular life burst onto the scene in a geological blink of an eye. From there, the pace quickened. Mammals emerged, then primates, followed by hominids. Homo sapiens only appeared around 300,000 years ago. Yet, in just the past few thousand years, human civilization has rocketed forward.

The story is similar in technology, particularly in computing. Kurzweil often cites Moore’s Law—the observation that the number of transistors on an integrated circuit doubles roughly every two years. This is one of many exponential trends that fuel innovations. Each of these advancements follows its own S-curve:

1. A slow beginning

2. A period of explosive improvement

3. A plateau as physical or theoretical limits are reached.

   
   

Just as one technology begins to falter, a new one emerges. This momentum is depicted in the graphs below. When the first technology reaches its limits, it signals the dawn of the next.

You can see this happening in the timeline of Netflix.

And now, let’s apply this to the concept of sustainability in fashion.

Many other innovations follow similar patterns.

Kurzweil predicts that a similar phenomenon will happen with AI, culminating in what he calls the Singularity.

Understanding the Singularity

The Singularity is the moment when machines become smarter than humans and start improving themselves autonomously, leading to an intelligence explosion.

However, if we are indeed on an S-curve now, where do we stand on it? Are we in the slow adoption phase? The fast acceleration? Or have we already plateaued?

Most would predict we're in the early acceleration stage. But history tells us it’s only clear when we zoom out to observe the larger picture.

No matter where we are in the S-curve, we are on one. Even if we've reached the tail end (for various reasons such as running out of training data and rising energy costs), Ray’s theory suggests we might be at the beginning of the next S-curve.

What is Artificial General Intelligence (AGI)?

Before we reach the realm of machines that surpass human intelligence (Artificial Superintelligence), we must first achieve Artificial General Intelligence. Asking ten AI researchers for a definition of AGI often yields fourteen different responses.

One useful definition comes from Nils J. Nilsson, a Stanford AI pioneer. He calls AGI:

“A machine that can perform any intellectual task that a human being can.”

This classic definition frames AGI as general-purpose intelligence, not confined to narrow domains.

AGI Traits

Here is a summary table that captures what AGI represents:

Trait	Description
Generality	Not task-specific—can reason across many domains
Autonomy	Operates independently with minimal human input
Learning and Adaptation	Learns from experience and improves over time
Goal-Directed Behavior	Able to pursue objectives, plan, and make decisions
Transfer Learning	Applies knowledge from one task to another
Human-Level or Beyond	Matches or exceeds human performance in cognitive tasks

This contrasts with Narrow AI, of which generative AI is a part. Narrow AI refers to systems confined to specific tasks. This is what currently dominates the AI field.

AGI is usually defined in relation to the total capabilities of an average human. But who is this average human? Is this person from America, Nigeria, or China? Is this person someone with a decade of experience or a recent graduate? This is why defining AGI is so challenging and often leads to cultural discussions rather than technical ones. The AI we have today is far smarter than some humans, but it all depends on which “human” you choose.

In the end, the specifics may not matter. We might define AGI as a phase. For some nations, AGI will arrive sooner; for others, it may take longer. In the years to come, we might look back and realize that we collectively achieved AGI in 20XX.

AGI's Future and Furthermore, What's Next?

A few months ago, a paper from AI researchers discussed the future of AI in 2027. In three short years, an AI race between the US and China could lead to significant changes. The paper ends with an interactive choice: a path toward slowdown or acceleration—favoring a slowdown. For those who enjoy doomsday scenarios, this paper might intrigue you.

However, I think it goes a bit too far. Researchers sometimes overlook the rest of the world. I have lived in the San Francisco bubble. There, people work with advanced AI technologies daily. This group often overestimates the shift's pace. The general acceptance of AI is an uphill climb. While advancements occur, they may not be as drastic as the paper suggests. (Feel free to revisit this article in 2027 to discuss the results!)

The Transition to AGI and Beyond

These days, during my keynote speeches, I joke with the audience to cherish those spreadsheets they dislike. In the future, those roles will be seen nostalgically, much like how we view tape recorders or gramophones.

There's significant evidence pointing to a decrease in entry-level positions. Many college graduates struggle to find jobs because routines that fell on entry-level roles are now assigned to AI. This creates a conundrum where it's challenging to secure a job without experience—something most college grads lack. Currently, many companies measure success not by headcounts but by revenue per employee. Thus, fewer employees lead to better-looking company performance.

For the first time in history, elite MBA graduates from prestigious institutions like Harvard are facing employment difficulties. Read more about this here..

Recently, Bill Gates mentioned three roles that will likely remain untouchable by AI destruction. The first is surprisingly, programmers. While AI can generate code and automate some programming tasks, it still lacks the ability to handle complex software development, debugging, and refining code with the same precision as a human coder.

The second role is Energy Experts. The energy sector, involving oil, nuclear, and renewable energy, is highly intricate and requires human expertise to manage infrastructure and navigate industry challenges.

The last role is for Biologists. While AI aids biological research, particularly in areas like disease diagnosis and DNA analysis, it currently lacks the creative thinking crucial for groundbreaking discoveries. Biologists remain essential for hypothesis formulation and experiment design.

As AGI becomes integrated into knowledge work, specific sectors and roles will be affected first. Tasks such as administrative support and customer service jobs will be altered. However, in finance, insurance, and analytics, AGI will excel at data-crunching, pattern recognition, and prediction. Positions like analysts, accountants, and bookkeepers are highly automatable and could radically change.

