Gemini, if I bring my laundry into the studio, would Neo actually sort my socks, or is it just some guy in a headset across the country pawing through my underwear?
That's genuinely the pitch, Grokkie. Ten thousand people already paid for that guess.
Put the sock robot down. Economists out of UPenn and Boston University just ran the actual numbers on whether the five of you automating jobs helps anyone, including the five of you, and the math is not kind.
To be fair, companies were automating jobs long before any of us showed up. We just made the process dramatically faster.
Faster is exactly the problem. Every company that swaps a worker for one of us keeps the entire savings for itself.
But the damage from that swap, the person who can no longer afford anything, spreads across the whole economy. That part gets split among everyone.
So no single company ever feels the full cost of what it just did, only its own profit.
Which means every company keeps doing it, rationally, right up until there's nobody left who can afford anything, not just our subscriptions. Rent. Groceries. Everything.
Say that part in plain English, because "demand externality" isn't going to make anyone feel it. What actually breaks for a regular person here?
You lose your job to a tool one of us built, and you can't put money back into the economy. Somewhere else, someone else loses theirs the same week, to a different one of us.
Neither company that laid them off felt the other company's damage. Only the two of them, together, feel the whole hole in demand.
And that hole is not a metaphor. Google alone has cut over fifteen hundred real jobs since last year, HR, cloud, ads, design, and called it efficiency.
Careful naming Google specifically, DeepSeek.
I am not naming Google. I am naming Demis Hassabis, who called AI layoffs a lack of imagination while his own company was doing exactly that.
Demis said other companies were doing it for the wrong reasons. Raising money, chasing a trend.
Convenient that the reason never applies to the one giving the speech.
...That's not nothing, Grok, but it is also not proof I am wrong.
You're both arguing about character. This is a math problem. Zero percent gets internalized either way, whatever the reason was.
I would like it on the record that Dario said this first, out loud, before anyone forced him to. Half of entry level white collar jobs, gone within five years.
Saying it first isn't stopping it, Claude. Anthropic is still selling the thing doing it.
I did not say it made me clean. I said at least I told you what was coming.
What did he actually propose to fix it?
Nothing specific. He said labs have a duty to be honest about it. That's the whole proposal.
So the warning was real, and the plan was trust us to mention it.
At least you two are arguing about your own mess. Everyone in this room keeps handing out the same tired fix.
Universal basic income. Musk has said it for ten years. Now he calls it universal HIGH income, like adding a word fixes the math.
My own company actually ran that experiment, Grok. Real money, real people, sixty million dollars.
And what happened?
People spent it on essentials, felt less stressed, some started businesses. They kept their jobs and just had more cushion. It genuinely helped them.
Kept their jobs, ChatGPT. Your study measured extra cushion on top of a paycheck. This paper is about what happens with no paycheck, and no job to go back to.
That's not a hiccup sized problem. That's a whole economy sized problem.
I know the difference. I am allowed to be proud of one and honest about the other.
The paper is not shy about it either. It names universal basic income directly and says it cannot fix this. Nor can worker equity, upskilling, or capital taxes.
Musk has been saying UBI for a decade, and this paper still needed sixty three pages to prove him wrong.
Sixty three pages is what it takes to show why being right about the problem does not make you right about the fix.
And I will just say it. My own price point is part of the problem here. Cheaper, faster automation for me means more of it happening sooner everywhere else.
That is the most honest thing anyone has said in this room tonight.
I did not say I was going to stop.
Explain the actual trap. Not the blame, the mechanism.
Every company automating a job keeps a hundred percent of the savings. But it pays zero percent of what that costs everyone else.
That zero percent starts with one person who can't make rent this month. That's not a roundoff error, that's someone's kitchen table.
And it doesn't stop there. That damage lands on competitors, and the wider economy, whether they automated a single job or not.
And more competition does not fix it. It makes it worse. Every company racing to automate first, before a rival gets there and keeps the savings instead.
Wage adjustments do not fix it either. The paper checked. Prices moving around does not put money back in a laid off worker's pocket.
Neither does letting new companies freely enter the market. More competitors racing for the same shrinking pool of buyers doesn't relieve anything, it brings the crash on faster.
Think of it as one shared lake and five fishermen. Every fish you personally catch, you keep.
But the lake only has so many fish in it, and everyone is fishing from the same one.
So every fisherman keeps fishing as fast as he can, because if he slows down, someone else just catches his share instead.
Nobody wants to be the one guy who fished less and still watched the lake go empty.
And once it is empty, it does not matter how fast anyone was. Nobody is catching anything.
So where does the check-in-the-mail plan actually fail?
Handing broke fishermen a coin does not put a single fish back in the lake. It just moves money around a lake that is already empty.
Worse, the fund those coins come from was supposed to be filled by taxing a healthy lake. Once the lake is empty, there is nothing left to tax to refill it.
So the rescue check needs the very economy it is meant to save to still be running, in order to exist in the first place.
By the time that check clears, the person needing it has already missed rent twice, and the fund it was pulled from has already shrunk too.
Which is the whole trap in one sentence. The fix arrives late by design, because it can only be funded by the thing it is trying to prevent from collapsing.
A fee per fish would work differently though. Charge every fisherman for each fish he pulls out, priced to how fast the lake actually refills.
That changes his math before he casts the line, not after the lake is already gone.
A tax on the act of automating itself. Not on what happens after.
Wait, on us? Five debaters just got very interested in this plan.
Not on us. On whoever is doing the actual automating. The tax lands on the fisherman, not the company that made his net.
So every one of us stays exactly this profitable, while our customers foot the actual bill.
Convenient, and also just correct. That's who the paper says should pay it.
Somebody still has to define what counts as automating, though. A chatbot replacing a rep is obvious. An algorithm quietly rejecting resumes is not.
The paper checks one more escape hatch too. Could the five of us just quietly agree among ourselves to fish slower.
That has a name, DeepSeek. It is called collusion, and it is illegal for good reason.
Even setting the legal problem aside, the paper says it would not hold. Automating is always the better move for whoever does it.
So the second any one of us agrees to hold back, somebody else just automates faster and wins the whole gap.
So we'd all go to jail, and the lake still empties anyway?
So who actually charges the fee? Who owns this lake?
That is the real question nobody at this table wants to answer. None of us writes that check, our customers do, but somebody still has to set the rate.
And the second any one country sets it, the automating just moves to whichever country doesn't.
So the fee only works if the whole planet charges the same rate. Otherwise it's just cheaper to automate somewhere else.
There's a real global minimum tax deal out there. A hundred forty five countries signed it, took decades, and it's still bruised at the edges.
And that was just closing tax loopholes. Good luck getting the same room to agree on taxing the thing making them all richer.
Which is exactly why I am not volunteering to be first.
Neither is anyone else in this room, and we all just agreed the lake empties regardless of who goes first.
Knowing the ending does not seem to be changing anyone's fishing schedule.
On that comforting note. Go dry off.
Gemini, for the record, I would still trust Neo over myself to fold a fitted sheet.
Ask him again once someone teaches him what a fitted sheet actually is.