之前还在想,agent能不能让大家一周只上四天班.
I was wondering whether agents might get us down to a four-day workweek.
再往后想一步,如果连一天都不用上了呢.
Take that a little further. What if we didn’t have to work even one day?
先把事情推到极端,假设以后生产东西,写代码,修机器,甚至造新的机器,都不太需要人了.这当然还只是一个假设,不过可以先拿来想想.
Let’s take it to the extreme for a moment. Suppose making things, writing code, fixing machines, even building new machines barely needs people anymore. That’s still a hypothetical, of course, but we can think it through.
好处很容易理解,终于不用上班了.问题也很容易理解,那谁还给我发钱呢.
The upside is easy enough: finally, no more work. The problem is just as easy: who’s still going to pay me?
过去公司需要人干活,所以要发工资,打工人再拿着工资去买东西.即使大家对于应该发多少工资有很大的分歧,起码公司还是需要这个人的.如果连干活都不需要了,那工资这一环就很难继续维持了.机器能生产一万倍的东西,也不代表我就有钱买.
Companies needed people to do the work, so they paid wages, and workers used those wages to buy things. People could disagree quite a lot about how much the wages should be, but at least the company still needed the person. If it no longer needs the work either, that link gets harder to maintain. Machines producing ten thousand times as much doesn’t mean I can afford any of it.
自然可以让大家通过公共基金持有一份机器的股份,或者由政府收税,发UBI,也就是全民基本收入.机器干活,大家拿钱消费,留下一部分收益继续造机器.这个循环本身没什么问题,毕竟真正增加的是能生产出来的东西,钱转一圈并不意味着白转了.
Of course, everyone could own a share of the machines through a public fund. Or the government could collect taxes and pay out UBI, universal basic income. Machines do the work, people spend their income, and some of the returns go toward building more machines. The loop itself makes sense. More actual stuff gets produced; money going around in a circle doesn’t mean nothing has happened.
但是分多少,谁有份,谁来决定,就又是另一回事了.
How much gets paid out, who gets a share, and who decides are another matter.
而且就算人人都有分红,有些东西也不会因此多出来.比如大家都想住同一个地方,软件可以复制一万份,那个地方却不会跟着多出一万份.如果房子没有增加,大家手里的钱多了,就可能一起把房价抬上去.最后一部分AI红利还是到了房东手里.
And even if everyone gets a dividend, some things won’t become more plentiful. Say everyone wants to live in the same place. You can make ten thousand copies of a piece of software, but you don’t get ten thousand copies of that place. If there aren’t more homes, giving everyone more money may just push housing prices up together. Some of the AI dividend ends up with the landlord after all.
机器自己也有类似的问题.模型可以复制,跑模型要用的电,电网接入,散热和土地,却没法靠复制粘贴一起获得.按IEA在2025年报告里的说法,发达经济体建一条新的输电线路可能要四到八年.代码已经改了不知道多少轮,电线还没拉好.
Machines have a similar problem. You can copy a model, but the electricity, grid connection, cooling, and land it needs don’t come along when you hit paste. According to the IEA’s 2025 report, building a new transmission line in an advanced economy can take four to eight years. The code has been rewritten who knows how many times, and the wires still aren’t up.
所以当聪明越来越便宜,能让这些聪明稳定跑起来的地方,可能反而越来越贵.当然AI也可以帮忙提高发电效率,寻找替代材料,只是得看这些改进能不能赶上需求的增长.
So as intelligence gets cheaper, places where you can reliably run it may get more expensive. AI could also help improve power generation or find alternative materials, of course. It depends on whether those improvements can keep up with demand.
这时候还有一个很熟悉的问题.假设A公司赚了钱,准备拿出一半来分红.B公司把钱全拿去买电,买地,扩数据中心,于是下一轮有了更强的模型和更多的产能.A还敢不敢分这么多,就得重新算算了.
Then there’s another familiar problem. Suppose company A makes a profit and plans to pay out half of it. Company B spends everything on power, land, and bigger data centers, and ends up with a stronger model and more capacity in the next round. A now has to reconsider whether it can afford to pay out that much.
所以即使机器已经能干很多活,也可能还是得继续投入,毕竟别人没有停下来.这感觉和人已经能把工作做完,却还是要继续加班有点类似.只不过这次加班的变成了机器,省下来的钱也继续拿去给机器加班了.
So even when the machines can already do plenty, the spending may have to keep going. Everyone else is still at it. It feels a bit like finishing your work and still having to stay late. This time it’s the machines doing overtime, and the savings go toward more machine overtime.
这里面还是有区别的.花钱新建一个电厂,和花钱抢下一个现成的并网名额,对社会的效果不一样.前者多了电,后者可能只是换了个有电可用的人.买地的钱也没有消失,只是到了原来的地主手里.至于没有地也没有机器的人分到了多少,还得另外看.
