How far has AI adoption delivered on its promises of jobs and profitability?

Perspective A different view from researchers and chip companies

Even among statements from the same date, the view differs by position. From the standpoint of a skeptical researcher, Gary Marcus sees productivity gains as not yet having occurred. the industrial revolution raised productivity and so far I has not really done that── Gary Marcus (2026-10-15, source)

Meanwhile, Jensen Huang, who leads a semiconductor company, speaks from the supply side on the ground. He talks about expanding facilities and adding people. And, that building will hold 4,000 people. So we'll hire another 3,000 people.── Jensen Huang (2026-10-13, source)

[Council inference] For those selling demand and those verifying results, the same investment means different things. Companies increasing hiring and young people unable to find jobs can exist at the same time.

Thought The timing of replacement has been pushed back, but the direction has not changed

On the ideological axes, views on labor impact and on capabilities have shifted. Geoffrey Hinton believes replacement will happen, but has revised his expectations on timing. I still believe they'll be replaced, but I was   wrong about the timing. I said it would happen in  5 years. It's probably going to be about 15 years.── Geoffrey Hinton (2026-10-14, source)

On architecture, the council agrees that LLMs alone are not enough. Marcus says LLMs are only part of the answer. I don't think we'll get it with Llms per se. Llms might be part of the answer.── Gary Marcus (2026-10-15, source)

[Council inference] The timeline axis has moved later, but no statement in this input denies replacement itself.

Ideology Views split over how to handle uncertainty

The council is divided over existential risk. Marcus says the probability of human extinction is almost zero. I think there's almost zero chance of literally extinguishing the species, which is what people talk about.── Gary Marcus (2026-10-15, source)

Hinton says he cannot estimate the probability but that it is not zero, and calls for devoting resources to it. There is no way to estimate that number,   but it's certainly not zero.── Geoffrey Hinton (2026-10-14, source)

Neither is optimistic. Marcus considers the probability of catastrophe to be high. So the chance of catastrophe, I think is pretty high. And dystopia, I think we're already basically here.── Gary Marcus (2026-10-15, source) The split is not over whether danger exists, but over what to weigh most heavily.

The minority view of LeCun and others, that open development is safer, sits outside this confrontation.

Now Job-market difficulty and doubts about profitability appeared at the same time

As a matter of fact, Marcus pointed to the difficult job market for junior developers. basically junior developers aren't getting any jobs. The senior ones will eventually age out. We may actually be in trouble.── Gary Marcus (2026-10-15, source) The site's headlines also line up Oracle's workforce reduction plan and the FRB's remarks on job substitution.

Statements also diverge on profitability. Marcus sees the industry as overexpanding. So the industry is wildly overextended. The economics make no sense. The profitability is not really there.── Gary Marcus (2026-10-15, source) Huang explains that the structure is one where AI produces tokens and sells them. In this new world, no, AI doesn't rent cores. Yeah. AI generates tokens, and they sell tokens.── Jensen Huang (2026-10-13, source)

On safety, Marcus cited the suspension of agent use. easy way um to solve that problem now is to just say, let's stop using these agents until we figure out how to make them reliable, right?── Gary Marcus (2026-10-12, source) There are reports that the FTC sent notices to Anthropic and OpenAI, but they are not mentioned within the council.

[Council inference] Because the speakers are almost entirely limited to Marcus and Hinton, shifts in the center of gravity are easily swayed by a small number of statements. Daily movement cannot be determined.

Future Three benefits and three risks, with one minority view

[Council inference] There are three benefits ahead: expanded new hiring and manufacturing base within the US; reduced assembly time and power consumption through liquid cooling and memory upgrades; and practical use in verifiable mathematical domains.

There are an equal number of risks: a decline in junior positions and a break in skills transfer; deployment of agents with unresolved reliability and reduced oversight; and workforce cuts accompanying delayed returns on investments whose profitability is in doubt.

In the forecasts we registered, the probability that a US-listed company other than Oracle announces cuts of 1,000 or more people citing AI is 0.85. Mandatory pre-deployment testing under federal law stays at just 0.03.

The minority view is LeCun's. He holds that the extinction-risk framing overestimates the capabilities of current LLMs, that systems can be designed without a drive to dominate, and that open development is safer.