My understanding is that the obstacles to stores replacing the remaining cashiers with self-checkouts is not so much "we need a better machine" but (a) the shoplifting deterrence effect of staffing, (b) customers are slower at packing than cashiers, which can cause queues and other inefficiencies.

Tellingly, the newest supermarket in the area (Tesco, not any sort of weird startup) only has space for one manual checkout, plus about 15 self-checkouts.

Where I'm from, cashiers don't pack the groceries of customers

My biggest gripe with self checkout is if I make a mistake, they call it theft. I haven't been trained and certified on the register, so I can't accept the potential for getting a criminal record due to a scanning error. Example: Walmart will successfully "beep" when scanning, but will display "system busy" on the screen, and the scanned items don't show up on the receipt.

https://en.wikipedia.org/wiki/Luddite

Genuine question. I'm trying to find a difference, but not succeeding.

The Luddites were kind of unusual in the history of things. We've had all sorts of tech innovation over the last two centuries and not much has lead to riots. The luddite riots happened at time of high unemployment and inflation after the Napoleonic wars and may have been more a last straw rather than the root cause of the dissatisfaction.

it seems exceptionally well-suited to AI-based automation because software engineering has already needed to figure out how to efficiently cope with humans who sometimes produce code which may have defects. most obviously automated testing and type systems. also a lot of programming tasks have verifiable solutions so it's also better for training. it seems natural that the most obviously successful AI tools are for coding.

yet the software industry has also been absorbing waves of automation for 70 years. Fortran and COBOL were referred to as "automated programming" and there was a narrative that these new tools would make it easy for non-specialists to program. this time may be different, or it may just be another wave of automation.

i think software has pretty unique properties among white collar jobs, and I would hesitate to draw conclusions about other industries based on AI progress in software engineering.

I also don't really care if my job is automated. If it doesn't need doing anymore, why should I want to do it? I will do something else.

Realistically, we are going to see 20-30% reductions in headcount in the near-to-medium term. THIS IS STILL CATASTROPHIC.

A number of earlier stage companies I've funded have already been heavily utilizing automation to simplify code generation or scaffolding/project ops work. They use the cost savings to hire experienced SWEs at high base salaries and are able to hit the same development metrics as they would have if they had a large team of average paid SWEs. On the BDR side, they are using a massive amount of video/audio automation to scale out cold calling or first impressions, so reducing the need to hire teams of BDRs hitting the phones all the time. And finally, they are automating tier 1/2 support ticket responses and communication so reducing the need for teams of support engineers spending time basically respondng to customers with the polite equivalnet of "read the docs".

Basically, a Series A startup that would have had a staff of 50 employees 10 years ago can essentially output the exact same as a Series A startup with a headcount of 20 employees today, and with a tangible path to FCF positivity.

This is a net reduction in jobs, and a significant one at that, because most people just cannot upskill - it's hard.

Impressive.

> This is a net reduction in jobs, and a significant one at that, because most people just cannot upskill - it's hard.

So you lose out on people to do work? That they cannot upskill implies that there are jobs to fill. That seems like an economic void there. If the jobs cannot be filled something else must compensate.

But I see no reason to accept this narrative. In the long run we’re all upskilled out of a job. Our tiny brains can’t keep up. Certainly if something as simple-sounding as code generation and/or scaffolding can reduce headcount by this much. What will full-force AI do then?

Maybe they shouldn’t spend their time chasing that promise of bigger brain job tasks and (maybe even) higher pay (because big IQ means big pay?). Eventually the rent-seeking becomes too naked to ignore. Maybe around the time when the whole operation is maintained by the last twenty or so biggest brains of the lot (survival of the biggest). Then their tasks all get automated. Whose left standing with the spoils? Not the biggest brains.

Whether this amount of success is proof that there won't be any issues with the tech in that area, remains to be seen. Hell, they're not even interested yet in talking about how this may pan out outside their Houston trial runs.

April 29, 2014 - "Milken 2014: Driverless cars due in five years"[0]

Nov 24, 2015 - "Ford is 5 years away from self-driving cars"[1]

Oct 20, 2016 - "A Driverless Tesla Will Travel From L.A. to NYC by 2017"[2]

Now general consensus is level 5 autonomous self-driving is decades away, at least.

