Author: David George

Translated by: Felix, PANews

Editor's Note: Currently, the AI "doomsday theory" seems to have become mainstream opinion, with fears of "AI taking jobs" and "unemployment" spreading globally. Various stakeholders are also strategizing on the disruptive innovations that AI is about to bring. However, a16z General Partner David George writes that the "doomsday theory" is nonsense, lacking evidence and imagination, and does not understand humanity. Below is the full content.

The "permanent underclass" argument put forth by AI alarmists is not convincing. It is nothing new, just a repackaged "fixed amount of work fallacy."

The "fixed amount of work fallacy" claims that the total amount of work that needs to be done in the world is fixed. It assumes a zero-sum game between existing workers and anyone or anything that could do the same work (be it other workers, machines, or current AI). If the total amount of useful work that needs to be done is fixed, if AI does more work, then humans will inevitably do less.

The problem with this premise is that it contradicts all of our understanding of humans, markets, and economics. Human needs and desires are far from fixed. Keynes predicted nearly a century ago that automation would result in a 15-hour work week, but it turns out his prediction was incorrect. While his assertion about automation causing a "surplus of labor" was right, we did not simply enjoy the fruits of this; rather, we found new and different productive activities to fill our time.

Of course, AI will certainly eliminate some jobs and compress certain positions (and there is evidence that this may already be happening). The labor market landscape will change, just as it does with every transformative technology. However, the claim that AI will lead to permanent unemployment across the entire economy is bad marketing hype, poor economics, and shows ignorance of history. On the contrary, improving productivity should increase the demand for labor, as labor becomes more valuable.

Here are our reasons.

“Humans are done for?” Don’t joke around

We agree with the doomsayers that cognitive costs are plummeting. AI is becoming increasingly proficient in areas that were until recently considered the exclusive domain of the human brain.

Doomsayers believe: "If AI can replace our thinking, then humanity's 'moat' will disappear, and our ultimate value will be zero." Humanity is done for. Clearly, we have finished all the thinking we need or want and now AI will take on more and more cognitive load, leading to the gradual obsolescence of humans.

However, the fact is: precedent (and intuition) suggests that when the cost of a powerful input decreases, the economy does not stagnate. Costs decrease, quality improves, speed increases, new products become feasible, and demand expands outward. Jevons' Paradox proves true again. When fossil fuels first made energy cheap and abundant, we did not merely make whalers and lumberjacks unemployed; we also invented plastic.

Contrary to the doomsayers' view, we have every reason to expect that AI will have a similar impact. As AI takes on an increasing cognitive load, humans can free up their hands to explore grander new fields than ever before.

Historically, technological changes have always expanded the economic pie.

Each "dominant economic sector" is replaced by a larger successor sector... which, in turn, further expands the economic scale.

Today's technology scales far beyond finance, railroads, or industry, yet its proportion share of the economy or the entire market remains small. Increases in productivity are far from a zero-sum game; rather, they create strong positive-sum forces. Delegating so much work to machines ultimately leads to a larger, more diverse, and complex economy and labor market.

Doomsayers want you to ignore the history of innovation and focus solely on the sharp decline in cognitive costs, viewing it as the whole truth. They see task replacement and then cease to think.

“We will increase cognitive output tenfold, but we won’t do more thinking, we’ll just pat our bellies and go to lunch early, and others will do the same.” This claim reflects not only a severe lack of imagination but also a failure to observe basic facts. Doomsayers call this "realism," but that simply cannot happen.

The failure of Ludditism

(PANews note: Ludditism refers to a social movement initiated by the working class in early 19th century England against the industrial revolution, destroying industrial machines to protest against deteriorating working conditions and unemployment)

Let's take a look at what happens when a significant leap in productivity sweeps through the entire economy.

Agriculture

In the early 20th century, before the widespread mechanization of agriculture, about one-third of the U.S. workforce was employed in agriculture. By 2017, this proportion had dropped to about 2%.

If automation led to permanent unemployment, then tractors should have completely destroyed the labor market. However, that is not the case; agricultural output nearly doubled, supporting significant population growth, and these workers did not just remain permanently unemployed but surged into previously unimaginable industries, factories, stores, offices, hospitals, laboratories, and ultimately, service and software industries.

