My company writes software to manage youth soccer tournaments. We constantly get hammered to provide a “one button” push to poop out a schedule. Should be easy enough, right? Recently, I fed all the parameters for a sample division into ChatGPT and it popped out an entirely unusable mess. I kept tweeting the prompts and after about four hours of trying to ask it to do what I needed it to figure out and it kept giving me generic, really real-world unusable schedules, I gave up. I could have scheduled an entire 200 team tournament event in the time I wasted on AI on just one division!
I don’t think people who know how to schedule soccer tournaments are going anywhere soon. Maybe it’s too complex for AI, maybe the niche is just way too small for AI to bother learning ... maybe soccer scheduling is just a human skills by humans, for humans.
Or maybe we augment these system with some more general "reasoning" modules, such as for arithmetic, or SAT, temporal logic, etc ... and give it a little training on when and how to employ them. We might not need much soccer specific logic - I suspect many of the rules are probably general to scheduling broadly and small number to soccer in particular, and then we just need a sufficient memory to interact with a calendar.
As a software engineer, you probably be looking into not how to replace your system with a GPT variant but either to integrate with it - either as input, output or augment.
also just look at the facts: 1) its able to learn and improve itself. 2) there are people improving it. 3) the pace of improvements between GPT 2 until GPT 4........... (and their capabilities).... so.. what do you think will GPT 5 will look like (doesnt matter if it will be called that, it may be Meta who suddenly come out with something better, hope u all get it)
This all makes sense, but/and I do think that the concern about the stratification of society into people who can afford the luxury of human interaction is a huge concern. We’re already at heightened levels of social inequality, and AI seems poised to only push that further and harder.
Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
I do think software has "eaten the world" but mostly from within. Every large company has been revolutionized by software. You apply for jobs online. Grocery stores have internet-enabled cash registers that hook straight into inventory management systems. Clothing stores allow you to order clothes online and then make returns in person. Even the plumbing company has dispatch-management software and you pay on a tablet the plumber brings with them.
What software has not done is allowed for technology-only companies to *replace* companies in other sectors.
I imagine the exact same future for AI. Every single large company will have an AI strategy. Many will become much more efficient. But why should OpanAI open a storefront when Home Depot and Lowe's will both buy their products?
Because the incentives currently are all about capturing the most from the unsustainable global debt bubble and the dying globalism paradigm which has hit its apogee and it about to undergo phase-transition collapse.
Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
Very interesting post. It's worth noting that Marc Andreessen recently tweeted something about the impact of AI on labor, in which he came to more or less the same conclusion as you--that its impact on jobs was, at present, overstated. He argued the point on the basis of regulations: regulated industries (law, education, medicine, etc.) will be resistant to the allure of AI technologies, *even if* AI technology produces a superior product.
Until regulation is bankrupted. Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
None ... except maybe in some vague speculative sense. We'll start seeing the impact over the next 6 months ... but adoption was very low at the time the big tech layoffs started earlier this year. Early adopters playing around wasn't a factor. Margin was a factor ... if the macro environment was like 2021, copilot would have just been seen as a point of leverge demanding _further_ hiring and growth to adopt it and reap the benefits sooner.
Maybe compare the "new feature" output of the companies after the tech layoffs and see if there is a decline -- and if not, that is evidence copilot has increased the productivity of the remaining programmers. (Twitter service didn't seem to suffer - maybe it's just waste.)
In general, those who are good at doing digital work will be able to find digital ai assistants that allow them to be more productive. Especially programmers, but likely also report writers and middle managers.
Perhaps VPs or directors will be able to handle, with ai mgr support assistants, some 100 direct reports with 10 x 10-person team leaders who are NOT (quite) managers. Any and all such managers, or tech workers, who lose their jobs, will rapidly find some other job - but how seldom less than 80% of their prior pay? Likely fewer than 1% will be unable to find a median pay job (~$62k).
(Also, thanks for great talk with Arnold Kling just now, Monday night/ Tuesday in Slovakia).
I don't think I can endorse the "software is eating the world" critique. Software is progressing slowly and can't yet replace the human brain - let alone 8 billion human brains. However, these are clearly early days. Machines really are eating the world. They are just taking a while to digest some parts of it.
"elder care workers, physical therapists, psychologists, nurses, and workers in similar jobs should be insulated from AI-related disruption because most people are going to prefer human caregivers over robots." You mean that we will keep producing services that cannot have productivity increases, or not in the same way as we produce goods. So we can expect the goods sector's share of the economy getting smaller while the service sector's getting bigger because of the productivity gap. It means that in advanced economies the service sector is slowing down the whole economy because the service sector is growing thanks to labour expensive resource. How can we solve this conundrum in advanced economies?.
And unlike during the two major Agricultural revolutions (following the Dark Age and Black Death), the Industrial Revolutions, by now human civilization is at the apogee of a massive globalism debt bubble that will crater demand. And the billions are going to be in a race to bottom as they entered the tail-end, dying Industrial Revolution, e.g. the discussion I had recently about Mexico and note that fits my model of the massive, deeply embedded maladaptation inertia that will retard the world forward:
(c.f. §Universal Basic Income Is Unaffordable and §Wrong Diagnosis, Wrong Prescription)
Ditto even regulation can delay and exacerbate maladaptation. The egregious socialism is the problem, not the solution! But Westerners will not forsake socialism and debt willingly because w/o it they are just in race to the bottom with Chinese who will tolerate much more sacrifice.
(Jordan Peterson’s Full Testimony Before Congress)
3. Again that technology is not driving the dire outcome per se but economics is and deflationary spirals in the face of technological disruption juxtaposed against Wrong Prescription reactions at the government/societalcide level, are a prescription for snowball into Gini coefficient hell. Thus my rebuttals to the following are necessity (of the looming deflationary collapse) is the mother invention and 30% automation of 60% of jobs would still 18% unemployment if demand remains level but demand will also crater leading to a downward spiral in demand and that necessity of invention will bite evermore intensely in a spiral in the wrong direction. Also outsourcing has been the first salvo, but the catastrophe of Baumol's Cost Disease which sustained the West (enabled by the monetization of the Fed) is indicative of what is to ensue and consideration to when the Fed is no longer able to maintain control of this careening Titanic:
But that requires an economic paradigm by which the fruits can be distributed out to the population-at-large. When starting from extremely high interest rates and low global debt, with a fiat central banking system born in the Middle Ages, then humanity had bluesky to pursue.
(c.f. interest rate chart since inception of human civilization in Mesopotamia)
But we are starting from egregious maladaptation which will pursue resistance to adaptation (i.e. sustaining socialism) first. And choose rationing via carbon credits and CBDC enforcement, instead of a free market adaptation. Because the system is too highly ordered right now and the only path forward for nature (in its Second Law of Thermodynamics quest for inexorable advance of entropy aka disorder aka maximal distribution of uncertainty) is to first 1 Samuel 15 destroy the highly ordered morass to make way for the golden knowledge age which can flourish after that maelstrom smash-up.
