AI Will Not Replace You
But it will trap us in poverty unless we act.
Call it Automation Infrastructure
The current discussions about the AI hype and its social implications tend to underestimate a crucial point.
The real reason why companies and governments are currently investing resources at unprecedented levels into one technology is to build automation infrastructure. In fact, this might be a more accurate reading of the shortcut A.I.
While we get entertained with fascinating feats and exciting promises about what AI can or will be able to do, the underlying process is a boring, but devastating structural redesign of our economies.
If we just fear losing our jobs, we underestimate the impact of this. The drivers of this redesign are set on automating every single process that is commercially valuable. And this will render the vast majority of humans systemically obsolete.
If you want to get an idea of how being systemically obsolete looks like, take a look at the slums and refugee camps of today. You will not just face poverty, but also the complete lack of a chance to change your situation.
We are part of a giant experiment
The AI we are currently using as chatbots and productivity tools is not the AI that will transform our societies. In this current phase, the big AI Labs are distributing half-baked products to the public for essentially one purpose: data harvesting.
In the past year, I’ve been reading dozens of articles and blog posts by developers who point out that “vibe-coding doesn’t work” or that coding agents just create a costly mess. But pointing at the limitations of current tools misses the point of why we are offered these tools at subsidized pricing. Every time a developer repairs messed up code by AI, they provide valuable data for training future systems.
The same holds for basically all use cases. By using the imperfect tools, we help the big tech corporations to collect more data for training the “real AI” they are working on. Billions of dollars are invested into subsidizing our use of tools that are neither cost-efficient, nor reliable.
So what is the “real AI”?
The “real AI” that big tech is working on is Automation Infrastructure that works well enough to take over economic production at scale. For this, no AGI is needed (”Artificial General Intelligence” that can do any intellectual work we humans can do).
The real driver behind this is the corporate logic. AI is optimized for maximizing corporate growth rather than reflecting the cognitive traits and needs of human societies. Never mind the story telling about “AI alignment” or “benefiting humanity”.
The infinite demand for compute power
Let’s take a simple metaphor: I am a pretty mediocre dart player. If I throw a single dart, it will almost certainly miss the 50. But what if I have a system that allows me to throw 1,000 darts a second, instantly calculates how far off each throw was, and automatically corrects my aim for the next one?
I don’t need to possess human mastery, and I don’t need to get lucky. I simply need enough processing power to iterate and self-correct until the target is hit.
This is, simplified, how machine learning works. And now the same mechanism is applied to multi-agent workflows. One AI agent generates an output, another evaluates it against the goal, and they loop this process, recalibrating at machine speed, until the solution works. In such a system, imperfection is solved with raw compute power. Even if individual LLMs hallucinate and make mistakes, the process can be repeated until the desired result is achieved.
What’s still missing is scale and specification.
There’s currently not enough compute power on the planet to do all work humans do with AI systems and robots. The other part that is still missing is a testable definition of what it means to “hit the 50” for all kinds of human jobs and tasks. This is why the big data collection is happening. This is why the AI labs need your data, of you doing any type of task, in your job, at your home, in your private life. The more data we give them, the faster they can take everything from us.
Pay will converge to zero
The threat of job automation is real. If companies can automate most of their processes, workers and unions lose all their leverage.
We often hear things like: “There will be new types of jobs.” Or “You won’t be replaced by AI, but by someone using AI.” These points might be valid in the short run. But in the long run, there is absolutely no question that all commercially valuable activities will be automated.
AI is different to all technological transformations humanity has gone through in the past. For the first time, the technology can learn by itself, improve itself, and adapt to any challenge. We’re already in the phase of visual/sensual machine learning. Robots learn from interaction with the real environment. Once one robot learns something, it can be copied to all robots. In real time. This is different to how humans learn. We have no chance competing with that.
Turning the whole world into “the poor”
What the “job replacement hypothesis” is not taking into account is that there will be still use for human labor. It will just be paid much less. Instead of mass unemployment, the real effect is the collapse of human bargaining power. Again, we can see this playbook already in the exploitation of people in the poorest countries today. When people have no options, they are forced to work in slavery-like conditions. Labor standards are a luxury for countries where people have options.
When humans are cheaper than robots
Most of the work that people do in the coltan mines in the Democratic Republic of the Congo could theoretically be done by sophisticated machines and robots. But it is much cheaper to let it be done by humans who work extreme hours for barely any pay. And there are more valuable use cases for today’s advanced machines and robots, like assembling electric vehicles, or computer chips, or building more robots.