So, indeed, you could say that technology disrupted the career prospects of ordinary farm workers, but at the same time, it released a surplus of global labor (and resources), spawning an entirely new economic system.

Electrification

The history of electricity development is similar.

Electrification was not merely a switch from one energy source to another. It replaced drive shafts and belts with independent motors, forcing factories to reorganize around entirely new workflows, and birthing new categories of consumer and industrial products.

This is precisely what we would expect in different phases of technological revolutions, as recorded by Carlota Perez in “Technological Revolutions and Financial Capital”: massive upfront investments and financial benefits, a sharp decline in the cost of durables, and subsequent generations of prosperity for durable goods manufacturers.

Electricity’s productive advantages did not materialize overnight. In the early 20th century, only 5% of American factories used electrically powered machines, and less than 10% of homes had electricity.

By 1930, electricity supplied nearly 80% of manufacturing power, and labor productivity doubled in the following decades.

The increase in productivity did not diminish the demand for labor; rather, it generated more manufacturing, more sales personnel, more credit, and more commercial activities, not to mention the ripple effects brought by labor-saving devices like washing machines and cars. These devices enabled more people to engage in previously unattainable high-value work.

As car prices fell, car production and employment numbers skyrocketed.

This is the true role of general-purpose technologies: they reorganize the economy and expand the boundaries of useful work.

We see this happening time and again. Did VisiCalc and Excel end the careers of bookkeepers? Absolutely not. Enhanced computational techniques led to a surge in the number of bookkeepers and birthed an entire financial planning and analysis (FP&A) industry.

We lost about 1 million "bookkeepers," but gained about 1.5 million "financial analysts."

New jobs in the service industry

Of course, job replacement does not always drive employment growth in related economic sectors. Sometimes, productivity gains translate into new jobs in completely unrelated industries.

But what if AI means some people become extremely wealthy while others get left far behind?

At least, those super-rich will have to spend their money somewhere, just as they have done before, starting new service industries from scratch:

Significant productivity gains and the resulting wealth creation spur entirely new job fields, and without growth in income and an increase in the labor supply, these fields might never have emerged (although many were technically viable long before the 1990s). Whatever one thinks about the service industries catering to the wealthy, the end result benefits everyone, as increased demand leads to a significant rise in median wages (creating more "affluent" people).

Stripe's in-house economist Ernie Tedeschi provides a comprehensive case that showcases how technology has disrupted, transformed, and reshaped the travel agent profession.

Did technology reduce the demand for travel agents? The answer is yes.

Today, the number of employees in travel agencies is about half of what it was around 2000, and this is almost certainly due to technological advances.

So does this mean technology has killed jobs? The answer is no, as travel agency employees did not face permanent unemployment because of it. They found work in other sectors of the economy, and overall employment rates in the economy now are essentially static compared to 2000 (adjusted for population aging).

Meanwhile, for those who remain in today's tech-enabled travel agency industry, productivity enhancements mean higher wages than ever before:

“In the peak year of 2000, the average weekly salary of travel agents was 87% of the overall average weekly salary. By 2025, this ratio will reach 99%, which means that during this period, travel agents’ wages have grown faster than in any other private sector.”

Thus, even though technology has indeed impacted jobs in travel agencies, overall, the employment rate of the working-age population remains on par with before, while the remaining travel agency personnel are better off than ever.

Enhance > Replace (and the yet-to-emerge jobs)

The last point is very important and again illustrates that the doomsayers only tell a small part of the story.

For certain jobs, AI poses a threat to survival. Indeed. But for others, AI is a multiplier: it makes these jobs more valuable. For every job at risk of AI replacement, there are other jobs that stand to benefit:

Goldman Sachs estimates the “AI replacement” effect is far less significant than the “AI enhancement” effect.

Notably, management teams seem to be more focused on enhancement rather than replacement:

To date, during earnings call conferences, mentions of "AI as an enhancement" outnumber mentions of "AI as a replacement" by about 8 times.

Although Goldman Sachs did not include software engineers in their “enhanced” talent list, they may well be the best example of AI-enhanced talent.