See capitalists will see the deflationary spiral writing on the wall and pursue self-preservation and maximum return on capital investment, which robotics will presumably offer which can outstrip any double-digit rate of interest that the human civilization can not bear (as maladapted as it currently is to ZIRP).
Lol, Westerners below calling for another FDR pogrom at the apogee of ZIRP. These Westerners are so maladapted to socialism and financialization. They have no conceptualization the looming monetary reset.
Governments think the looming crisis will empower them with socialism-on-steriods, but Bitcoin is coming to disrupt them with a monetary reset and alternative to their CBDC fiat Hotel Californias (you can check in anytime you like, but you can never leave their looming capital controls). The Bitcoin ETFs are going to trap everyone and go “poof it’s gone.”
It’s likely true that AI will spawn the creation of many new highly-skilled, high tech jobs. But there will be relatively few (i.e. millions) who are qualified to fill those jobs (i.e. versus billions of cute bunny rabbit human protoplasm).
But the flaw in arguments that AI+robotics will only augment existing roles instead of eliminating a preponderance of roles, is (I argue) those roles only continued to exist because the global, consumerism debt bubble enabled them to.
I conjecture that In the ensuing deflationary monetary system (i.e. Bitcoin as the only trustworthy global unit-of-account), the priorities of those looking to accumulate wealth and capital will be diverted away from throwing socialist-propped-up protoplasm-blobs into the IPO-scam financialization money pit, in a mad race to capture a share of the Brrrr money printer, to instead of maximizing return-on-capital-investment which will be a much more cold, calculating position of risk mgmt and consistency of outcomes. The human laborer is a massively-embedded entitled liability, that will sue for his/her/its (pronoun entitled) fattening stipend, petition the government labor board, unionize, etc.. Humans these days are entitled and think they deserve community, work-life balance, etc.. Nature owes us nothing. We had our chance and the Lord warned us in 1 Samuel 8, and now the Lord (i.e. nature) cometh for payback as in 1 Samuel 15 (i.e. cull them all, even the babies and women).
“The rich rule over the poor, and the borrower is slave to the lender.” — Proverbs 22:7
Westerns fell in love with 30 year mortgages, socialism and egregiously inefficient suburbia. Now they will be harvested by the natural reversion of economic reality. The normalcy-bias, exponential-function-ignoring can has been kicked to the edge of the societalcide cliff. Goodbye social safety net, welcome to feudalism 2.0. China has no social safety net and the world will be competing with them, and India, etc.. A race to the bottom.
Yes eventually after the billions have been culled by the socialist governments those protoplasm-blobs will demand protect them from nature, there will be a new golden Knowledge Age. Covid was the first salvo in giving them what they demand, per 1 Samuel 15. But we first have to get from here all the way to there, which will involve perhaps decade(s) of tumult.
I wrote in my updates on my aforementioned published Tradingview idea, that the public wave which begins in 2033, will likely be nascent for decades if not hundred years or more, because of the significance of the 11145.6 6-6-6th ECM (Economic Confidence Model) cycle wave which is ending. That new public wave will be the NWO with a world government to replace the nation-states, which will self-immolate as their constituents demand socialism. Without the dollar empire and the servant class of Chinese to sustain debt-based globalism, the paradigm that ensues simply won’t be able to prop up such uneconomic paradigms.
Westerners are not at all prepared for the immovable brick wall they are about to slam face first into and have their noses rearranged to the back of their skulls. Their fingertips to be burned up to their armpits.
There will be a race to the bottom to remain relevant for most of humanity that was lulled to sleep by the globalism bubble and thus never transitioned to the exponentially advancing future. Humans don’t correctly perceive the pace of exponential function. The doubling-daily coverage of the lily pad is only 50% covered and then the next day the farmer is shocked it is fully covered in lilies.
For example, Hal Varian’s comment in aforelinked is likely correct, but only for those millions sufficiently skilled to remain relevant in the new economic and technological necessity that I posit ensues. Tim Bray hopes for a government socialism that can redistribute to sustain the humane environment, but this can not happen until AFTER the nation-states are torched. Only the future world government (perhaps after Bitcoin becomes incentives incompatible by 2141 when programmed mining reward is replaced by txn fees) that will be able to enforce such a global socialism (and that will end up being Biblical Revelation hell but I will be gone long before that anyway).
My reaction to Kevin Carson at al’s 3D printing artesian prediction is indeed to some extent but bespoke is inefficient and the demand will only be from the wealthy. Westerners have a mindset of abundance of finance because they are maladapted to the globalism debt bubble. The vast majority will be in a race to the bottom. The globe can not return to homeostasis until perhaps the Georgia Guidestones’ 50 to 100 million population goal is achieved. Or sufficient percentage of the population has retrained in the new high-valued added knowledge age skills, which ever comes first.
Agreed technology is not deterministic of future. But economics is. Unfortunately not that many people have realistic understanding of economics and their cycles. Refer back to the 6-6-6th cycle which can be researched at my aforelinked published idea.
As I started to explain on my updates *today* on my recently published idea, the author of this blog is I argue blinded by too short of a look-back window time horizon. Suggest readers peruse the following linked:
I posit the brick&mortar world has resisted disruption only because the Fed propped up the global debt bubble to enable that misallocation and maladaption to continue to even more egregiously maladapted levels. Which I conjecture is just going to make it that much more difficult for maladapted society to adapt quickly when the sudden “phase transition” (e.g. the discontinuity as heated water suddenly transitions to steam) ensues with a sudden global collapse, probably in taking the form of WWIII circa late ~2025.
NECESSITY IS THE MOTHER OF INVENTION. The technological disruption hasn’t been absolutely necessary yet, but the technological “molecules“ are heating up and getting aligned for a sudden, face-ripping, neck-breaking, whirlwind waterfall transition.
I will respond to some specific points in Tim’s much appreciated and eloquent enumeration:
1. Bitcoin will not lurch to the global monetary standard replacing the dollar until after the $30+ trillion EuroDollar market (i.e. dollar denominated loans outside the Fed’s banking system control) defaults creating scorched earth. The ANYONECANSPEND destruction of the impostor Bitcoin likely has to transpire first as well (the Bitcoin everyone is using will be destroyed and legacy will rise as the 1988 Economist Magazine’s cover story Phoenix). So do not make any judgements about Bitcon yet, as it is too early. All the plebs must be kicked off Bitcoin first, and the transaction rate returned to the 7 txns per second limit, so that only $billionaires can use Bitcoin. Plebs will be enslaved in the looming two-tier CBDCs system. nobody will trust any centralized monetary system ever again after the looming devastation the EuroDollar collapse is going to cause. Dollar will become very strong one final time as the world implodes in a Greater Depression and war. The West is self-immolating as we observe accelerating approaching a phase transition to civil unrest mayhem. World is about to be turned upside-down. Boomers are culling themselves (jabs, begging for war with Pootin, etc). All their normalcy biases (e.g. towards time-honored institutions) will eviscerated. Bitcoin isn’t eating the banking system per se, it’s not about retail, but instead it’s eating the monetary system.