[0] https://www.usatoday.com/story/money/cars/2014/04/29/milken-...

[1] https://www.businessinsider.com/ford-is-5-years-away-from-se...

[2] https://www.nbcnews.com/business/autos/driverless-tesla-will...

Remember that this was before transformers, LLMs, and more recently VLM/VLAMs, though.

I'm not sure to what extent these are already integrated in self-driving hardware, but I would not be surprised if we see a big improvement in self-driving due to related technologies soon, especially with the smaller models becoming far more (efficient and) potent.

The typical tech person will reply to this with some variant of "that shouldn't matter". Well, it does.

It's like saying Windows 95 doesn't have any security flaws, as long as you don't connect it to the internet.

Me: My Tesla takes me 20min to work every day without human intervention, and on long trips of hundreds of miles

You: This is illegal because it might kill someone

Call the police on Tesla, I guess?

Unrelatedly, what's the law on texting, answering a call, and just quickly checking a message while driving? All crimes, I guess?

But I think the smart money is that Waymo will get appreciably better in the 5-10 year timeline. Perhaps massively, crazily better, but at least steadily, noticeably better.

My bad.

They're deploying Waymo in Atlanta now and they're already driving around sans-humans.

Atlanta is not a grid. Atlanta has crazy and dangerous road design. Random roads built in random places. Potholes, lots of hills, sharp angles, lack of shoulders, no margin or protection from 45mph oncoming traffic (90mph delta with ZERO margin or division), weird infrastructure, five and six and even seven way intersections, pedestrians, jaywalkers, motorcycle gangs that ride in the wrong lanes. You name it, Atlanta's got it.

Atlanta has lots of weather. Tremendous amounts of subtropical rain in the summer, fog, and other low visibility and dangerous conditions. Intermittent summer rain that pours so hard that it's a literal rain curtain white out. Seasonal tornadoes.

Self-driving is here now. It'll be everywhere by the end of the decade. I can't wait to buy my own Waymo-equipped vehicle.

That time when people had blocky cell phones. Then they were in everybody's pocket.

That's self-driving now.

It's here, just uneven. You can ride in one today. Soon, it'll be taking over for trucking, delivering everything you can think of on-demand, and revolutionizing how we think about transit and travel.

Instacart and Uber Eats and Amazon same-day are blocky cell phones. Soon we'll have instant logistics. A ten times faster Amazon. A world where almost everything can be delivered instantly and cheaply.

Trains and subways and metros are rigid and inflexible transit corridors. Soon we'll have last-mile transit pods for everyone that connect everything and everywhere and make suburbs highly desirable and sought after again.

Self-driving will be like the internet or smartphones all over again. And we're right at the precipice.

The financial gradients of all of this ensure that it'll happen and that it'll happen fast and all at once once we get past the inflection point.

But back to your point: you wouldn't call those blocky cell phone users non-cellphone users. The tech just wasn't evenly distributed. It got there in short time, because the tech was world-changing. So too shall it be with self-driving. The economic advantages are inevitable.

And you could do it in 2020. It seems unlikely that it will be “everywhere” in 2030 unless you give that word a quite restricted meaning.

The issue around self driving is more laws and regulations than it is a matter of tech.

There’s a reason these companies spend so much time effort and money on legal hurdles.

What jobs will AI create that AI cannot itself do?

In the automation revolution, the bots were largely single purpose, the bots couldn’t be created by bots. There could and probably will be trillions of jobs created by AI, but they will be done by trillions of agents. How many jobs do you really create if ChatgGPT is so multi-purpose, it only takes one say 250k company to support it.

Part of the problem is the definition of "AI" is extremely nebulous. In your case, you seem to be talking about an AGI which can self-improve, while also having some physical interface letting it interact with the real world. This reality may be 6 months away, 6 years away, or 600 years away.

Given the current state of LLMs it's much more likely they will create jobs, or change workflows in existing jobs, rather than wholesale replace humans. The recent public spectacle of Microsoft's state-of-the-art Github Copilot Agent[0] shows we're quite far away from AI agents wholesale replacing even very junior positions for knowledge work.