AI is a multiplier for coding. Not only have the number of git pushes surged (the same goes for the creation of new applications and businesses), but demand for software engineers also seems to be on the rise:

Since the beginning of 2025, software development positions (both in terms of numbers and as a percentage of overall employment) have been continuously growing.

Is this related to AI? Frankly, it might be too early to draw conclusions, but AI is undoubtedly improving work efficiency in software engineering, not to mention that AI has become a focus for executives in every company.

Given that everyone is striving to explore how to integrate AI into their businesses, it is not surprising that companies are engaging in large-scale hiring, which will undoubtedly enhance the value of certain employees rather than diminish it.

The proliferation of AI seems to be driving wage growth above average levels (especially in system design fields).

Currently, this growth may still be relatively limited, but we are still in the early stages. As expertise expands, opportunities will increase accordingly. In any case, this is not the data the doomsayers want you to see.

Meanwhile, according to Lenny Rachitsky (founder of Lenny's Newsletter, a platform for communication within the tech community), the number of project manager job vacancies continues to rise (which had previously dropped significantly due to interest rate fluctuations), currently exceeding any time since 2022:

The growth in hiring for software engineers and product managers strongly supports the validity of the "fixed amount of work fallacy." If AI had completely replaced human thinking capabilities, you might think “the number of engineers needed for product managers would decrease” or you might say “the number of product managers needed for engineers would decrease,” but that is not the case. We see demand for both types of talent rebounding because the key is that people's work efficiency has increased.

That is why the doomsayers' claims essentially reflect a lack of imagination. They focus only on jobs that will be replaced by automation, neglecting the fields where demand is about to create entirely new jobs that we have not even envisioned:

Most of the jobs created since 1940 did not exist even in 1940. By 2000, it was easy to imagine travel agents becoming unemployed, but envisioning a mid-market tech service industry built around "cloud migration" was much more difficult, as the widespread adoption of cloud computing was still more than a decade away.

What is the current situation?

So far, the main focus has been on theory and precedents, as both support the optimists:

That's right. Every time productivity improves, it brings demand growth or the reallocation of surplus resources to other areas of the economy. This means more job opportunities, many of which will see a significant increase in value, even entirely new occupations that have never been heard of. If this time is different, then those doomsayers must provide stronger arguments instead of merely talking empty rhetoric.

"Job replacement" is not the end of civilization (in fact, quite the opposite), and this assertion makes sense. It is in human nature not to be content with the status quo. Once we complete one job, we seek another.

However, aside from theories and precedents, what do the actual data indicate regarding AI and employment? Although we are still in the early stages (good or bad), the existing data do not support the views of the doomsayers. If any change exists, it is "no significant change," but emerging data point in the opposite direction: the job opportunities created by AI are more than the jobs taken away.

First, starting from some academic research. This is not meant to be a comprehensive literature review, just a few recent paper examples:

• “AI, Productivity, and Labor: Evidence from Corporate Executives” (NBER Working Paper 34984): "In summary, these results indicate that while the adoption of AI has not yet led to significant changes in total employment, it has begun to reshape the distribution of tasks and occupations within firms. In particular, conventional clerical and administrative tasks seem more readily replaceable, while analytical, technical, and managerial tasks are more frequently described as being augmented by AI."
• “Corporate Data on AI” (Atlanta Federal Reserve Bank Working Paper 2026-3): "In four surveys, on average over 90% of companies estimated that AI has had no impact in the past three years."

• “Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Employee Tasks” (Census Bureau Economic Research Center, Working Paper CES 26-25): "Employment changes driven by AI remain limited, with only about 5% of firms using AI reporting an impact on employee numbers: the proportions of increased (weighted by firms at 2.3%, weighted by employment at 3.7%) and decreased (weighted by firms at 2.0%, weighted by employment at 2.4%) are almost equal."

• “Tracking the Impact of AI on the Labor Market” (Yale Budget Lab, April 16, 2026): "Despite widespread anxiety about the effects of AI on today's labor market, our data suggest that this is largely speculative. The depiction of AI's impact on the labor market from our data largely reflects stability rather than significant economic disruption."

The conclusion emphasized repeatedly in the latest research is "overall there is no change, but evidence suggests that there has been a reallocation of jobs and tasks." In some cases, the net impact of AI implementation on hiring can even be positive.