Bitcoin is not here to help you. It is here to enslave you. Wait I will cite the whale with 500,000 BTC (http://trilema.com/2016/mpex-smpoe-closing-statement/) who carried out the Dao attack on Ethereum (http://trilema.com/2016/to-the-dao-and-the-ethereum-community-fuck-you/) on this. You do not seem to factor in that the world that comes after this global smashup, is a broken, multipolar world, where in some places warlords rule. Where capitalists rule, and the Gini coefficient skyrockets. You are either rich or you are useless.
ALL FUNGIBLE RESOURCES ARE POWER-LAW DISTRIBUTED. Let that fact sink in.
I also hope for the brighter version of the future, but we have so much work to do, and that is not a given. The future looks very precarious to me.
2. The Fed is the only thing standing between plumbers being dime-a-dozen (e.g. 1000 plumbers competing for every plumbing job) and the current situation where entitled people do not want to work. That entire monetary system paradigm is dying. The peace dividend of the fall of Communism has already reached its apogee. China can’t continue absorbing the West’s debt bubble so they must soon devalue their currency, import inflation and export global deflation. Being at the end of their globalism rope, is why they have been cooperating with the Western elites to trap Russia and foster the WWIII the CCP needs as a scapegoat to hold onto power. But the nation-state is going to be torn to shreds over the next century. I posit that the CCP has a few decades or so before they are disrupted as well, e.g. the demographic time bomb for China by 2040. Also the race to the bottom for economies that are trying to compete to the world’s manufacturer— Industrial Age is dying.
3. Writing about the sharing economy is myopic when the global economy is predominantly about the financial economy and the Fed. Capital flows are orders-of-magnitude more significant than trade flows. The significant theme is what does a world look like without a Fed backstop? With the wrecking ball of hard money in control?
And how does deflationary scorched globe focus resources when there is no more backstop? Let’s hearken back to the Dark Ages when usury was banned in Europe and human labor was in such excess supply that West of the Hajnal line females delayed child birth to their late 20s and it was not even productive to breed stronger horses, because humans cost less to hire to plow than horses. The Black Death ostensibly reduced the population enough to stimulate the Agricultural revolutions. The abolition of (or preliminary work arounds for) usury was necessary to provide the finance for the consequent expanding economies.
But in a world where robots can manufacture robots, thus so capital multiplies itself faster than any feasible rate of interest, capital will concentrate and those left behind will fall further and further behind. Massive social discontent, but what will they do to defeat the robot armies of the capitalists to demand a redistribution of wealth?
I conjecture that the era of RobinHood returns along with feudalism 2.0.
Timothy B Lee: “And most of the American economy is not information-focused: It’s focused on delivering physical goods and services like homes, cars, restaurant meals, and haircuts.”
That’s all propped up by the Fed. But not for much longer. What remains after?
“Human workers will still retain another big advantage in the marketplace: the fact that other people like to interact with them.”
Who cares what they like if the Fed is no longer able to enable them to afford it.
“For example, in the future there will probably be low-cost virtual schools where students interact with chatbot tutors rather than human teachers.”
And that will be the only affordable option. Nobody will be handing out student loans, except maybe for a while the soon-to-be scorched earth socialist governments may try to hand out dystopian CBDCs forcing students to choose unionized educational tarpits. And all those who choose that fork in the road will self-immolate.
The ATM example is especially pertinent because luxury is only a choice for those who aren’t impoverished. I suggest you check out banking in the Philippines for a glimpse of the Western future.
“For example, it’s very hard to imagine we are ever going to replace human judges with AI.”
With the kangaroo courts deconstructing Western civilization, most certainly we are going to never again trust humans in positions that can be corrupted.
“If AI starts replacing workers in the coming years, that will put downward pressure on wages and prices while growing the economic pie. That will give the Fed more leeway to cut interest rates and give Congress more room to raise spending or cut taxes.”
C.f. my discussion of John Nash’s proscription for the future (e.g. Bitcoin as Ideal Money) in the aforementioned updates on my published idea.
The dollar reserve currency system will be dead by 2032. This is the 6-6-6th wave comprising the end of a 11145.6 year cycle.
Does Timothy B Lee realize what time it is? He had better pay attention.
1) Human population will soon decline. This trend will possibly accelerate thanks to AI. That's because as you say, humans will always need human interactions, while AGI won't. So, thinking of your convincing idea of a potential two-tier society, where the lower tier will interact much more often with robots, the more we replace humans with AI/robots the less humans will procreate.
2) It's clear that plumbers will be the last jobs to be taken over by AI/machines. It is less clear to me that a white-collar person like you and me will love to drop their current job and start repairing sinks. It's just not what we chose to do and we would probably be bad at that too.
3) ATM example: I can't see any company in today's environment / business-mindset following that example. Simplifying it badly, if you save costs, you either distribute the savings to shareholders or invest in productive business development, but you don't reduce the margins of your slowly shrinking cash cow business.
4) What is missing here is the recognition of the hyperbolic speed and compounded nature of tech's evolution: "But in material terms, our lives aren’t much different than they were 30 years ago". That might be true (I disagree) but from a tech point of view, we are in a very different place today than 30 years ago. Not only the tech stack is much more advanced, but the speed of tech development is orders of magnitude faster.
Concluding, I can't predict the future, but it definitively won't be a repetition of the past or a rhyme of it. Humans have reached the tech inflection point, the point of no return.
I’ve been thinking about this a bit lately and my take has been much the same. I use Github Copilot daily and it’s been a definite boon to productivity - Copilot knows what I’m doing, knows where I’m going with a block of code and is super quick to offer suggestions on ways to not only achieve it but make it better. But would the end product be better if Copilot just wrote the program? Nah, maybe someday but not today.
I suspect you're a good programmer - and your increasing productivity in writing good code will reduce your company's need to hire new programmers, and may allow the company to lay some off.
If Copilot writes a functioning program, debugged and ready to ship - in one day- that should allow multiple iterations of user improvements. But it's still pretty far away from that level.