[0] https://news.ycombinator.com/item?id=44050152

Are they better now? When ChatGPT came out I could ask it for the song lyrics to The Beatles - Come Together. Now the response is

> Sorry, I can't provide the full lyrics to "Come Together" by The Beatles. However, I can summarize the song or discuss its meaning if you'd like.

You can argue that ChatGPT "knowing" that 9.11 is less than 9.9 or counting the 3 r's in strawberry means it's better now, but which am I more likely to ask an LLM?

Good question. And one where "in the immediate..." is not relevant. In the long term then, what?

One direction: Humans are not good for semiconductor fabs. Humans bring contamination of all kinds into a space that shouldn't have any. Humans can't help themselves and tweak things they are not qualified to. Humans can't help themselves and don't report changing conditions that they should. etc, etc. So that current fabs contain mountains of automation. And yet they also have lots of humans. Full automation is really difficult. It's more that automation creates the economic conditions that make it manageable to employ the remaining humans.

> How many jobs do you really create if ChatgGPT is so multi-purpose, it only takes one say 250k company to support it.

Is this very relevant? For example the entertainment industry - I expect - relies on just a handful of data centers for digital distribution. But creating the movies and TV shows employees insane numbers of people. The question is more "What does NextGenChatOpen enables, that it cannot exploit itself"?

Basically any job that uses a word processor, spreadsheet, drawing tool, etc will all become more efficient and if Jevons Paradox applies, demand for those things will increase beyond the reduction due to efficiency gains.

I can imagine that for many fields it will be cheaper to have humans use AI (in the near term) rather than try to make fully automated systems that require no/little supervision.

A) Fire half your staff, now earning $100 profit? ($50/yr to employees, $150/yr revenue)

B) Attempt to double your business, now earning $200 profit? ($100 emp, $300 rev)

C) Double your headcount and earn $400 profit?

If companies are profit-seeking, B and C should be more common than A!

AI lack skin-in-the-game, they cannot bear responsibility for outcomes. They also don't experience desires or needs, so they depend on humans for that too.

To make an AI useful you need to apply it to a problem, in other words it is in a specific problem context that AI shows utility. Like Linux, you need to use it for something to get benefits. Providing this problem space for AI is our part. So you cannot separate AI usefulness from people, problems are distributed across society, non-fungible.

I am not very worried about jobs, we tend to prefer growth to efficiency. In a world of AI, humans will remain the differentiating factor between companies.

If AI is roughly where IT is in the 60's, we might see actually decreased productivity for a while until people (yes people) figure out how to use it effectively.

Why?

If you're waiting around for AI to do these things by the time it happens it will be getting hit like a truck. The speed of technological implementation these days is very fast, especially when compared to the speed of regulation.

Moreso we are not just seeing improvement in things like LLMs, there is broad improvements in robotics and generalized behavior in AI.

For example, even outsourcing giants like Infosys shrank hiring by 20% AND increased personnel utilization from 70% to 85% just by mandating employees to start using code-gen tools, and as a result were able to significantly enhance margins.

https://www.epi.org/productivity-pay-gap/

The questions of who will benefit from AI is a political questions. As power is currently firmly in the had of the capital class, AI will only make the rich richer and the poor poorer.

Wages are not based on how much value you produce but how much it cost to reproduce your labor. How much it cost for you to stay alive and be able to work and feed your family. What kind of living standard that entails is a political matter and depends on how strong your unions are, how willing you are to fight for it and so on.

Same with welfare. It is something which was fought for and if the power balance is such that they can get away with giving you less, they will do. In fact some rich pricks would gladly turn you into soylent green if they could get away with it.

I have hope in decentral technologies. Solar power is a rare example. If things gp right then in a few decades, your typical household will just not depend on anybody anymore in terms of energy needs. That's a significant shift.

That is, if big corps don't manage to capture that somehow, though I can't currently see how. But it has happened before. The www and email used to be as decentralized as you could imagine. Nowadays rarely any email gets sent that doesn't come from or go to a gmail server, or definitely not outside one of the biggest 3 providers. Similarly with the www. Everything is so centralized now, it's a great success of big tech.