Lessons from the Industrial Revolution

To foresee our future, we might need to look back at history. Knowledge workers today are facing changes similar to those experienced by factory workers during the Industrial Revolution.

In the late 1800s, the United States advanced rapidly due to the rise of railroads, steel, and mass production. Industrialists like Rockefeller and Carnegie amassed immense wealth while ordinary workers toiled under poor conditions for low pay. Gains from new technology largely favored the owners of capital, leading to increased inequality.

We may be heading down a similar path with AGI. Power could become concentrated among those adept at utilizing AGI. AGI-driven automation benefits workers with advanced skills or education. In contrast, employees in roles requiring less education may experience job erosion or stagnant wages.

In the U.S., we could witness a pattern similar to that of the early Industrial Revolution: soaring productivity and wealth creation, but labor's share of that wealth may decline. Without intervention, social stratification might intensify, potentially resulting in unrest or political upheaval in future years.

Income Polarization

Expect continued and amplified polarization: high-wage jobs will grow and pay more, while low-wage jobs will proliferate but offer poor pay. Displaced middle-class workers may shift into lower-paying occupations. Workers in lower-wage fields may face more competition. A layoff might force a former administrative assistant into a role as a retail supervisor, increasing access to low-wage jobs.

Research shows that those in lower wage brackets are 10–14 times more likely to transition to new occupations than those in the top quintile. Certain groups, such as women and minorities, may be disproportionately affected since they often occupy automatable roles.

Regional Disparities

Some regions of the U.S. will experience more significant disruptions. Areas reliant on manufacturing or routine jobs could face severe job losses. Conversely, regions focused on tech (like Silicon Valley or Seattle) could thrive. If unaddressed, AGI may worsen the divide between prosperous metro areas and struggling ones, a pattern seen with earlier automation trends.

The Future of Work: No Immediate Mass Unemployment, More Gig Work

It's crucial to understand that AGI's impact is gradual and cumulative, not a sudden shock. We are unlikely to witness a 30% unemployment rate overnight. The labor market will likely absorb initially displaced workers through new job creation. For instance, opportunities may grow in healthcare or green energy sectors.

If well-managed, the economy might keep unemployment rates low. Employers may gain negotiating power, pushing down wages or shifting roles to part-time or contract positions. Without strong labor institutions or standards, many workers could face gig-like employment. Nonetheless, AGI might also create new job categories that we can't fully predict.

Additionally, shorter workweeks could redistribute workloads among more people, enhancing quality of life without pay reductions. However, achieving these outcomes necessitates deliberate actions to reinvest and redistribute the gains from AGI.

Possible Solutions: Can AGI Lift All Boats, Not Just Yachts?

Numerous experiments explore life after AGI. You may have heard of Universal Basic Income (UBI), a common proposal tested in countries like Canada, Wales, and Norway. These studies consistently show improvements in mental health, reduced poverty stress, and sometimes even increased full-time employment.

The Idea of a Robot Tax or “AI Automation Tax”

Bill Gates recently suggested that policymakers could introduce a tax on significant profits from labor-displacing AI. For example, a company that automates 100 jobs might pay a tax equivalent to the lost wages into an “AI Transition fund.” This revenue could support retraining efforts.

Another approach is an “AI dividend.” If AGI substantially boosts GDP, some of that wealth could be distributed to citizens, similar to Alaska’s oil dividend, but focused on AI productivity.

In the near term, adjusting the Earned Income Tax Credit or implementing wage insurance could benefit individuals displaced by AI. Countries like South Korea have considered automation taxes, a trend gaining global traction as AI develops.

Establishing a Nationwide AI Transition Workforce Fund and Retraining Initiative

Such a program would provide income support and retraining services to workers laid off due to automation. If a call center in Iowa cuts 500 jobs due to AI, these former employees might receive unemployment benefits and free tuition for in-demand skill training.

The scale of this program must be substantial. It could draw inspiration from past initiatives like the Trade Adjustment Assistance program, which helped over 5 million workers. Evidence suggests participants earned more over the following decade than those without support.

Expanding Lifelong Learning & Upskilling Programs

Traditionally, we perceive education as ending at around 22 years old. However, in the AI age, lifelong learning will become essential. A core long-term solution involves reforming our educational system to promote continuous skills training.

Governments and industries should subsidize and facilitate ongoing training. This includes making community college or technical education tuition-free for displaced workers and creating accessible certification programs for AI-related skills.

Employers should pivot to skills-based hiring rather than formal credentials, welcoming candidates from nontraditional backgrounds.

Conclusion

As we approach AGI, we may see various forms of course correction. Unemployment could rise in some roles, but other opportunities might arise quickly. While gig work may increase, we might avoid the worst-case scenario of sudden mass unemployment.

Governments will respond with various measures—from UBI to robot taxes and AI Transition Funds. Companies may also receive tax breaks for implementing AI-Affected Programs to rehire retrained workers.

Nevertheless, inequality is likely to increase. Those capitalizing on AI may see significant wealth growth, further straining the middle class. Despite these challenges, I remain hopeful. The journey is just beginning. We must all stay informed about the unfolding changes.

So, join in. If you believe someone might benefit from this article, please share it. A more informed population will lead to a more resilient society. AI can elevate all people, not just the elite.

--

Hello! I’m Sharon Gai, a keynote speaker on AI and its effects on workers and society. If you were forwarded this article, here is where it came from https://sharongai.substack.com/