But there is a significant exception to the "no change" assertion. Researchers from Stanford University, the Dallas Federal Reserve Bank, and the U.S. Census Bureau have found (to varying degrees) that "entry-level jobs with high AI exposure are becoming increasingly difficult to come by." However, before anyone concludes that “AI is killing entry-level jobs,” it is worth noting that these researchers also found that entry-level jobs have increased in cases where AI is playing an auxiliary role (and also increased in jobs where AI has had no effect).

However, even temporarily assuming AI is "killing" certain entry-level positions (rather than being affected by broader cyclical hiring trends and “ageing populations”), the overall macro picture is very clearly showing that AI's total impact on employment is essentially zero.

This may be the most concise summary of AI's impact on employment:

“There remains no statistically significant relationship between AI and unemployment rates or employment growth.”

Perhaps there is a certain preference for AI-enhanced positions while there is also a push for AI-replacement positions:

For “AI enhanced” industries, the growth in hiring appears to be stronger (with lower unemployment rates), while for industries at higher risk of "AI replacement," the situation is quite the opposite.

In other words, the overall situation is neutral, but not static: some jobs disappear, some jobs emerge, some jobs depreciate, and others appreciate. At this rate, job postings for developers will surpass pre-pandemic levels in less than two years. AI may have single-handedly saved San Francisco's job market.

This is our initial premise: AI will undoubtedly eliminate or compress some positions (and companies), but to conclude that this is the end of the story is erroneous. The labor market adjustment (ultimately leading to growth rather than widespread unemployment) is exactly what we should expect from this transformative technology. This has happened before and will almost certainly happen again (and it seems to have already begun).

Knowledge work is just beginning

This may sound like a cliché, but it is true: this is not the end of knowledge work; on the contrary, it is just the beginning.

Automation strips away repetitive tasks and elevates human work to a higher level. The reason is simple: humans desire expansion. When a scarcity disappears, people move to higher levels. When food prices decline, we increase our spending on housing, healthcare, education, travel, entertainment, amenities, pets, safety, beauty, and longevity.

The labor market is no different. New jobs continuously emerge because human ambition never ceases, and conquering old frontiers reveals new frontiers that need to be conquered.

The emergence of new enterprises has seen explosive growth and is highly correlated with the application of AI:

The speed at which new applications are launched in app stores has increased by 60% year-over-year:

We should not view the modern economy as a museum of yesterday's jobs. Instead, it is a creative resource allocation machine, continuously generating new jobs, new tasks, new goals, and new inventions.

Robotics has largely been seen as science fiction, as the demands for computational power in dynamic environments were too high. But AI is bringing a whole new robotics industry into view:

Robotics-related datasets are experiencing explosive growth, soaring from tenth place to first in just two years.

Before AI genuinely takes effect, the robotics field has a multitude of currently unasked jobs.

Again, this does not mean all jobs will escape unscathed. The U.S. Bureau of Labor Statistics (BLS) projects a reduction in positions for customer service representatives and medical transcriptionists, and perhaps that reduction has already begun:

Some jobs will disappear, some will shrink. The economy will undergo adjustments and painful transitions, and productivity gains may take time to gradually benefit the entire economy (sometimes good, sometimes bad). We should understand these changes and strive to make them as smooth as possible, including actively engaging in retraining.

The aim of productivity enhancement is to eliminate heavy labor, and this time is no exception. However, the idea that AI would lead to the end of jobs only holds true under the assumption that human needs and ideas suddenly stop the moment AI becomes cheap. That is absurd. Personally, I do not subscribe to the "robot apocalypse" notion, and I believe I am not alone in this:

From a macro perspective, the future is not an age of unemployment; we will not be sitting around fat and lazy on electric scooters enjoying Netflix after retirement.

The future is about cheaper intelligence, larger markets, new companies, new industries, and higher-level human work. The amount of work is not fixed, nor is cognitive capability; it has never been in the past. AI is not the end of work but the beginning of a richer intelligent era.

Related Reading: Latest podcast transcript from Jensen Huang: NVIDIA's future, the "AI doomsday" theory, and business moats...

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。