Essentially, the fate we are working towards is pushing us all into the conditions of the people who mine the minerals for our electric cars, laptops, smart phones. We will never run out of work, because there will always be some work for which using robots would be comparatively too expensive.
The “consumer argument” does not work
People often ask: if no-one earns money anymore, who will then be able to buy the products? But this demand side argument underestimates the creativity of the corporate logic to find and create further avenues for market growth.
In the phase we are entering, the transition into a fully automated economy creates enormous demand for infrastructure building. Data center construction, robot production, resource harvesting all are economic activities that can be scaled almost infinitely. Already, corporations like SpaceX and Nvidia are working on plans to expand data-centers into space.
And even if wealth concentration reaches an extreme where only a handful of human “consumers” are left, the demand for luxury goods, status symbols, and power instruments is literally limitless. It might be more profitable for the fully automated economy to focus on space expansion, robot armies, and hyper-secure entertainment islands for a few, than on mass-markets of “inefficient” consumers (us).
Once more, the dynamics in very poor countries, with a tiny elite living in excess and the vast majority living in extreme poverty, offer a peek into how this could play out.
The end of the social contract
This risk scenario is exacerbated by automated weapon systems. The idea that weapons can find their targets and carry out an attack without human intervention is not just a moral concern. The risk also cannot be reduced to them potentially running out of control. The greatest harm they bring might be the fact that they render the need for human soldiers obsolete.
This risks undermining the very social contract that all political systems in history have been based on. Even authoritarian political systems like China still function on the premise that the general population is relevant. Even without elections and legal accountability, the authorities still need to provide some level of a functioning state to recruit soldiers they can use to control the rest.
Once autonomous weapons reach sufficient capability, this premise dissolves. The structural need for human populations, as workers, as soldiers, as a tax base, disappears. This would not just be the end of democracy, but of any political system that takes the country’s people into account.
No protests or social unrest or military coup could change any of this. Whoever controls the autonomous weapon systems controls everything, even if they are not in the role of the official government.
Being overwhelmed is part of the risk
AI development is running at an extreme pace and news about new tools and breakthroughs is coming every day. The debates range from global geopolitics to financial bubbles to fantastic speculation about potential benefits from futuristic technologies. It is easy to feel overwhelmed and thrown into passivity.
The drivers of the AI hype are taking advantage of the confusion and uncertainty that many of us currently feel. All this lands in societies that have already gone through decades of commercialization and overstimulation with social media and ever faster technologies.
In fact, the AI hype accelerates the very pressures that keep us disconnected from ourselves. The more data we give to the big tech corporations, the better they can capture our attention. And the general sense of uncertainty drives us further into insincerity. At the same time, our use of AI tools as companions or consultants for our own mental well-being exposes us to new levels of manipulation.
It is all the more important to take care of our own mental balance. The more we are able to step back and see the patterns, the less we are passive victims of this process of disenfranchisement.
The window is still open
There is no question that we are living in a time where the global power structures will be redefined. The questions that are still open are how quickly this will happen and what the new power structures will look like.
If we take action now, we still have a chance to play a role in shaping our future. In fact, we have more power than it might seem when we just follow the news.
Consider where the actual power in a corporation like Google resides today. If we follow the traces of power to where action creates reality, we can see that it is not with the CEO, or the board, or even with the shareholders. The real power is with the people who do the everyday work.
If they stopped working now, no value would remain in the company. CEOs and shareholders hold power because staff follows their instructions, bound by contracts that can individually be enforced through state authorities. But no contract could hold, if everyone acted collectively.
This is just the example of one company, but the same holds for most corporations. This structural reality gives us humans more leverage than most of us realize. At least for now.
Reclaiming control
Once we accept the reality that we are currently in a fundamental transition of the basic conditions of all of our lives, and that the current trajectory could trap us all in absolute poverty, there is a real path to reclaim control for us humans.
Of course, this needs to happen step by step. Most of us are still busy making a living in the current conditions.
This is why I personally decided to continue investing my time in working on CosyAI. If we manage to build a global coalition of organizations and people who want to take AI and automation infrastructure into cooperative ownership, we can offer a path forward that doesn’t involve giving up our current lives. What is crucial is that we succeed in making sure that we, rather than corporations and captured governments, can control how AI is developed and used and that the benefits are equally distributed.
The window is still open, but the longer we wait, the harder it will get.
In the next article, I will look at specific things that we can do right now. While we still can.