I found it useful for adopting and integrating _new_ technology. I need to integrate third-party thing I know next to nothing about: generate an example with that matches roughly what I need to do. Show me how to change the things I care about without me needing to traverse tons of docs that presume background knowledge from other docs (or worse no docs and investigating implementation-code - or even worse wrong/bad docs). My time to come up to speed a get something going is reduced from a couple days to minutes. It doesn't really help with the whole system, but the "I could sorta want to do something like X, using toolset Y and further refine with details Z" is super powerful. Plenty of people use tools like stackoverflow when they get stuck ... but that still has latency of hours/days. If you get those questions resolved in seconds, its remove a lot of idle time. And that is value today ... in the future I imagine it will be able to do more on the architecture/ whole-project side of things.
It's very far away from that, to be honest. While it's great to help me keep from having to type and re-type typos and it's great for remembering the variable that should go in this logic, it's not all that great for full blocks of code beyond maybe a conditional block or a function that resembles one you've already worked on. I get your point about it increasing productivity to where some programmers get laid off. I don't know that it would really work that way but I will allow that it might, in quite some time.
Loved how you described the position in so much detail. I wrote something similar on my Substack - about whether AI will replace legal jobs, again with a resounding NO.
This fear mongering is as common as boom and bust cycles.
I didn't find the "robotics is hard" section very compelling. You give a few anecdotes about specific robotics projects that have progressed slowly, but you don't really explain the *why*.
All the newer tools (large GPU clusters, SGD, dropout, pre-training, transformers, relu units, etc. etc.) have all been applied to reinforcement learning problems, with some success. So why do these problems still appear so much harder?
I reckon it has something to do with the difficulty of producing an adequate physical model to perform training in a simulated setting, and the high cost of tuning the physical model (or training the whole network) on input from actual physical sensors.
I'm not saying it's hard to build robotics software (though it might be). I'm saying it's hard to build the actual robots. The human hand for example is an engineering marvel—it has incredible degrees of freedom, has highly sensitive sensors, and it can work for decades without need for repair or maintenance. I don't think we're anywhere close to figuring out how to build something with the same combination of strength, weight, sensitivity, and durability, to say nothing of its ability to self-repair after minor damage.
I suspect it may be impossible to build a robot with our self-maintenance abilities using materials like steel and plastic. That obviously doesn't mean we can't eventually build an ecosystem of robots that are able to build and repair one another, but I think it's going to take a long time to build up that ecosystem. In the shot-to-medium turn, every robot we build will mean new jobs for people to maintain and repair it when it breaks.
Yes robotics is hard. I worked as an automation engineer for automotive plants for a few years. Robots there are powerful but pretty dumb. You have to enclose them in cages so they don’t kill people. However if you want to reliably spot weld doors together you really can’t beat it. No human wants that job anyway.
I was a coach/mentor for our high school FIRST robotics team. The competition has the kids assemble and program a mobile robot to do some basic tasks. The task we had in 2018 was to pick up 2 foot square cubes and place them on a balancing scale. Even that well defined task proved to be quite difficult in the “fully autonomous” part of the competition.
The FIRST robotics competition is divided into 2 parts. The first 15 seconds is the “fully autonomous’ phase. In that phase your robot needs to run autonomously in the competition area and try and accomplish the assigned task. The rest of the completion has a human remote controlling the robot for 3 minutes. That is much more productive and fun. I can see the AI revolution taking that turn. Instead of replacing humans it will give them super powers like the Iron Man suit.
But what level of proficiency is needed to displace a given percentage of jobs? We don’t necessarily need robots to be as dexterous as humans in order to for example displace 30% of 60% of jobs, which is an 18% unemployment rate presuming demand is level. What if demand is collapsing due to an imploding global debt bubble and the apogee of the bankrupt paradigm of globalism?
Machines have been steadily displacing workers for 150 years, yet unemployment is currently near record lows because demand isn’t level. As the economy gets more efficient people consume more.
Demand isn’t level because the Fed is propping up a massive globalism bubble that is going to self-immolate and leave your mouth agape within this decade. Entitled Westerners do not want to work for peanuts. No surprises there, when the government is handing out fiscal debt spending like there is no tomorrow nor reckoning ever coming. The reckoning is ~2026ish in earnest.
In the early 1980s I was evaluating one of a number of "create code from English Language" products (I think it was called Rescue); older colleagues predicted then that I wouldn't have much of a software career before these new products took over ... I'm still creating software nearly 40 years later ...
My company writes software to manage youth soccer tournaments. We constantly get hammered to provide a “one button” push to poop out a schedule. Should be easy enough, right? Recently, I fed all the parameters for a sample division into ChatGPT and it popped out an entirely unusable mess. I kept tweeting the prompts and after about four hours of trying to ask it to do what I needed it to figure out and it kept giving me generic, really real-world unusable schedules, I gave up. I could have scheduled an entire 200 team tournament event in the time I wasted on AI on just one division!
I don’t think people who know how to schedule soccer tournaments are going anywhere soon. Maybe it’s too complex for AI, maybe the niche is just way too small for AI to bother learning ... maybe soccer scheduling is just a human skills by humans, for humans.
Or maybe we augment these system with some more general "reasoning" modules, such as for arithmetic, or SAT, temporal logic, etc ... and give it a little training on when and how to employ them. We might not need much soccer specific logic - I suspect many of the rules are probably general to scheduling broadly and small number to soccer in particular, and then we just need a sufficient memory to interact with a calendar.
As a software engineer, you probably be looking into not how to replace your system with a GPT variant but either to integrate with it - either as input, output or augment.
and maybe it wasnt gpt 4?
also just look at the facts: 1) its able to learn and improve itself. 2) there are people improving it. 3) the pace of improvements between GPT 2 until GPT 4........... (and their capabilities).... so.. what do you think will GPT 5 will look like (doesnt matter if it will be called that, it may be Meta who suddenly come out with something better, hope u all get it)
This all makes sense, but/and I do think that the concern about the stratification of society into people who can afford the luxury of human interaction is a huge concern. We’re already at heightened levels of social inequality, and AI seems poised to only push that further and harder.
Yes that's a legitimate concern for sure!
Not just AI, but the entire monetary system and globalism paradigm being bankrupt.
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51268748
Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
I will stop spamming the comments now.
I do think software has "eaten the world" but mostly from within. Every large company has been revolutionized by software. You apply for jobs online. Grocery stores have internet-enabled cash registers that hook straight into inventory management systems. Clothing stores allow you to order clothes online and then make returns in person. Even the plumbing company has dispatch-management software and you pay on a tablet the plumber brings with them.
What software has not done is allowed for technology-only companies to *replace* companies in other sectors.
I imagine the exact same future for AI. Every single large company will have an AI strategy. Many will become much more efficient. But why should OpanAI open a storefront when Home Depot and Lowe's will both buy their products?
Yes agree 100 percent!
I posit you are deceived by a normalcy bias whilst the exponential function is accelerating on your myopia.
Because the incentives currently are all about capturing the most from the unsustainable global debt bubble and the dying globalism paradigm which has hit its apogee and it about to undergo phase-transition collapse.
Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51268748
Very interesting post. It's worth noting that Marc Andreessen recently tweeted something about the impact of AI on labor, in which he came to more or less the same conclusion as you--that its impact on jobs was, at present, overstated. He argued the point on the basis of regulations: regulated industries (law, education, medicine, etc.) will be resistant to the allure of AI technologies, *even if* AI technology produces a superior product.
Until regulation is bankrupted. Humans can’t perceive the exponential function. On the 29th day the lily pond is half-covered. The next morning the farmer is shocked that it’s entirely covered.
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51268748
What’s the best evidence the tech layoffs are related to copilot?
None ... except maybe in some vague speculative sense. We'll start seeing the impact over the next 6 months ... but adoption was very low at the time the big tech layoffs started earlier this year. Early adopters playing around wasn't a factor. Margin was a factor ... if the macro environment was like 2021, copilot would have just been seen as a point of leverge demanding _further_ hiring and growth to adopt it and reap the benefits sooner.
Maybe compare the "new feature" output of the companies after the tech layoffs and see if there is a decline -- and if not, that is evidence copilot has increased the productivity of the remaining programmers. (Twitter service didn't seem to suffer - maybe it's just waste.)
In general, those who are good at doing digital work will be able to find digital ai assistants that allow them to be more productive. Especially programmers, but likely also report writers and middle managers.
Perhaps VPs or directors will be able to handle, with ai mgr support assistants, some 100 direct reports with 10 x 10-person team leaders who are NOT (quite) managers. Any and all such managers, or tech workers, who lose their jobs, will rapidly find some other job - but how seldom less than 80% of their prior pay? Likely fewer than 1% will be unable to find a median pay job (~$62k).
(Also, thanks for great talk with Arnold Kling just now, Monday night/ Tuesday in Slovakia).
Hard to say! I will be thinking about this for sure.
I don't think I can endorse the "software is eating the world" critique. Software is progressing slowly and can't yet replace the human brain - let alone 8 billion human brains. However, these are clearly early days. Machines really are eating the world. They are just taking a while to digest some parts of it.
"elder care workers, physical therapists, psychologists, nurses, and workers in similar jobs should be insulated from AI-related disruption because most people are going to prefer human caregivers over robots." You mean that we will keep producing services that cannot have productivity increases, or not in the same way as we produce goods. So we can expect the goods sector's share of the economy getting smaller while the service sector's getting bigger because of the productivity gap. It means that in advanced economies the service sector is slowing down the whole economy because the service sector is growing thanks to labour expensive resource. How can we solve this conundrum in advanced economies?.
There's a name for this, Baumol's Cost Disease. And the answer is you probably can't.
So it's not sofware nor A.I. that is eating the modern world but the service sector suffering from Baumol's Cost Disease.
One final salvo…
Again to recap so the central thrust of my arguments are not lost in the weeds of all I wrote.
1. Whether technology spawns new jobs or focused on eliminating roles is dependent on how saturated the market is already:
https://sites.bu.edu/tpri/2017/07/06/why-isnt-automation-creating-unemployment/
https://builtin.com/artificial-intelligence/ai-workers-jobs
And unlike during the two major Agricultural revolutions (following the Dark Age and Black Death), the Industrial Revolutions, by now human civilization is at the apogee of a massive globalism debt bubble that will crater demand. And the billions are going to be in a race to bottom as they entered the tail-end, dying Industrial Revolution, e.g. the discussion I had recently about Mexico and note that fits my model of the massive, deeply embedded maladaptation inertia that will retard the world forward:
https://www.youtube.com/watch?v=B1CC97WuM6w&lc=Ugyarb1WBxAJOzFK40Z4AaABAg.9x0lkCcou6EA0QlL11grDW
(find my hopefully not shadow-banned comments by @SMoore-vj7bt)
Note I love Mexico, so please do not anyone take those comments as derogatory in any way.
2. The reaction of humanity will exacerbate instead of adapt, because maladapted systems self-immolate instead of adapt, as they can’t:
https://fee.org/articles/no-robots-wont-make-us-all-unemployed/
(c.f. §Universal Basic Income Is Unaffordable and §Wrong Diagnosis, Wrong Prescription)
Ditto even regulation can delay and exacerbate maladaptation. The egregious socialism is the problem, not the solution! But Westerners will not forsake socialism and debt willingly because w/o it they are just in race to the bottom with Chinese who will tolerate much more sacrifice.
https://www.fastcompany.com/90906438/why-mass-unemployment-worry-sentient-ai
https://www.youtube.com/watch?v=KdhHO9HBZvc
(Jordan Peterson’s Full Testimony Before Congress)
3. Again that technology is not driving the dire outcome per se but economics is and deflationary spirals in the face of technological disruption juxtaposed against Wrong Prescription reactions at the government/societalcide level, are a prescription for snowball into Gini coefficient hell. Thus my rebuttals to the following are necessity (of the looming deflationary collapse) is the mother invention and 30% automation of 60% of jobs would still 18% unemployment if demand remains level but demand will also crater leading to a downward spiral in demand and that necessity of invention will bite evermore intensely in a spiral in the wrong direction. Also outsourcing has been the first salvo, but the catastrophe of Baumol's Cost Disease which sustained the West (enabled by the monetization of the Fed) is indicative of what is to ensue and consideration to when the Fed is no longer able to maintain control of this careening Titanic:
https://www.reddit.com/r/changemyview/comments/s2nmbw/cmv_ai_and_automation_will_not_lead_to_massive/
4. True that automation can increase overall production and thus human prosperity:
https://archive.today/https://medium.com/swlh/no-ai-wont-lead-to-mass-unemployment-67c38a3ddb79
But that requires an economic paradigm by which the fruits can be distributed out to the population-at-large. When starting from extremely high interest rates and low global debt, with a fiat central banking system born in the Middle Ages, then humanity had bluesky to pursue.
https://www.armstrongeconomics.com/research/a-brief-history-of-world-credit-interest-rates/3000-b-c-500-a-d-the-ancient-economy/
(c.f. interest rate chart since inception of human civilization in Mesopotamia)
But we are starting from egregious maladaptation which will pursue resistance to adaptation (i.e. sustaining socialism) first. And choose rationing via carbon credits and CBDC enforcement, instead of a free market adaptation. Because the system is too highly ordered right now and the only path forward for nature (in its Second Law of Thermodynamics quest for inexorable advance of entropy aka disorder aka maximal distribution of uncertainty) is to first 1 Samuel 15 destroy the highly ordered morass to make way for the golden knowledge age which can flourish after that maelstrom smash-up.
See capitalists will see the deflationary spiral writing on the wall and pursue self-preservation and maximum return on capital investment, which robotics will presumably offer which can outstrip any double-digit rate of interest that the human civilization can not bear (as maladapted as it currently is to ZIRP).
