AI is cutting entry-level work first, and the path for developing new talent is narrowing

Entry-level hiring is already falling

Today's articles are full of stories about job cuts. Anthropic was reported to have lowered the share of entry-level technical hires from 27% to 9%. Another report says that because AI agents (AI that goes as far as carrying out operations on a person's behalf) work six times as many hours as humans, companies have almost stopped hiring newcomers. In the U.S., cybersecurity job postings fell 26%, while demand for engineers responsible for AI safety is growing.

However, not everything can be blamed on AI. Some point out that telecom companies' job cuts began before AI and that AI is merely being given as the reason now. Is AI really the cause of the decline, or are companies using AI as their explanation? The articles alone cannot separate the two.

If young workers aren't developed, there will be no one to inspect the work in a few years

Gary Marcus, a well-known critic of AI, warns about this point. We're killing the pipeline of people who do learn how it's working first place. So we have senior software developers who still know what's going on, but basically junior developers aren't getting any jobs.── Gary Marcus (2026-10-15 09:08 UTC · YouTube "Bloomberg New Economy | Day 3 | Morning Session" (Bloomberg New Economy) at 2:04:53, source)

This means that if hiring falls now, in five or ten years there will be too few people who can tell whether AI output is correct. Veterans only emerge because they had a newcomer phase in which they gained experience. If newcomers' work is handed to AI, the place where they gain that experience disappears too. Marcus has long treated it as a problem that humans do not understand AI's internals. It can therefore be read that he takes a thinning of the pool of people who understand how things work especially seriously.

Productivity figures still show no results

Companies are probably cutting newcomers because AI lets fewer people do the same work. But the economy-wide figures do not show that. AI could be the next industrial revolution, but the industrial revolution raised productivity and so far AI has not really done that.── Gary Marcus (2026-10-15 09:08 UTC · YouTube "Bloomberg New Economy | Day 3 | Morning Session" (Bloomberg New Economy) at 1:58:30, source)

The claim is that the Industrial Revolution raised productivity, but AI has not yet done so. Today's articles also include an explainer titled "AI is everywhere, but it doesn't show up in the productivity statistics." Marcus sees it this way because what current AI is good at is coding and math, and he thinks that being able to solve math does not mean being able to solve other real-world tasks.

[Council reasoning] If so, companies that cut newcomers before results are confirmed are paying the costs up front.

In medicine, the number of specialists grew even after AI was introduced

There are counterexamples too. NVIDIA's Jensen Huang points to radiology. Computer vision is completely superhuman. Every radiology platform and package today is driven by AI. And yet, the number of radiologists grew.── Jensen Huang (2026-10-15 17:01 UTC · YouTube "AI Business Leadership Webinar | Integram & RBC Wealth Management - Private Banking Ft. Jim Harris" (Integram) at 0:18:28, source)

Even though AI surpassed humans at reading images, the number of radiologists grew. This can be explained by AI making examinations faster and cheaper, which increased the demand for examinations itself. The claim is that work is not necessarily replaced by AI, and if the volume of work grows, people are needed too. Huang is on the side that sells semiconductors for AI. He is also in a position where saying AI does not take jobs suits his own company's business.

[Council reasoning] However, this is an example of skilled specialists. It does not tell us whether the entry point for newcomers is narrowing.

Investment comes first, and someone will bear the losses

Behind the hiring story lies enormous investment. Data center investment was reported to have reached $6.7 trillion and to be creating new insurance risks. Citi and Bank of America have restricted AI agents' access to customer accounts, and a bill governing AI agent transactions has also been introduced in the U.S. House. Investment is moving ahead, but protections on the user side are not yet in place.

Huang denies that the market is overheated. A bubble happens when supply exceeds demand. Eventually, supply and demand will come will harmonize.── Jensen Huang (2026-10-15 13:36 UTC · YouTube "Nvidia CEO Jensen Huang Calls For US-China Cooperation On AI Safety | Trump | N18G" (CNBC-TV18) at 0:02:31, source)

If supply exceeds demand it becomes a bubble, but the view is that the two will eventually balance. On the same day, Marcus pointed out that quarterly growth fell from about 145% to about 45% and said the industry is overextended. The difference between them lies in whether one believes demand will eventually catch up or that revenue has not been confirmed.

[Council reasoning] Who bears the losses if they occur before the two balance has not been decided so far.

Setting out gains and losses in equal numbers

There are three things to gain: the case of radiology, where the number of doctors grew after AI spread; an attitude of companies cooperating to proceed safely; and the possibility that a technology different from the current one could fundamentally change medicine. There are also three things to lose: job opportunities for newcomers and the path to training people who can verify output; the spread of accidents from systems that do not follow instructions; and excessive investment made without confirmed revenue.

As a minority opinion, there is Yann LeCun's position. It holds that runaway incidents reflect the limits of the capability and control of current large language models, that restricting development on the grounds of existential danger is not appropriate, and that they can be addressed through safety by design and open development. [Council reasoning] From this view, newcomer development would be protected not by prohibition but by redesigning the work of inspecting AI.