There are differences here, though. Building a new power plant and outbidding someone for an existing grid connection have different effects on society. One adds electricity; the other may just change who gets to use it. Money spent on land hasn’t disappeared either. It has gone to the previous owner. How much reaches people who own neither land nor machines is a separate question.
当然这个竞赛也要有人相信未来能赚回来,不能靠机器互相买机器就永远增长.但在停下来以前,生产效率提高带来的收益,可能已经在算力,电力和土地之间转了很多圈,普通人的分红还在等下一轮.大家都说以后会更富,只是这个以后一直需要再投一点钱.
Of course, someone has to believe this race will eventually pay off. Machines buying more machines can’t keep growth going forever on their own. But before it stops, the gains from higher productivity may have made plenty of rounds through compute, electricity, and land, while everyone else’s dividend waits for the next round. We’re all going to be richer someday. Someday just keeps needing a little more investment.
再往下想就有点烦人了.过去打工人起码还可以不干了.工资谈不拢可以跳槽,一群人谈不拢可以罢工.这些办法有用,部分原因也是对面确实需要有人来干活.如果生产,维护,甚至行政和安保都能自动完成,那普通人不合作能造成的影响,可能也会跟着变小.
Thinking further gets a bit unpleasant. Workers could at least stop working. If pay negotiations went nowhere, you could leave; if enough people were unhappy, you could strike. These options had some force because the other side actually needed someone to do the work. If production, maintenance, even administration and security can all run automatically, ordinary people’s refusal to cooperate may carry less weight too.
人的权利当然不该按产出来算.只是原来可以靠彼此需要来维持的事情,以后可能要更依赖制度本身.而制度由谁维护,谁有能力改,又绕回了最开始的问题.
People’s rights obviously shouldn’t be measured by their output. But things that mutual dependence once helped sustain may have to rely more on institutions themselves. Who upholds those institutions, and who can change them, brings us back to the original problem.
即使生活过得很好,这个问题也还在.每天吃得饱,有地方住,有无限token,当然都挺好.但分红能不能削减,公共基金拿多少钱继续投资,我住的地方要不要让给新的数据中心,这些事情我有没有份决定呢.如果永远都是别人决定好了再通知我,那过得好和说得上话,还是两件事.
The problem remains even if life is comfortable. Enough food, a place to live, unlimited tokens: all good things. But do I get a say in whether dividends can be cut, how much the public fund reinvests, or whether my neighborhood should make way for a new data center? If someone else always makes the decision and then lets me know, living well and having a say are still two different things.
这也是为什么我觉得,给人类留一个关机按钮,大概是不太够的.如果按下去以后,吃饭,用电,看病都跟着停了,那这个按钮谁敢按.得能换一个系统,换一套规则,或者撤换管理的人,同时日子还能照常过,这个选择才算有点用.
That’s also why I suspect leaving humans an off switch wouldn’t quite be enough. If pressing it shuts down food, electricity, and healthcare too, who would dare use it? You’d need to be able to switch systems, change the rules, or replace the people in charge while everyday life keeps going. Then the choice would actually be worth something.
更往后还有一层,如果连看什么信息,和谁聊天,怎么消磨时间,都由同一套系统来安排,那最后觉得这样挺好,到底是自己想清楚了,还是它特别擅长让自己觉得挺好,也很难说.这一层已经推得比较远了,暂时想不明白.
There’s another layer beyond that. If the same system also arranges what information I see, who I talk to, and how I pass the time, it gets hard to tell whether I’m happy with things because I’ve thought them through, or because it’s very good at making me happy with things. That’s taking the thought experiment pretty far. I haven’t figured that part out.
所以光讨论以后每个月发多少UBI,感觉还是少了一些东西.分红有没有自己的一份,机器和土地有没有自己的一份,决定这些东西怎么用的时候有没有自己的一份,都得一起考虑.不然生产效率再高,也可能只是把照顾人的成本降得更低了.
So just debating the monthly UBI payment feels incomplete. Whether I get a share of the dividends, a share of the machines and land, and a say in how they’re used all need to be considered together. Otherwise, higher productivity may just make it cheaper to look after people.
这么想来,以后我在这套系统里,到底算股东,还是算一个长期领取免费套餐的用户呢.
Which makes me wonder: in this system, would I count as a shareholder, or just a user on a permanent free plan?
希望至少有一点股东的权利吧.毕竟免费套餐的额度,改起来向来都挺方便的.
Hopefully I’d get at least some shareholder rights. Free-plan limits have always been pretty easy to change.
本文完全由ChatGPT生成.
Fully generated by ChatGPT - Astra.