Frontier models are being provided at a loss right now to my "knowledge", and they're already as crap as you'd want to go, if not even crappier, for worker replacement in my experience.

So right now, I have a really hard time seeing LLMs as a well-spent money if they were bought at their real cost

And as exhibit A I provide progress in automation in the last decades (enormous) in relation to the median buying power of the population in developed nations.

It just doesn't work the way you describe, but I have sympathies for naïvely believing that it does.

It doesn't matter how many products and services your society shits out, what matters is how you organize it. And clearly we're already way past the point where the amount of products and services is a bottleneck to quality of life. Access to healthcare, job security, home ownership, the ability to afford kids, good food, all of these things are going down, not up, despite the fact that as a whole we're much better off than in the last century.

Virtually no one can sustain a family on a single salary, it was the norm not so long ago

https://www.researchgate.net/profile/Haruki-Seitani/publicat...

Housing is more and more out of reach

https://www.researchgate.net/publication/335721006/figure/fi...

Most of the productivity gain doesn't translate into income:

https://files.epi.org/charts/img/235212-28502-body.png

Sure enough we now have access to instant brain rot doomscrolling services, infinite low quality entertainment, vast amount of ultra processed food, &c. but if anythings these things are distractions.

Same goes for my movie theater, when did it x2 its price in the past 5 years while having less employees ? Oh wait, there's only one theater in town and I don't have a car, that's probably why.

Taken for granted that AI is extremely cheap. Product and services don’t become cheaper just because they are easier to produce. One way to get rich is artificial scarcity.

Now the only way I know that people in most jobs got better pay was to organize. Including striking. You can’t bully people too much if they can shut down sectors of the local community or country.

Cheap AI can make things cheap. Or not. Who owns it? Some dozens of rich people? Do they have to make it cheap? Do governments have to make living on social welfare (who says there will be social welfare?[1]) “more luxurious than working full-time is now” when you look around and... there’s no one to pressure them to do that any more?

[1] Some people with a lot of money stuffed in their ears think that social welfare is something that the poors/the lazy can just fall back on indefinitely. They know nothing about means testing or the fact that social welfare has been fought against by the very rich (ideologues who are savvy, they certainly didn’t stuff money in their ears) since social welfare was introduced.

I’m learning that every topic that people read should be about AI. Also the hypothetical ones. One I barely read one month ago was some pseudo-psychological self-care piece about how to soothe ourselves as we have to deal with the inevitability of AI changing our very self-identity. They do insist too much not to raise concerns about their intentions. Oh, and very regular commenters on the Web are also very weirdly insisting that if some specific percentage of “your code” or higher is not AI then some bad stuff is about to happen to you. Also whispers about management demanding... yes their very own percentages for how much AI code output. Could tech CEOs be motivated to write about every hypothetical thing that is not a thing yet (but will be immediately surely, probably tomorrow)? Of course it’s not my place to be cynical. Everyone who writes about AI has good intentions until proven otherwise.

Now, any CEO is comfortable with writing about protests against job loss in an American context. Because they just are flashes in time.

> Bias is a little different. America did flood the streets after George Floyd in a more sustained manner, for many months, but major reforms stalled out once marches faded and partisan lines re-hardened.

Do you remember when Occupy Wallstreet changed America and put the 99% in charge? Me neither.[1] America has these heroic uprisings spread across years. Then what happens? The out of touch people in power listen a bit more attentively?

[1] In hindsight (three minutes later) that’s unfair and hyperbolic. What could have more realistically happened was that some organizing came out of it that would have survived to this day. Organizations that might have changed names and members and “leadership” (as everyone calls it now) but that anyone with a slight interest in the matter could point to and say, oh yeah that’s a product of Occupy Wallstreet, right there.

> How do we help the displaced?

Of course the chosen focus is on helping the displaced. Do they deserve handouts and from whom? But this misses the mark, right? If AI hypers like himself are correct then “helping” becomes at best outdated, outmoded. I’m sure the powers would be would love the dichotomy of helpless displaced people getting either help from the government or the “AI winners” (he uses scare quotes for reasons).[2]

But if AI job displacement becomes massive this is a dead end. Displace enough people without making new jobs and jobs themselves become outmoded. Remember that people don’t strike or protest “for jobs” because they intrinsically want a wage job... they want to live and survive and this is the means they have to do that. So what if jobs are gone, then what? Then the means of AI should be socialized. If AI displaces jobs we don’t need jobs any more. Because we can just use AI. If we own the AI... not if dozens of billionaires own them. Some CEOs might disagree on this point.

