AI, Unemployment and Work | Hacker News Reader

AI, Unemployment and Work

marginalrevolution.com

15 pointsspeckx12 comments

> 60% of people employed and 40% unemployed is the same number of working hours as 100% employed at 60% of the hours

More people = more overhead of almost every kind, from taxes, to communication.

Doubling the number of people in a team does not result in twice as much work being done.

People using AI in tech seem to be moving more towards 996 than a 3 day workweek.

This is a "wishful thinking" take that ignores a brutal reality: AI isn’t just a tool; it’s raising the cognitive floor for employment. We are rapidly approaching a point where the baseline IQ required to contribute at all will surpass the capabilities of a huge segment of the population. And that floor is rising.

You can already see the decoupling: Mag7 hiring has flatlined while their AI capex hits the moon. Even the "brightest minds" are being replaced by the hardware they built. No amount of policy or interest rate maneuvering can stop this shift because you can't legislate away the fact that human labor is becoming an inferior good.

The assumption that this shift can be evenly distributed and broadly wiped away via policy tools is asinine.

This time IS different.

Sidenote: This is exactly why I’m so bullish on risk assets—we are transitioning into an economy where capital no longer needs to drag the weight of a massive, increasingly "useless" workforce, and the policy responses will actually exacerbate the shift (ie. rates down on unemployment = assets up)

I think you're right. I think that a majority of people, including myself, will be generally unemployable soon.

I'm early career and have a few hundred thousand dollars I can deploy. What's the best way to invest?

I'd invest in whatever is profitable when the unemployed masses burn these data centers to the ground, like fire trucks.

I'd also invest in companies that make rope.

Sort of agree with this but I do find it amusing that this sort of take on it implicitly discounts that the capital still has to live with all these “useless” people on the same rock, whose numbers will grow to an unignorable amount...The idea that anything will be insulated is for the fairies and seems just as asinine to me

I’m getting sick of these broad sweeping statements about where things are heading.

Ive never seen someone put together a well formed argument that is rich in depth and play it through - I get that this requires vision and most don’t have it - but at the same time I’d wish people would shut up if they have nothing new to contribute.

Not to mention the bizarro economists were completely wrong about prediction re. The impact of the web.

We should absolutely use the lessons we learned about how the web unfolded to inform our predictions about AI.

Our naivete back in 1996 could be forgiven. Not applying the lessons since then and anticipating how subsequent technologies will be captured and used against us is irresponsible.

Okay so give me an application of a web lesson

The only relevant lesson is that predictions are likely to be more wrong than right tbh

The enshittification we're bathing in every day is the lesson that matters.

Anything built to democratize, will inevitably be centralized, scaled until it's scaler's lack of care to chase maximum profit becomes a massive problem, and regulated until it is de-democratized.

Example: Finance, file-sharing, recorders, AI vs. libraries, SaaS vs. Software libraries/standalone deployments.

>If investors are involved, kiss any altruistic intent goodbye. Bait-&-switch->enshittification is a matter of when, not if.

>Solutions are not the goal of work in the United States. Attracting capital, and leaving someone else holding the bag is.

Dot coms, 2008, end of ZIRP.

>You cannot fix the system from within. (Or as Einstein said, problems cannot be solved by thinking on the same level that created them)

FLOSS is architecturally unsound as evidenced by the existence of AI models, and the relative impossibility to seemingly maintain development efforts without eventual corporate co-option. GPL, AGPL have managed to carve out some minimal system stacks for the public, but with AI companies, the time has come where the System (the judiciary of the U.S. at least, is signalling that Copyright only exists if a corporate actor's legal team is sufficiently funded to argue it does. Altruism, as noble as it is; cannot persist forever without nourishment.

If you can't recognize these things, this is more indicative of a problem to observe than a lack of data to learn from.

Reply on news.ycombinator.com