Lol, Westerners below calling for another FDR pogrom at the apogee of ZIRP. These Westerners are so maladapted to socialism and financialization. They have no conceptualization the looming monetary reset.
https://www.linkedin.com/pulse/why-robotics-cause-mass-unemployment-part-2-mark-fairchild
Governments think the looming crisis will empower them with socialism-on-steriods, but Bitcoin is coming to disrupt them with a monetary reset and alternative to their CBDC fiat Hotel Californias (you can check in anytime you like, but you can never leave their looming capital controls). The Bitcoin ETFs are going to trap everyone and go “poof it’s gone.”
https://www.reddit.com/r/ChatGPT/comments/11t0lz7/why_arent_governments_afraid_that_ai_will_create/
https://anonymint1.substack.com/i/137748629/what-wont-work
https://youtu.be/-DT7bX-B1Mg
(And it's gone)
I’ll deconstruct another article on this topic of our labor future.
https://www.pewresearch.org/internet/2014/08/06/future-of-jobs/
It’s likely true that AI will spawn the creation of many new highly-skilled, high tech jobs. But there will be relatively few (i.e. millions) who are qualified to fill those jobs (i.e. versus billions of cute bunny rabbit human protoplasm).
But the flaw in arguments that AI+robotics will only augment existing roles instead of eliminating a preponderance of roles, is (I argue) those roles only continued to exist because the global, consumerism debt bubble enabled them to.
I conjecture that In the ensuing deflationary monetary system (i.e. Bitcoin as the only trustworthy global unit-of-account), the priorities of those looking to accumulate wealth and capital will be diverted away from throwing socialist-propped-up protoplasm-blobs into the IPO-scam financialization money pit, in a mad race to capture a share of the Brrrr money printer, to instead of maximizing return-on-capital-investment which will be a much more cold, calculating position of risk mgmt and consistency of outcomes. The human laborer is a massively-embedded entitled liability, that will sue for his/her/its (pronoun entitled) fattening stipend, petition the government labor board, unionize, etc.. Humans these days are entitled and think they deserve community, work-life balance, etc.. Nature owes us nothing. We had our chance and the Lord warned us in 1 Samuel 8, and now the Lord (i.e. nature) cometh for payback as in 1 Samuel 15 (i.e. cull them all, even the babies and women).
“The rich rule over the poor, and the borrower is slave to the lender.” — Proverbs 22:7
Westerns fell in love with 30 year mortgages, socialism and egregiously inefficient suburbia. Now they will be harvested by the natural reversion of economic reality. The normalcy-bias, exponential-function-ignoring can has been kicked to the edge of the societalcide cliff. Goodbye social safety net, welcome to feudalism 2.0. China has no social safety net and the world will be competing with them, and India, etc.. A race to the bottom.
Yes eventually after the billions have been culled by the socialist governments those protoplasm-blobs will demand protect them from nature, there will be a new golden Knowledge Age. Covid was the first salvo in giving them what they demand, per 1 Samuel 15. But we first have to get from here all the way to there, which will involve perhaps decade(s) of tumult.
https://steemit.com/science/@anonymint/the-golden-knowledge-age-is-rising
I wrote in my updates on my aforementioned published Tradingview idea, that the public wave which begins in 2033, will likely be nascent for decades if not hundred years or more, because of the significance of the 11145.6 6-6-6th ECM (Economic Confidence Model) cycle wave which is ending. That new public wave will be the NWO with a world government to replace the nation-states, which will self-immolate as their constituents demand socialism. Without the dollar empire and the servant class of Chinese to sustain debt-based globalism, the paradigm that ensues simply won’t be able to prop up such uneconomic paradigms.
Westerners are not at all prepared for the immovable brick wall they are about to slam face first into and have their noses rearranged to the back of their skulls. Their fingertips to be burned up to their armpits.
http://trilema.com/2013/the-story-of-pointless-and-witless/
There will be a race to the bottom to remain relevant for most of humanity that was lulled to sleep by the globalism bubble and thus never transitioned to the exponentially advancing future. Humans don’t correctly perceive the pace of exponential function. The doubling-daily coverage of the lily pad is only 50% covered and then the next day the farmer is shocked it is fully covered in lilies.
For example, Hal Varian’s comment in aforelinked is likely correct, but only for those millions sufficiently skilled to remain relevant in the new economic and technological necessity that I posit ensues. Tim Bray hopes for a government socialism that can redistribute to sustain the humane environment, but this can not happen until AFTER the nation-states are torched. Only the future world government (perhaps after Bitcoin becomes incentives incompatible by 2141 when programmed mining reward is replaced by txn fees) that will be able to enforce such a global socialism (and that will end up being Biblical Revelation hell but I will be gone long before that anyway).
My reaction to Kevin Carson at al’s 3D printing artesian prediction is indeed to some extent but bespoke is inefficient and the demand will only be from the wealthy. Westerners have a mindset of abundance of finance because they are maladapted to the globalism debt bubble. The vast majority will be in a race to the bottom. The globe can not return to homeostasis until perhaps the Georgia Guidestones’ 50 to 100 million population goal is achieved. Or sufficient percentage of the population has retrained in the new high-valued added knowledge age skills, which ever comes first.
Agreed technology is not deterministic of future. But economics is. Unfortunately not that many people have realistic understanding of economics and their cycles. Refer back to the 6-6-6th cycle which can be researched at my aforelinked published idea.
As I started to explain on my updates *today* on my recently published idea, the author of this blog is I argue blinded by too short of a look-back window time horizon. Suggest readers peruse the following linked:
https://www.tradingview.com/chart/DJI/UPT3GBm7-U-S-dies-by-financial-repression-then-hyperinflation/
I posit the brick&mortar world has resisted disruption only because the Fed propped up the global debt bubble to enable that misallocation and maladaption to continue to even more egregiously maladapted levels. Which I conjecture is just going to make it that much more difficult for maladapted society to adapt quickly when the sudden “phase transition” (e.g. the discontinuity as heated water suddenly transitions to steam) ensues with a sudden global collapse, probably in taking the form of WWIII circa late ~2025.
NECESSITY IS THE MOTHER OF INVENTION. The technological disruption hasn’t been absolutely necessary yet, but the technological “molecules“ are heating up and getting aligned for a sudden, face-ripping, neck-breaking, whirlwind waterfall transition.
