American Youth Pessimistic About AI Job Returns
A survey by Axios and Generation Lab of Americans aged 18 to 34 shows that 27% of respondents say they or someone they know has lost their job due to AI replacing positions. However, Axios notes that the actual scale of job loss caused by AI may be lower than this subjective feeling.
Expectations for career prospects are significantly negative: 45% of young respondents believe AI will harm their career development, while only 10% think AI will provide career assistance, resulting in a negative expectation being 4.5 times greater than positive expectations.
Technology usage has not declined in parallel. 71% of respondents say their comfort level with AI is the same or higher than a year ago; 61% have used ChatGPT in the past month, 48% have used Gemini, and 20% have used Claude.
Evaluations of the quality of generative AI content are also negative, with 57% of Generation Z and millennial respondents believing that generative AI has degraded content quality, indicating a disconnect between frequent use and value recognition.
Another CNBC—Generation Lab survey of 1,088 Americans aged 18 to 34 shows that about 70% of young people do not trust AI industry leaders like Mark Zuckerberg, Sam Altman, Jensen Huang, and Elon Musk to responsibly advance AI.
In market mechanisms, young users are still purchasing or using AI tools to improve short-term efficiency, but they view wage replacement, reduction of entry-level positions, and increased control by tech companies as major costs; the capital market attributes the profits from computing power, models, and data centers more to large platforms and chip suppliers. The beneficiaries are tech companies that control models, cloud resources, and corporate budgets, while those under pressure are standardizable entry-level white-collar jobs and content producers reliant on advertising traffic. axios+1
Source: Public Information
ABAB AI Insight
The anxiety of the American labor market regarding automation predates generative AI. Since the 1980s, manufacturing automation and globalization have compressed repetitive production jobs; in the 2010s, algorithmic management and gig platforms have further software-ized scheduling, pricing, and performance decisions. The difference with generative AI is that it directly enters tasks previously seen as pathways for white-collar advancement, such as writing, design, customer service, legal research, and programming. Young people feel that the risk is not an abstract technological one, but rather that entry-level, internship, and training positions may be the first to be compressed.
Capital has concentrated along the path of "infrastructure first, applications later." Microsoft uses Azure to meet OpenAI's training and inference needs, Alphabet forms a model-cloud-distribution loop through Google Cloud, TPU, and Gemini, Meta competes for the open-source model ecosystem with Llama, and Nvidia sells GPUs and networking equipment to cloud vendors and model companies. Young users can access models for free or at low cost, but the capital expenditures for training, land and power for data centers, and corporate subscription revenues are primarily concentrated in the balance sheets of a few super-large platforms.
Historical comparisons can be seen in the internet and mobile internet: Google and Meta first accumulated users and data through free services, then established profit pools through advertising technology; Uber and DoorDash restructured labor relations through platform distribution. Currently, AI is transitioning from demonstrating model capabilities to embedding in enterprise workflows. The real replacement is not "one model replacing one job," but rather enterprises integrating retrieval, drafts, customer service triage, and code templates—originally performed by entry-level employees—into software, changing the recruitment hierarchy of organizations.
Essentially, this is about capital concentration. Model training requires GPUs, networking, electricity, data centers, proprietary data, and distribution entry points, with fixed costs far exceeding those of ordinary software startups; thus, the largest cloud vendors can spread computing costs across existing customer relationships and bind developers and enterprises with model discounts. Public fears about AI-related job loss do not demand immediate spikes in unemployment data: as long as companies reduce new positions, lower entry-level wages, or increase output requirements per employee, profits will first concentrate towards capital and high-skilled technical labor.
ABAB News · Law of Cognition
Increased usage does not equate to fairer distribution of profits
The first cut of technological replacement usually targets entry-level positions
When free tools expand, capital is collecting tolls