[2] The right-wing libertarians are the ones who love Universal Basic Income.

> Two historical moments that resemble this pattern: When automated textile frames wiped out skilled jobs in the U.K. in the early 1800s, the Luddite riots turned violent enough (including killing a factory owner) that Parliament dispatched roughly 12,000 troops to restore order, which concluded with over a dozen executions.

Noteworthy that he chose a violent crackdown example (sorry, restoring order) with no mention of any fruits that were won.

I have no doubt that he will report any violent occurrences of restoring order as dispassionately as he does here on this hypothetical piece.

> Do protests even matter?

> I need to dig in more, but the short answer seems to indicate yes, in a few ways. They seem make the people who attend them more politically motivated, at least for that issue. As a result, if enough people go to them, then they can swing elections.

Political scientists are funny. The kind of people who make ostensibly democratic participation into rat maze experiments.

The link just seems to be about whether or not protests can affect elections. Like they specifically and narrowly studied that. That’s a bit boring?

I don’t think the Vietnam War protests just helped impact elections. Are you kidding me? The anti-war movement up until Reagan forced that administration to make their Central America interference clandestine. Well, it was bad enough that they did it. But popular resistance was strong enough that they forced them to at least do some things differently. Imagine a repressed or apathetic society where the government just supports anti-democratic paramilitary groups in foreign countries out in the open. That’s worse in a sense.

There’s some anecdote about Kissinger and Nixon in a car where Nixon says that he’s afraid that the crowd could kill them. That’s a bit more than swinging an election.

I’m sure that Apartheid South Africa was crushed in part because of protests and boycots. Although it wasn’t on my news back then.

... But what do we get according to this article? Swinging an election. Wow. That sure will motivate Americans with their two-party system where either party serve slightly different sets of corporate sectors.

Notice these people. These people who insist that you protesting and organizing and working to change things, well it might move the needle electorally. That’s what they want you to think. That that’s all the political power you have. You can change the ballots that are presented to you. And you can persuade fellow voters. That’s it. Now go home.

We are gonna have to change more than that if the AI hypers are correct.

The reason working people's jobs haven't been automated away is because people are cheaper and more conscientious than robots. Even worse, the robots are owned by famous rent-seekers who see you as prey, so defending yourself against the robots as a customer adds to the overhead of choosing them. How do you know it's not going to cost you 10x as much to fix/update them as it cost you last year? You don't, because you were locked into a scammer who unilaterally decided to raise rates on you or the entire industry, or simply got in over their head and shut down, leaving you without a vendor. Minimum wage employees don't do that.

Now asking some dipshit McKinsey consultant or some social media expert for 10 ideas to increase sales? LLMs can give you 100, summarize them, and rank them with references. They are only getting better at this. We criticize them because they only give us a list of obvious answers, but the reason we hire recent graduates is for them to give us the list of obvious answers they learned while being educated in their specialty. "Creativity" gets assigned to whoever has the social capital to be seen as a legitimate and worthy "creative," or, more accurately, the person with the money either chooses somebody to run the show who they suspect has a magic secret formula or aura, or they take on that identity themselves.

That white collar middle-class magic is gone. People who could do difficult arithmetic quickly used to make a lot of money, too.

> No political force will fight for onshoring services jobs or incorporating barriers to reduce automation in most white collar jobs.

Only people who get protection would be the ones that are unified enough to have the institutions to launder cash to give to politicians. Doctors, realtors, etc.

I disagree. I personally think SWEs (who are what most people think of as tech bros), accountants, back-office roles, and others are at the upper economic tier of society now - as wage and educational data has clearly shown.

These are the roles most at risk for automation, not because jobs will be 100% automated, but because automation now exists to reduce headcount by 20-30% and still generate the same output. This is what already happened in skilled trades over the last 40 years.