I will respond to some specific points in Tim’s much appreciated and eloquent enumeration:
1. Bitcoin will not lurch to the global monetary standard replacing the dollar until after the $30+ trillion EuroDollar market (i.e. dollar denominated loans outside the Fed’s banking system control) defaults creating scorched earth. The ANYONECANSPEND destruction of the impostor Bitcoin likely has to transpire first as well (the Bitcoin everyone is using will be destroyed and legacy will rise as the 1988 Economist Magazine’s cover story Phoenix). So do not make any judgements about Bitcon yet, as it is too early. All the plebs must be kicked off Bitcoin first, and the transaction rate returned to the 7 txns per second limit, so that only $billionaires can use Bitcoin. Plebs will be enslaved in the looming two-tier CBDCs system. nobody will trust any centralized monetary system ever again after the looming devastation the EuroDollar collapse is going to cause. Dollar will become very strong one final time as the world implodes in a Greater Depression and war. The West is self-immolating as we observe accelerating approaching a phase transition to civil unrest mayhem. World is about to be turned upside-down. Boomers are culling themselves (jabs, begging for war with Pootin, etc). All their normalcy biases (e.g. towards time-honored institutions) will eviscerated. Bitcoin isn’t eating the banking system per se, it’s not about retail, but instead it’s eating the monetary system.
Bitcoin is not here to help you. It is here to enslave you. Wait I will cite the whale with 500,000 BTC (http://trilema.com/2016/mpex-smpoe-closing-statement/) who carried out the Dao attack on Ethereum (http://trilema.com/2016/to-the-dao-and-the-ethereum-community-fuck-you/) on this. You do not seem to factor in that the world that comes after this global smashup, is a broken, multipolar world, where in some places warlords rule. Where capitalists rule, and the Gini coefficient skyrockets. You are either rich or you are useless.
ALL FUNGIBLE RESOURCES ARE POWER-LAW DISTRIBUTED. Let that fact sink in.
http://trilema.com/2015/lets-address-some-of-the-more-common-pseudo-arguments-raised-by-the-very-stupid-people-that-like-the-gavin-scamcoin-proposal/#selection-107.0-107.452
http://trilema.com/2015/lets-address-some-of-the-more-common-pseudo-arguments-raised-by-the-very-stupid-people-that-like-the-gavin-scamcoin-proposal/#selection-207.0-213.539
http://trilema.com/2015/lets-address-even-more-of-the-more-common-pseudo-arguments-raised-by-the-very-stupid-people-that-like-the-gavin-scamcoin-proposal/#selection-153.0-157.521
http://trilema.com/2015/third-pass-addressing-the-more-common-pseudo-arguments-raised-by-the-very-stupid-people-that-like-the-gavin-scamcoin-proposal/#selection-127.0-141.300
http://trilema.com/2014/the-sins-of-the-group-of-posers-behind-the-so-called-bitcoin-foundation/
http://trilema.com/2014/the-bitcoin-foundation-finally-incorporated/
http://trilema.com/2015/theres-a-one-bitcoin-reward-for-the-death-of-pieter-wuille-details-below/
http://trilema.com/2015/ok-so-what-is-bitcoin-disrupting/
I also hope for the brighter version of the future, but we have so much work to do, and that is not a given. The future looks very precarious to me.
2. The Fed is the only thing standing between plumbers being dime-a-dozen (e.g. 1000 plumbers competing for every plumbing job) and the current situation where entitled people do not want to work. That entire monetary system paradigm is dying. The peace dividend of the fall of Communism has already reached its apogee. China can’t continue absorbing the West’s debt bubble so they must soon devalue their currency, import inflation and export global deflation. Being at the end of their globalism rope, is why they have been cooperating with the Western elites to trap Russia and foster the WWIII the CCP needs as a scapegoat to hold onto power. But the nation-state is going to be torn to shreds over the next century. I posit that the CCP has a few decades or so before they are disrupted as well, e.g. the demographic time bomb for China by 2040. Also the race to the bottom for economies that are trying to compete to the world’s manufacturer— Industrial Age is dying.
3. Writing about the sharing economy is myopic when the global economy is predominantly about the financial economy and the Fed. Capital flows are orders-of-magnitude more significant than trade flows. The significant theme is what does a world look like without a Fed backstop? With the wrecking ball of hard money in control?
https://steemit.com/money/@anonymint/rise-of-hard-money-is-a-harbinger-of-misery
And how does deflationary scorched globe focus resources when there is no more backstop? Let’s hearken back to the Dark Ages when usury was banned in Europe and human labor was in such excess supply that West of the Hajnal line females delayed child birth to their late 20s and it was not even productive to breed stronger horses, because humans cost less to hire to plow than horses. The Black Death ostensibly reduced the population enough to stimulate the Agricultural revolutions. The abolition of (or preliminary work arounds for) usury was necessary to provide the finance for the consequent expanding economies.
But in a world where robots can manufacture robots, thus so capital multiplies itself faster than any feasible rate of interest, capital will concentrate and those left behind will fall further and further behind. Massive social discontent, but what will they do to defeat the robot armies of the capitalists to demand a redistribution of wealth?
I conjecture that the era of RobinHood returns along with feudalism 2.0.
http://esr.ibiblio.org/?p=8085
(The return of the servant problem)
Timothy B Lee: “And most of the American economy is not information-focused: It’s focused on delivering physical goods and services like homes, cars, restaurant meals, and haircuts.”
That’s all propped up by the Fed. But not for much longer. What remains after?
“Human workers will still retain another big advantage in the marketplace: the fact that other people like to interact with them.”
Who cares what they like if the Fed is no longer able to enable them to afford it.
“For example, in the future there will probably be low-cost virtual schools where students interact with chatbot tutors rather than human teachers.”
And that will be the only affordable option. Nobody will be handing out student loans, except maybe for a while the soon-to-be scorched earth socialist governments may try to hand out dystopian CBDCs forcing students to choose unionized educational tarpits. And all those who choose that fork in the road will self-immolate.
The ATM example is especially pertinent because luxury is only a choice for those who aren’t impoverished. I suggest you check out banking in the Philippines for a glimpse of the Western future.
http://trilema.com/2012/bitcoin-and-the-poor/#selection-77.0-77.867
“For example, it’s very hard to imagine we are ever going to replace human judges with AI.”
With the kangaroo courts deconstructing Western civilization, most certainly we are going to never again trust humans in positions that can be corrupted.
“If AI starts replacing workers in the coming years, that will put downward pressure on wages and prices while growing the economic pie. That will give the Fed more leeway to cut interest rates and give Congress more room to raise spending or cut taxes.”
C.f. my discussion of John Nash’s proscription for the future (e.g. Bitcoin as Ideal Money) in the aforementioned updates on my published idea.
The dollar reserve currency system will be dead by 2032. This is the 6-6-6th wave comprising the end of a 11145.6 year cycle.
Does Timothy B Lee realize what time it is? He had better pay attention.
It likely resolves itself in a massive catastrophe, c.f. my longish comment for a posited scenario.
I'd note 4 things and conclude:
1) Human population will soon decline. This trend will possibly accelerate thanks to AI. That's because as you say, humans will always need human interactions, while AGI won't. So, thinking of your convincing idea of a potential two-tier society, where the lower tier will interact much more often with robots, the more we replace humans with AI/robots the less humans will procreate.
2) It's clear that plumbers will be the last jobs to be taken over by AI/machines. It is less clear to me that a white-collar person like you and me will love to drop their current job and start repairing sinks. It's just not what we chose to do and we would probably be bad at that too.
3) ATM example: I can't see any company in today's environment / business-mindset following that example. Simplifying it badly, if you save costs, you either distribute the savings to shareholders or invest in productive business development, but you don't reduce the margins of your slowly shrinking cash cow business.
4) What is missing here is the recognition of the hyperbolic speed and compounded nature of tech's evolution: "But in material terms, our lives aren’t much different than they were 30 years ago". That might be true (I disagree) but from a tech point of view, we are in a very different place today than 30 years ago. Not only the tech stack is much more advanced, but the speed of tech development is orders of magnitude faster.
Concluding, I can't predict the future, but it definitively won't be a repetition of the past or a rhyme of it. Humans have reached the tech inflection point, the point of no return.
What about those concerned with AI existential risk like your old colleague Kelsey Piper?
Stay tuned!
I’ve been thinking about this a bit lately and my take has been much the same. I use Github Copilot daily and it’s been a definite boon to productivity - Copilot knows what I’m doing, knows where I’m going with a block of code and is super quick to offer suggestions on ways to not only achieve it but make it better. But would the end product be better if Copilot just wrote the program? Nah, maybe someday but not today.
I suspect you're a good programmer - and your increasing productivity in writing good code will reduce your company's need to hire new programmers, and may allow the company to lay some off.
If Copilot writes a functioning program, debugged and ready to ship - in one day- that should allow multiple iterations of user improvements. But it's still pretty far away from that level.
I found it useful for adopting and integrating _new_ technology. I need to integrate third-party thing I know next to nothing about: generate an example with that matches roughly what I need to do. Show me how to change the things I care about without me needing to traverse tons of docs that presume background knowledge from other docs (or worse no docs and investigating implementation-code - or even worse wrong/bad docs). My time to come up to speed a get something going is reduced from a couple days to minutes. It doesn't really help with the whole system, but the "I could sorta want to do something like X, using toolset Y and further refine with details Z" is super powerful. Plenty of people use tools like stackoverflow when they get stuck ... but that still has latency of hours/days. If you get those questions resolved in seconds, its remove a lot of idle time. And that is value today ... in the future I imagine it will be able to do more on the architecture/ whole-project side of things.
It's very far away from that, to be honest. While it's great to help me keep from having to type and re-type typos and it's great for remembering the variable that should go in this logic, it's not all that great for full blocks of code beyond maybe a conditional block or a function that resembles one you've already worked on. I get your point about it increasing productivity to where some programmers get laid off. I don't know that it would really work that way but I will allow that it might, in quite some time.
Loved how you described the position in so much detail. I wrote something similar on my Substack - about whether AI will replace legal jobs, again with a resounding NO.
This fear mongering is as common as boom and bust cycles.
Thanks Suhas! I'd love to read your piece but it looks like it's behind a paywall.
Hi Tim, I have removed the paywall! Cheers. https://futureofthelaw.substack.com/p/will-ai-replace-legal-jobs
I didn't find the "robotics is hard" section very compelling. You give a few anecdotes about specific robotics projects that have progressed slowly, but you don't really explain the *why*.
All the newer tools (large GPU clusters, SGD, dropout, pre-training, transformers, relu units, etc. etc.) have all been applied to reinforcement learning problems, with some success. So why do these problems still appear so much harder?
I reckon it has something to do with the difficulty of producing an adequate physical model to perform training in a simulated setting, and the high cost of tuning the physical model (or training the whole network) on input from actual physical sensors.
I'm not saying it's hard to build robotics software (though it might be). I'm saying it's hard to build the actual robots. The human hand for example is an engineering marvel—it has incredible degrees of freedom, has highly sensitive sensors, and it can work for decades without need for repair or maintenance. I don't think we're anywhere close to figuring out how to build something with the same combination of strength, weight, sensitivity, and durability, to say nothing of its ability to self-repair after minor damage.
I suspect it may be impossible to build a robot with our self-maintenance abilities using materials like steel and plastic. That obviously doesn't mean we can't eventually build an ecosystem of robots that are able to build and repair one another, but I think it's going to take a long time to build up that ecosystem. In the shot-to-medium turn, every robot we build will mean new jobs for people to maintain and repair it when it breaks.
Yes robotics is hard. I worked as an automation engineer for automotive plants for a few years. Robots there are powerful but pretty dumb. You have to enclose them in cages so they don’t kill people. However if you want to reliably spot weld doors together you really can’t beat it. No human wants that job anyway.
I was a coach/mentor for our high school FIRST robotics team. The competition has the kids assemble and program a mobile robot to do some basic tasks. The task we had in 2018 was to pick up 2 foot square cubes and place them on a balancing scale. Even that well defined task proved to be quite difficult in the “fully autonomous” part of the competition.
The FIRST robotics competition is divided into 2 parts. The first 15 seconds is the “fully autonomous’ phase. In that phase your robot needs to run autonomously in the competition area and try and accomplish the assigned task. The rest of the completion has a human remote controlling the robot for 3 minutes. That is much more productive and fun. I can see the AI revolution taking that turn. Instead of replacing humans it will give them super powers like the Iron Man suit.
But what level of proficiency is needed to displace a given percentage of jobs? We don’t necessarily need robots to be as dexterous as humans in order to for example displace 30% of 60% of jobs, which is an 18% unemployment rate presuming demand is level. What if demand is collapsing due to an imploding global debt bubble and the apogee of the bankrupt paradigm of globalism?
Machines have been steadily displacing workers for 150 years, yet unemployment is currently near record lows because demand isn’t level. As the economy gets more efficient people consume more.
Demand isn’t level because the Fed is propping up a massive globalism bubble that is going to self-immolate and leave your mouth agape within this decade. Entitled Westerners do not want to work for peanuts. No surprises there, when the government is handing out fiscal debt spending like there is no tomorrow nor reckoning ever coming. The reckoning is ~2026ish in earnest.
In the early 1980s I was evaluating one of a number of "create code from English Language" products (I think it was called Rescue); older colleagues predicted then that I wouldn't have much of a software career before these new products took over ... I'm still creating software nearly 40 years later ...
You should read that:
https://medium.com/@avivolah/why-ai-robotics-will-be-able-to-do-everything-a-human-does-and-will-do-and-how-it-matters-c37d89af4a18
I don't feel like that piece addresses the argument I make in the "Human interaction as a luxury" section?
My comments on this important blog linked published also as a Note for my followers.
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51268748
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51278117
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51272051
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51276856
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51269678
https://www.understandingai.org/p/software-didnt-eat-the-world/comment/51269546